Part 2 book “Medical image analysis and informatics - Computer-aided diagnosis and therapy” has contents: Computer-Aided diagnosis of breast cancer with tomosynthesis imaging, computer-aided diagnosis of spinal abnormalities, health informatics for research applications of CAD,… and other contents.
11 Computer-Aided Diagnosis of Breast Cancer with Tomosynthesis Imaging Heang-Ping Chan Ravi K Samala Lubomir M Hadjiiski Jun Wei 11.1 Introduction 241 11.2 Imaging Characteristics of Breast Tomosynthesis .243 11.3 Computer-Aided Detection in DBT .249 Computer-Aided Detection of Microcalcifications • Computer-Aided Detection of Masses 11.4 Summary 260 Acknowledgments 261 References 261 11.1 Introduction Breast cancer is the most prevalent cancer in women worldwide, and the second-most common cause of cancer death in many countries, including the United States [1,2] Mammography has been found to be effective in reducing breast cancer mortality by a number of cohort and case-control studies [1,3], although the cost of over-diagnosis has been a topic of controversy and study in recent years A major limitation of screening mammography is the low sensitivity in dense breasts [4,5] due to the reduced conspicuity of lesions obscured by overlapping dense fibroglandular tissue Another limitation is the high recall rate Many of these recalls are caused by overlapping tissue that resembles a lesion and requires diagnostic workup Finally, many malignant and benign lesions have similar mammographic appearance and cannot be distinguished even by further diagnostic workup The positive predictive value of recommended biopsies ranges from only about 15%– 30% [6] Recalls and benign biopsies not only cause patient anxiety, but also increased healthcare costs Digital breast tomosynthesis (DBT) is a new imaging modality that has been introduced into clinical use in the past few years In the United States, three commercial systems have been approved by the Food and Drug Administration since 2011 DBT is a limited-angle tomographic technique in which a small number (e.g., 9– 25) of projection images of the compressed breast are acquired over a small angular range (e.g., 11° – 60° ) With proper reconstruction, a stack of reconstructed image slices covering the breast volume can be obtained DBT provides high spatial resolution on slices reconstructed parallel (or at small angles) to the detector plane but with low resolution in the depth direction DBT reduces the overlap of fibroglandular tissue that can obscure cancerous lesions on mammograms, thereby alleviating a major problem that limits the sensitivity of breast cancer detection in mammography 241 242 Medical Image Analysis and Informatics: Computer-Aided Diagnosis and Therapy A number of clinical trials have been conducted to evaluate the addition of DBT as an adjunct to the digital mammogram (DM), that is, combining DBT and DM (DBT + DM), in comparison to DM alone in screening settings [7– 10] Other investigators compared the cancer detection and recall rates in screening populations before and after the DBT + DM mode was introduced into their clinical practice [11,12] All these studies found significant improvement in cancer detection and reduction in recalls Lang et al. [13] compared two-view DM alone and one-view DBT alone in a screening population and found that one-view DBT significantly improved the cancer detection rate, but increased the recall rate while maintaining the same positive predictive value Although the DBT + DM mode could achieve increased cancer detection rate and reduced recall rate compared to DM alone, it doubles the radiation dose to the screening population Recently efforts are being made to synthesize a mammogram-like image (SM) from the DBT to obviate the need for the DM Skaane et al. [14] showed that a newer version of SM (C-view) combined with DBT was not significantly different from the DBT+DM mode in a large screening study and concluded that DBT + SM was acceptable for routine clinical use Gilbert et al. [15] compared the DBT + DM mode and DBT + SM with DM alone in a screen-recalled population and observed significant increase in specificity and sensitivity for invasive cancers, but marginal increase in sensitivity for all cancers using the DBT + DM mode; however, DBT + SM increased specificity significantly but no significant increase in sensitivity for all cancers Although the studies found that DBT increased the detectability of breast cancer and reduces recall rates compared to DM, most studies did not analyze the detection of non-calcified lesions and the detection of microcalcifications separately In a few studies that reported the performance of DBT in the detection of microcalcifications, the results were not as consistent In an early study with 98 subjects, Poplack et al. [16] found that the recall rate could be reduced by 40% with the addition of DBT to DM, but the conspicuity of microcalcifications were inferior in of the 14 cases Gur et al. [17] compared DM alone to DBT alone and DM + DBT They found that DM+DBT could reduce recall rate by 30%; however, three benign microcalcification clusters that were seen in DM were not visible in DBT, whereas six benign masses not seen in DM were seen in DBT Wallis et al. [18] found that two-view DBT provided significantly higher detection for both masses and microcalcifications than DM Kopans et al. [19] also reported that the clarity of calcifications in DBT acquired with a GE prototype system was better than or comparable to that in DM in 92% of 119 cases with relevant calcifications Andersson et al. [20] found that the visibility of calcifications in DBT were comparable to that in DM for the 13 cancer cases with calcifications in their study However, Spangler et al. [21] found that the sensitivity and specificity of calcification detection in DM were higher than those in DBT in a dataset with 20 malignant and 40 benign calcification cases Various methods have been studied to improve the detection of microcalcifications in DBT, including the use of DM in combination with one-view or two-view DBT [18,22– 28], the use of a synthesized DM-like image from DBT to replace the directly acquired DM [14,15,29], development of computeraided detection methods for DBT [30– 41], and the enhancement of the visibility of microcalcifications by improving reconstruction and image processing methods [42– 46] Regardless of the method of implementing DBT (combo DBT + DM, DBT + SM, replacing one or both DM views with DBT), one of the major concerns of integrating DBT into clinical practice is the change in workflow A DBT volume contains a large number of reconstructed slices that need to be read by radiologist Even at 1-mm slice thickness, the number of slices per view of the breast will range from about 30 to over 80 Although the correlation between adjacent slices and the lesscomplex background make it much more efficient in reading each slice than reading a regular mammogram, studies showed that the time required for interpretation of a DBT + DM examination was about 50%– 100% longer than that for reading DM alone [7,18,47,48] If the caseload for a radiologist has to be maintained at essentially the same level as DM due to the limited resources available for screening, radiologists inevitably will tend to speed up the reading The DBT + DM or DBT + SM approach allows radiologists to search for microcalcifications in the two-dimensional (2D) DM or Computer-Aided Diagnosis of Breast Cancer with Tomosynthesis Imaging 243 SM, but the search for subtle microcalcifications even in 2D DM is known to be a challenging task; the additional blur and noise in the SM synthesized from DBT may make it more challenging Softtissue lesions such as masses and architectural distortion will be more visible in DBT slices, but it requires scrolling through the hundreds of slices in the 4-view screening examination The chance for oversight of subtle lesions in the large search space may not be negligible under the time constraint Detection of microcalcifications in DBT is especially important if DBT would replace DM for screening because no other imaging modalities can detect calcification as effectively as DM and calcification is an important sign of early stage breast cancer Computer-aided detection (CAD), therefore, is expected to play a similarly important, if not more important, role for DBT as for DM in clinical practice 11.2 Imaging Characteristics of Breast Tomosynthesis To design effective computer vision techniques for CAD, it is important to understand the imaging characteristics of DBT The image acquisition geometry of a typical DBT system is shown in Figure 11.1 The x-ray system is basically a digital mammography system, except that the x-ray source is rotated along an arc or moved linearly over a limited angle range and takes a small number of low-dose mammograms along the way However, different DBT manufacturers may have different designs for the image acquisition process For example, the detector can be stationary or may be rotated around the pivot point in the opposite direction while the x-ray source is moved to different locations for acquisition of the projection views (PVs) The x-ray source can be moved continuously while the PVs are taken with short x-ray pulses to minimize the blurring by the source motion, or moved in a velocity mode or a step-and-shoot mode such that the x-ray source is stationary during acquisition of the projections Another design that X-ray source X Y Z Compressed breast volume Chest wall Projection image Digital detector FIGURE 1 1.1 Geometry and image acquisition of a typical breast tomosynthesis system The legend shows a coordinate system being referred to in the other figures in this chapter 244 Medical Image Analysis and Informatics: Computer-Aided Diagnosis and Therapy uses an array of stationary x-ray sources placed at proper locations to take PVs at different angles is also being developed DBT is not a true three-dimensional (3D) imaging modality because image acquisition is limited to a small angular range (e.g., 11° – 60° ) compared to computed tomography (CT) of over 180° The information of the object in the depth direction that is void of projections is insufficient to permit accurate reconstruction of the details The depth resolution is mainly determined by the tomographic scan angle: the larger the angle, the higher the depth resolution In addition, the small number of projections taken within the imaging arc with relatively large sampling intervals and the distribution of the PVs (e.g., uniform and non-uniform) also affect the reconstructed image quality [49,50] However, the spatial resolution on the reconstructed DBT planes parallel to the detector is almost as high as that of digital mammograms, which allows DBT to maintain the spatial details of subtle breast lesions such as microcalcifications and small spiculated lesions similar to mammograms, while gaining the advantage of separating the images of the overlapping tissue into thinner layers Therefore, a tomosynthesis volume is different from a CT volume that can provide nearly isotropic spatial resolution and can be viewed at any cross-sectional planes An example of a DBT and a mammogram of the same breast in mediolateral oblique (MLO) view is shown in Figure 11.2 The DBT was imaged with an experimental system that acquires 21 projections over a 60° tomographic angle at 3° angular intervals The system uses a CsI/a:Si flat panel detector with a 0.1 mm × 0.1 mm pixel pitch The simultaneous algebraic reconstruction technique (SART) was used for the reconstruction at 0.1 mm × 0.1 mm pixel size and mm slice interval The breast contains an invasive ductal carcinoma manifested as a mass with calcifications The DBT slice shows the irregular-shaped mass and its extended spiculations clearly, while the same mass on the mammogram appears as an illdefined density, similar to the adjacent normal breast tissue Figure 11.3 shows the same DBT volume in three perpendicular planes The DBT slice parallel to the detector plane (x -y plane) has high spatial DBT slice Mammogram FIGURE 1 1.2 An example of a breast with an invasive ductal carcinoma (white arrow) manifested as a mass with calcification imaged on mediolateral oblique (MLO) view Left: DBT slice intersecting the breast cancer reconstructed from a DBT scan with 60° tomographic angle and 21 projections Right: mammogram Computer-Aided Diagnosis of Breast Cancer with Tomosynthesis Imaging 245 x-z plane Scan direction (y) y-z plane x-y plane FIGURE 1 1.3 Reconstructed DBT volume of the same breast shown in Figure 1 1.2 displayed in three perpendicular planes The simultaneous algebraic reconstruction technique (SART) was used for the reconstruction at 1-mm slice interval The invasive ductal carcinoma is indicated by a white arrow The image plane parallel to the detector plane (x -y plane) has high spatial resolution, similar to that of a mammogram The other two planes (x -z and y -z planes) that are perpendicular to the detector plane have very low resolution The horizontal and vertical lines on the DBT slice indicate the locations where the x -y plane and the y -z plane are relative to the x -y plane resolution, similar to that of a mammogram, whereas the other two perpendicular planes are dominated by the angular patterns of the x-ray paths without clear structures that resemble breast tissue Figure 11.4 shows an example of a DBT volume with a cluster of microcalcifications from a high nuclear grade ductal carcinoma in situ (DCIS) The inter-plane artifacts can be seen clearly as the extension of the long bright shadows of a dense calcification along the x-ray paths on the x -z plane and the y -z plane The shape of an object in DBT is, therefore, distorted along the depth direction and casts a shadow on the adjacent slices It is important to take into consideration these imaging properties during feature extraction for image analysis in DBT It is known that the image quality of DBT depends on the image acquisition parameters, including the tomographic scan angle, the angular increment, and the number of projections, in addition to other factors that affect the image quality of x-ray imaging systems The visibility of breast lesions also depends on the physical properties such as size and contrast of the lesions, as well as the structured noise in the images The best combination of the DBT acquisition parameters for each type of lesions has been an area of interest for research and development in DBT A number of simulation and modeling studies [51– 55] or experimental evaluations [49,50,56– 59] have been conducted to examine the dependence of image quality measures on DBT acquisition parameters In the studies by Zhang et al. [49] and Lu et al. [50], DBT scans of phantoms acquired at 60° angle and 3° increments with a total of 21 PVs were used They selected six subsets of 11 PVs from the original DBT scans to simulate DBT acquired with different tomographic angles and uniform or non-uniform angular increments The contrast-to-noise ratio (CNR), the full-width-at-half-maximum (FWHM), and the artifact spread function (ASF) of calcification-like and mass-like objects in the reconstructed DBT volumes were calculated to estimate the visibility of the objects on the DBT slices, the spatial blur on the x -y plane and along the z -direction, respectively The results showed that DBT acquired with a wide scan angle or, for a fixed scan angle, having a large fraction of PVs at large angles was superior to those acquired with a 246 Medical Image Analysis and Informatics: Computer-Aided Diagnosis and Therapy x-z plane y-z plane Scan direction (y) x-y plane (DBT slice) FIGURE 1 1.4 Reconstructed DBT volume of a breast with a cluster of microcalcifications in a high grade ductal carcinoma in situ (DCIS) displayed in three perpendicular planes SART was used for the reconstruction at 1-mmthick slice interval The x-ray source moved in the y-direction The image plane parallel to the detector plane (x -y plane) has high spatial resolution, similar to that of a digital mammogram The other two planes (x -z and y -z planes) that are perpendicular to the detector plane have very low resolution The white arrow points to the same dense calcification that causes inter-plane artifacts extending several mm along the depth (z) direction on the x -z plane and the y -z plane (Reprinted from Chan H-P., Computer Aided Detection and Diagnosis in Medical Imaging , CRC Press, Boca Raton, FL, 2015 ) narrower scan angle, as measured by the ASF in the z -direction On the x -y planes, the effect of PV distributions on spatial blur depended on the directions In the x-ray source scan direction, the PV distributions with a narrow scan angle or a large fraction of PVs at small angles had smaller FWHM, that is, less spatial blur In the direction perpendicular to the scan direction, the difference in the spatial blur among the different PV distributions was negligibly small In addition, for small objects such as subtle microcalcifications, PV distributions with a narrow scan angle or a large fraction of PVs at small angles yielded higher CNR than those with a wide scan angle Recently, Park et al. [60] conducted experimental studies to evaluate the effects of variable PV distribution and variable angular dose distributions in DBT acquisition on the reconstructed image quality of microcalcifications in breast phantom and observed similar results Chan et al. [58] and Goodsitt et al. [59] further investigated the impact of the imaging parameters on the image quality of signals in DBT by observer performance studies using an experimental DBT system that allows acquisition of projections at variable scan angles, angular increments and number of PVs One observer performance study [58] evaluated the detectability of simulated microcalcifications in DBT of heterogeneous breast phantoms acquired at seven acquisition geometries, using different combinations of scan angle and uniform or non-uniform angular intervals Another observer preference study [59] compared the visual quality of low-contrast objects for 12 different acquisition geometries These studies showed that a large tomographic angle was better for reducing overlapping tissue and improving the detectability of low-contrast objects such as soft-tissue lesions, whereas narrow tomographic angles provided higher detectability of microcalcifications Figure 11.5 shows DBT volumes of the same breast shown in Figures 11.2 and 11.3 reconstructed at three combinations of tomographic angle and angular intervals The original DBT was acquired with 60° , 3° angular increments and 21 projection views (PVs) Two other geometries, wide angle (60° , 6° , 11 PVs) and narrow angle (30° , 3° , 11 PVs), were simulated by reconstruction with a subset of 11 PVs Although the x-ray dose was reduced by about half, and the noise was higher for the reconstructions with the subsets of 11 PVs, the main effects of acquisition geometry on the appearance of the mass, microcalcifications and tissue 247 Computer-Aided Diagnosis of Breast Cancer with Tomosynthesis Imaging 11 PVs, 60° 11 PVs, 30° 60 50 50 50 40 40 40 30 20 –30° Z (cm) 60 Z (cm) Z (cm) 21 PVs, 60° 60 30 20 30° –30° 20 30° 10 10 –30 –20 –10 10 Y (cm) 20 30 15° –15° 10 0 30 –30 –20 –10 10 Y (cm) 20 30 –30 –20 –10 10 Y (cm) 20 30 (a) (b) (c) iteration iterations iterations FIGURE 1 1.5 Comparison of reconstructed DBT images for three acquisition geometries using SART The original DBT was acquired with 60° , 3° angular increments and 21 projection views (PVs) The other two geometries, middle column: wide angle (60° , 6° , 11 PVs), and right column: narrow angle (30° , 3° , 11 PVs), were simulated by reconstruction with a subset of 11 PVs Although the x-ray dose is reduced by about half and the noise will be higher for the reconstructions with the subsets, the effects of acquisition geometry on the appearance of the mass, microcalcifications, and tissue texture patterns are demonstrated The number of iterations for the DBT by the 11PV-reconstructions was doubled so that the number of PV updates is approximately equal to that of the 21PV-reconstruction (a) PV distributions of three geometries (b) DBT slice (x -y plane) intersecting an invasive ductal carcinoma (white box) (c) and (d) The enlarged region of interest showing the spiculated mass with calcifications The mass shows higher contrast in the wide-angle DBT, whereas the calcifications are sharper and higher contrast in the narrow-angle DBT Both the signal and noise increase as the number of iterations increases 248 Medical Image Analysis and Informatics: Computer-Aided Diagnosis and Therapy (d) iterations 10 iterations 10 iterations x-z plane x-z plane x-z plane (e) y-z plane y-z plane y-z plane (f ) FIGURE 1 1.5 (CONTINUED) (e) and (f) The inter-plane artifacts of the narrow-angle DBT extend longer than those of the wide-angle DBT, indicating that the wide-angle DBT has better depth resolution and less overlapping tissue shadows than the narrow-angle DBT texture patterns can be seen by comparison of the images The DBT was reconstructed with SART; the number of iterations for the DBT by the 11 PV-reconstructions was doubled so that the number of PV updates was approximately equal to that of the 21 PV-reconstruction to reduce the impact of fewer updates on the subset reconstruction It is shown that the spiculated mass and the fibroglandular tissue have higher contrast in the wide-angle (60° ) DBT than those in the narrow-angle (30° ) DBT; however, the calcifications are sharper in the narrow-angle DBT Both the signal and noise increase as the number of iterations increases Computer-Aided Diagnosis of Breast Cancer with Tomosynthesis Imaging 249 On the cross-sectional images perpendicular to the detection plane, the image texture is dominated by the patterns of x-ray paths and the inter-plane artifacts of the narrow-angle DBT extend longer than those of the wide-angle DBT These examples illustrate that wide-angle DBT has better depth resolution and less overlapping tissue shadows than the narrow-angle DBT, which results in DBT slices having a background with less fibrous textures and reduced tissue overlap, as evident by comparing the DBT slices in the second row In addition to the image characteristics of various types of lesions and their different dependences on the DBT acquisition geometry, the design of a DBT system often has to take into consideration the trade-offs among many other factors, such as the detector efficiency, the x-ray source output, the readout speed and lag of the detector, the scanning and breast compression time, and the mechanical stability and precision, while under the constraint of maintaining low radiation dose to the patient The optimal design of a DBT system that can balance the image quality requirements of various types of lesions at the lowest possible radiation dose is still a topic of continued investigation A number of reconstruction methods have been applied to DBT reconstruction, including shiftand-add, tuned aperture computed tomography (TACT), maximum likelihood-convex (ML-convex) algorithm, matrix inversion (MITS), filtered back projection (FBP) and simultaneous algebraic reconstruction technique (SART) [61– 65] Reconstruction methods have a strong impact on image quality of DBT Studies to improve the reconstruction methods and artifact reduction techniques are on-going [42,44,66,67] Reconstruction methods specifically designed to enhance microcalcifications and reduce noise are also under investigation [44, 68– 70] DBT images are usually reconstructed in slices parallel to the detector plane The spatial resolution on the reconstructed DBT slices can approach that of the digital detector if the geometry of the scanning system is accurately known and patient motion is kept at a minimum However, some degree of blurring is inevitable due to the reconstruction from multiple PVs with different x-ray incident angles and oblique incidence of the x-ray beam to the detector, especially at large projection angles [71] Super-resolution has been observed in DBT when reconstruction is performed with finer grids [46,72,73] Because of the lack of PVs at large projection angles, the spatial resolution in the direction perpendicular to the detector plane (the depth or z-direction) is poor The depth resolution is mainly determined by the tomographic angle: the larger the angle, the higher the depth resolution and the less the inter-plane artifact but with a trade-off of greater blurring on the DBT plane due to oblique intersection of the x-ray paths with a reconstructed slice of finite thickness This blurring may be reduced by reconstruction with an adaptive grid approach along the depth direction for small objects such as microcalcifications [74] Regardless of the reconstruction methods, tomosynthesis cannot provide true 3D information due to the lack of sampling over a wide angular range 11.3 Computer-Aided Detection in DBT DBT is composed of a number of low-dose DMs taken at slightly different projection angles The PVs, together with the acquisition geometry, contain all the available information for signal detection in DBT However, the individual PVs are noisy due to the low-dose acquisition A DBT volume can be reconstructed from the PVs by an appropriate technique, which can enhance the signal and reduce noise by combining the information from the multiple projections If both the set of PVs and the reconstructed DBT are available, CAD methods can be developed by combining the information from both in many different ways One approach is to use the set of PVs as input and combine the information from the PVs in the process Another approach is to use the reconstructed DBT volume (slices) as input and analyze images as a 3D volume or 2D slices A third approach is to use both sets of images as input and combines the information at different stages of detection Although the PVs and DBT volume basically contain the same information, the computer-vision techniques designed for the different sets of images may utilize the information differently Combining the information extracted from the different forms of images derived from the original PV images may improve signal detection or characterization 250 Medical Image Analysis and Informatics: Computer-Aided Diagnosis and Therapy The recent development of methods for generating a 2D synthetic mammogram from the DBT images leads to an additional option of lesion detection, namely, detection in the synthetic mammogram, which may be combined with the approaches described above However, it should be noted that some synthetic mammogram generation methods rely on detecting potential lesions with CAD to enhance the conspicuity of the lesions on the synthetic mammograms [75] The sensitivity of detecting lesions in this type of synthetic mammograms will depend on the sensitivity and false-positive rate of the CAD methods used in the generation of the synthetic mammogram On the other hand, if the synthetic mammogram is generated from the DBT without using CAD, the image quality and lesion detectability is most likely poorer than a DM because all overlapping tissue remains in the synthetic mammogram and additional blurring may result from the multiple-projection reconstruction and the limited depth resolution of the reconstructed volume 11.3.1 Computer-Aided Detection of Microcalcifications Detection of subtle microcalcifications in DBT by human or computer vision is challenging because of the large search space and the noisy background CAD methods for detection of microcalcifications in the projection views (PVs), the reconstructed slices or the reconstructed volume have been studied Peters et al. [76] detected calcifications on a small set of DBT A band-pass, filter-based, wavelet kernel was used to separate the potential calcification candidates from the background on the PVs A feature map was generated for each PV image, and the correspondence between 2D and 3D locations determined by the DBT acquisition geometry was used as a criterion to identify the calcifications Park et al. [77] applied a 2D CAD algorithm developed for digitized screen-film mammograms (SFM) to the PV and the reconstructed DBT slices Reiser et al. [36] developed an algorithm to detect microcalcifications in PV images to avoid the dependence of the CAD performance on the reconstruction algorithm van Schie et al. [37] estimated a non-uniform noise model from each individual DBT-reconstructed volume which was used for normalization of the local contrast feature Potential microcalcifications were detected by thresholding the local contrast feature, and the microcalcification candidates within a 5 mm radius were grouped to form microcalcification clusters The detection strategies developed by our research laboratory are described below 11.3.1.1 Microcalcification Detection in DBT Volume A CAD system generally consists of several major stages: preprocessing for signal enhancement, prescreening for candidate signals, feature extraction and analysis for false positive reduction and final decision for identifying detected signals A number of preprocessing methods have been investigated to improve the detectability of microcalcifications in DBT Sahiner et al. [31] developed a CAD system for detection of microcalcifications, as shown in Figure 11.6 Two parallel processing methods are designed to identify microcalcification candidates and cluster seed candidates For identifying microcalcification candidates, a 2D contrast-to-noise ratio (CNR) enhancement filter is applied to the DBT slices to enhance potential microcalcifications and reduce the low frequency background Adaptive thresholding and region growing are then applied to the CNR-enhanced volume to segment the individual microcalcification candidates For identifying cluster seed candidates, 3D multiscale filtering is applied to the DBT volume, and the eigenvalues of Hessian matrices are calculated at each voxel Multiscale calcification response representing the intensity, size and shape information are then derived from the Hessian eigenvalues, which is further weighted by the CNR-enhanced volume voxel by voxel, resulting in an enhancement-modulated calcification response (EMCR) volume With adaptive thresholding and region growing, potential calcifications are segmented from the EMCR volume and a set of topranked candidates are used as cluster seeds A dynamic clustering process then groups the individual microcalcifications into clusters using the cluster seeds as the starting point and a distance criterion to determine cluster membership The cluster candidates identified in the clustering process will undergo feature analysis and the clusters that not satisfy the criteria are excluded as false positives (FPs) The 504 Medical Image Analysis and Informatics: Computer-Aided Diagnosis and Therapy Abramson RG, Burton KR, Yu JP, Scalzetti EM, Yankeelov TE, Rosenkrantz AB, Mendiratta-Lala M, Bartholmai BJ, Ganeshan D, Lenchik L, Subramaniam RM Methods and challenges in quantitative imaging biomarker development Acad Radiol 2015 Jan;22(1):25– 32 Reboussin D, Espeland MA The science of web-based clinical trial management Clin Trials 2005;2(1):1– 2 Leroux H, McBride S, Gibson S On selecting a clinical trial management system for large scale, multi-centre, multi-modal clinical research study Stud Health Technol Inform 2011;168:89– 95 Deserno TM, Deserno V, Haak D, Kabino K Digital Imaging and electronic data capture in multicenter clinical trials Stud Health Technol Inform 2015;216:930 Rubinstein YR, Groft SC, Bartek R, Brown K, Christensen RA, Collier E, Farber A et al Creating a global rare disease patient registry linked to a rare diseases biorepository database: Rare DiseaseHUB (RD-HUB) Contemp Clin Trials 31:394404, 2010 Deserno TM, Haak D, Brandenburg V, Deserno V, Classen C, Specht P Integrated image data and medical record management for rare disease registries: A general framework and its instantiation to the German calciphylaxis registry J Digit Imaging 2014; 27(3): 702– 713 Stausberg J, Altmann A, Antony G, Drepper J, Sax U, Schiitt A Register for networked medical research in Germany Appl Clin Inform 1(4):408418, 2010 Drepper J, Semler SC (ed) IT infrastructure in patient-centered research State of the art and required actions (German) Technical Report IT-Reviewing Board of TMF Akademische Verlagsgesellschaft AKA GmbH, Berlin 2013 (ISBN 978-389838-690-6) 10 Deserno TM, Deserno V, Legewie V, Schafhausen J, Eisert A, Schmidt-Kotsas A, Kirstein S, Willems J, Spitzer K, Schulz JB IT support for translational management of clinical trials using the Google web toolkit GMS Publishing House 2011; 56: 023 [in German] 11 Haak D, Samsel C, Gehlen J, Jonas S, Deserno TM Simplifying electronic data capture in clinical trials: Workflow embedded image and biosignal file integration and analysis via web services J Digit Imaging 2014; 27(5): 571– 80 12 Haak D, Dovermann J, Kramer C, Merkelbach K, Deserno TM Data recording in clinical trials by ODM supporting tablets and smartphones Proceedings 59th GMDS Annual Conference in Gttingen , 2014, DocAbstr 210 13 Haak D, Gehlen J, Jonas S, Deserno TM OC ToGo Bed site image integration into OpenClinica with mobile devices Proceedings SPIE 2014; 9039: 091-6 14 Haak D, Page CE, Reinartz S, Krger T, Deserno TM DICOM for clinical research: PACS-integrated electronic data capture in multi-center trials J Digit Imaging 2015; 28(5): 558– 6 15 Deserno TM, Haak D, Samsel C, Gehlen J, Kabino K Integration image management and analysis into OpenClinica using web services Proceedings SPIE 2013; 8674: 0F1-10 16 Jose A, Haak D, Jonas SM, Brandenburg V, Deserno TM Human wound photogrammetry with low-cost hardware based on automatic calibration of geometry and color Proceedings of SPIE 2015; 9414: 3J1-8 17 Park M, Brocklehurst K, Collins RT, Liu Y Deformed lattice detection in real- world images using mean-shift belief propagation IEEE Trans Pattern Anal Mach Intell 2009; 31(10): 1804– 16 18 Mitrani RD, Myerburg RJ Ten advances defining sudden cardiac death Trends Cardiovasc Med 2015 [Epub ahead of print] 19 Sartor M, Jonas S, Wartzek T, Leonhardt S, Wanner C, Marx N, Deserno TM Non-linear time normalization in long term ECG: Generation of unique pseudo images from multi-lead ECG In: Deserno TM, Handels H, Meinzer HP, Tolxdorff T (ed) Image Processing for Medical Applications Springer, Berlin 2014; 300– 5 [in German] 20 Deserno TM, Marx N Computational electrocardiography Revisiting Holter ECG monitoring Methods Inf Med 2016; 55(4): 305– 11 Concluding Remarks Paulo Mazzoncini de Azevedo-Marques Arianna Mencattini Marcello Salmeri Rangaraj Mandayam Rangayyan In the preparation of this book, our aim was to present a wide range of topics and applications to demonstrate the impressive impact of CAD and related fields of study on health care of the entire human body Our wish is to motivate and attract more researchers and innovators to the related fields of research and development to continue to address the current challenges and lead to further advances We believe the chapters in this book set the stage for these purposes Reiche et al have demonstrated, in Chapter 1, the robustness of their segmentation method and investigated quantitative shape features to discriminate between two classes of brain lesions in FLAIR MR images using a novel and exploratory noise analysis approach In Chapter 2, Hatanaka and Fujita have shown that analysis of retinal fundus images is effective in the diagnosis of not only diseases of the eye, such as glaucoma, diabetic retinopathy, and age-related macular degeneration, but also systemic hypertension Oloumi et al have demonstrated, in Chapter 3, that a measure of vascular tortuosity, derived by applying image processing methods to retinal fundus images, can facilitate CAD of retinopathy in premature infants with high accuracy Roychowdhury et al., in Chapter 4, have shown how image processing methods may be applied to measure sub‑retinal layer thickness distributions in OCT images The thickness maps can aid analysis of disorganization in sub‑retinal layers and assessment of diabetic macular edema Chapter 5 by Muramatsu et al presented CAD methods for analysis of dental panoramic radiographs for screening and efficient diagnosis of oral as well as systemic diseases, such as osteoporosis In Chapter 6, Pé rez-Carrasco et al discussed the problem of burn diagnosis in clinical practice They highlighted the importance of using CAD methods to derive an accurate initial assessment, which is crucial to the evolution and response to treatment of the burn wound Gutierrez et al demonstrated, in Chapter 7, the role of cardiac image segmentation in a wide range of applications, including quantification of volume, localization of pathology, CAD, and image-guided interventions In Chapter 8, Pezeshk et al have described methods to augment databases for training and testing CAD systems by inserting lesions or tumors from one image into another The variations in the 505 506 Concluding Remarks replicated lesions provided by various transformations facilitate realistic simulation of additional cases and lead to improved training of CAD systems Koenigkam Santos and Weinheimer presented, in Chapter 9, clinical and technical aspects related to CAD and QCT in the evaluation of diffuse lung diseases, focusing on emphysema, airway diseases such as asthma, and cystic fibrosis Chapter 10 by Mencattini et al demonstrated that CAD techniques can be designed to detect bilateral asymmetry in mammograms Such an approach could facilitate detection of breast cancer even in the absence of well-recognized signs such as masses and calcifications In Chapter 11, Chan et al introduced the DBT imaging technique and its applications in breast cancer detection The authors compared, in detail, CAD techniques for DBT and standard digital mammography and described their potential capabilities and limitations in the detection of signs of breast cancer, such as calcifications and masses Nogueira-Barbosa and Azevedo-Marques presented, in Chapter 12, a review of CAD methods for spinal abnormalities, with focus on vertebral body fracture and intervertebral disc degeneration They indicated that MRI is currently the preferred option for evaluation of spinal abnormalities Chapter 13 by Yuan and Meng has shown how capsule endoscopy images may be obtained with advanced instrumentation and analyzed using image processing, feature extraction, and classification techniques to detect ulcers Virmani and Kumar described, in Chapter 14, a computerized tissue classification system to detect focal liver disease based on texture analysis In Chapter 15, Cipriani Frade et al presented computational aspects of CAD and CBIR via color image analysis applied to images of dermatological lesions on the lower limbs Chapter 16 by Boyd and Lagacé showed how imaging techniques, image analysis procedures, and finite element modeling methods could be used to derive parameters that could assist in the diagnosis of osteoporosis Various in vivo imaging techniques were used to derive a number of parameters, such as bone mineral density, trabecular volume fraction, and cortical thickness, that could be useful in the assessment of fracture risk Borotikar et al presented, in Chapter 17, techniques for augmented statistical shape modeling, mesh registration, and derivation of patient-specific anatomical references with applications in shoulder arthroplasty and rehabilitation The methods were integrated in a pipeline to facilitate pre-surgical planning and computer-aided surgery In Chapter 18, Knight and Khademi have shown how image processing and pattern recognition techniques can be designed to detect breast cancer in histopathology images Such approaches extend CAD techniques from medical images of patients to histological images of tissue samples and can improve the accuracy of the most decisive step in clinical diagnosis Chandra et al have shown, in Chapter 19, that diagnostically useful images may be obtained by using nonionizing microwaves If the associated technical difficulties are resolved, microwave imaging could provide a low-cost alternative to CT and MR imaging In Chapter 20, Traina et al discussed the potential applications and limitations of CBIR systems in the clinical environment They presented a CBIR system specialized for mammography that is able to encode perceptual parameters using feature extraction methods and distance functions The methods were shown to reduce the need for query refinement or relevance feedback cycles Chapter 21 by Deserno and Reichertz has shown how signal and image processing algorithms for biomedical and CAD applications can be integrated with models and schema drawn from the domain of informatics for improved management of patient care An important point to note from our discussion in the preface and from the chapters of the book is that quantitative analysis becomes possible by the application of computational procedures to medical images The logic of medical or clinical diagnosis via image analysis can then be objectively encoded and consistently applied in routine or repetitive tasks However, we emphasize that the end-goal of medical image analysis should be computer-aided diagnosis and not automated diagnosis A physician Concluding Remarks 507 or medical specialist typically uses a large amount of information in addition to images and measurements, including the general physical and mental state of the patient, family history, and socioeconomic factors affecting the patient, many of which are not amenable to quantification and logical rule-based processes Medical images are, at best, indirect indicators of the state of the patient; the results of image analysis need to be integrated with other clinical signs, symptoms, and nonmedical information by a physician The general background knowledge and intuition of the medical specialist play important roles in arriving at the final diagnosis Keeping in mind the realms of practice of various licensed and regulated professions, liability, and legal factors, the final diagnostic decision and communication with the patient are best left to the physician or medical specialist It is expected that quantitative and objective analysis facilitated by the application of medical image analysis, medical image informatics, CBIR, and CAD will lead to improved diagnostic decision by the physician CAD has become a part of the clinical workflow in the detection of breast cancer with mammograms, but is still in the infancy of its full potential for application to many other types of diseases or lesions imaged with various modalities CBIR is an alternative and complementary approach for image retrieval based on key words and metadata Initial results are promising for the use of CBIR as a diagnostic support tool In the future, it is likely that CAD and CBIR will be incorporated into PACS and will be employed as useful tools for diagnostic examinations in routine clinical work It is evident from the material presented in this book that digital image processing techniques can assist in quantitative analysis of medical images, that pattern recognition and classification techniques can facilitate CAD, and that CAD systems can assist in achieving efficient diagnosis, in designing optimal treatment protocols, in analyzing the effects of or response to treatment, and in clinical management of various abnormal conditions Medical imaging, medical image analysis, medical image informatics, CBIR, and CAD are proven, as well as essential techniques for health care Index μ CT see Micro-computed tomography (μ CT) Adverse event (AE), 494 AE see Adverse event (AE) Age-related bone changes, 357 Airways and lung parenchyma, in CT airway tree analysis, 211– 212 airway tree labeling, 210– 211 airway tree segmentation, 209– 210 lung lobe segmentation, 213 lung parenchyma analysis, 214 lung segmentation, 212– 213 Airway tree analysis, 211– 212 Airway tree labeling, 210– 211 Airway tree segmentation, 209– 210 Algorithms Contrast Source Inversion (CSI), 455 Modified Gradient (MGM), 455 Newton-type, 455 qualitative imaging, 456– 457 quantitative imaging, 455– 456 Anatomical landmark selection, 401– 402 ANN see Artificial neural network (ANN) Arteriosclerosis screening, 115– 117 Arterio-venous nicking (AVN), 39– 40 Artery-vein ratio (AVR), 36– 39 Artificial neural network (ANN), 33– 35 Augmented scapula, 405– 406 Augmented statistical shape modeling applications, 409– 410 clinical evaluation, 400– 409 anatomical landmark selection, 401– 402 building augmented scapula, 405– 406 intra- and inter-observer reliability, 402– 404 local and global validity of, 406 purpose, 400– 401 results, 406– 409 and image-based bone models hippocampi data, 372 scapula and humerus data, 371– 372 segmentation, 372– 374 overview, 369– 371 probabilistic principal component analysis (PPCA) humerus, 397– 398 methods, 395– 396 scapula, 396– 397 with size correction, 396 without-size correction, 396 and registration (see also Non-rigid registration; Rigid registration) basics of, 374– 375 rigid versus non-rigid, 375 shape model treatment principal component analysis (PCA), 394– 395 procrustes superposition, 393– 394 shape space and tangent space, 394 theoretical evaluation, 399– 400 Automated denoising, 88– 91 Automated segmentation, 91– 94 AVN see Arterio-venous nicking (AVN) AVR see Artery-vein ratio (AVR) Bilateral asymmetry detection challenges, 235– 236 characterization of direct methods, 228– 230 indirect methods, 227– 228 dataset of mammograms, 223 description, 220– 223 overview, 219– 220 pattern classification and cross-validation, 233– 234 and feature extraction, 230– 233 Tabá r masking procedures, 224– 226 Binary large objects (BLOBs), 497 Black top-hat transformation, 31– 32 BLOBs see Binary large objects (BLOBs) Blood vessel segmentation artificial neural network (ANN), 33– 35 double-ring filter and black top-hat transformation, 31– 32 high-order local autocorrelation (HLAC), 33– 35 matched filtering, 31 morphological image processing, 31– 32 Bone cortical properties, 337 509 510 osteoporosis, 338– 339 diagnosis, 339– 340 and loss, 338 structure, 335– 336 trabecular properties, 337 whole properties, 337– 338 Bone imaging, 460 dual-energy X-ray absorptiometry (DXA) clinical data, 343– 344 CT compared to, 348 description, 341– 342 measurement error, 342– 343 radiation dose, 342 HR-pQCT for accuracy and precision, 352 age-related bone changes, 357 cortical porosity, 356 finite element analysis (FEA), 352– 355 fracture risk assessment, 356 image analysis and measures, 351– 352 patient positioning and image acquisition, 350 registration, 351 scanner hardware and specifications, 348– 350 segmentation, 350– 351 trabecular level, 354– 355 whole bone level, 354 micro-computed tomography (μ CT), 347– 348 overview, 340– 341 peripheral quantitative computed tomography (pQCT) clinical utility of, 346– 347 measurement errors, 346 quantitative computed tomography (QCT) dose in, 345 measurement error, 345– 346 overview, 344– 345 Boundary-based techniques, 14– 15 Brain extraction, and standardization, Brain imaging, 460 Brain tumor detection, by microwave imaging, 457– 460 Breast cancer histopathology diagnostic tool color space transformations, 434– 435 motivation of approach, 433– 434 Breast imaging, 460 Burn injuries, telemedicine for burn depth estimation classification stage, 133 interpretation of coordinates, 135– 137 material, 137 multidimensional scaling analysis (MDS), 134– 135 psychophysical experiment, 134 results, 133– 134, 137 segmentation stage, 132– 133 description, 131– 132 overview, 129– 130 surface area estimation, 138– 140 Index CAC see Computer-Aided Classification (CAC) system Calcium score, 146– 147 Cardiac function, 149 Cardiovascular disease (CVD) cardiovascular magnetic resonance imaging (CMRI) carotid plaque quantification, 164– 165 extracellular volume (ECV) mappings, 162– 163 flow assessment, 163– 164 left ventricular function assessment, 161– 162 overview, 160– 161 reviewed methods, 165 tagged, 163 T1 relaxation time, 162– 163 intravascular optical coherence tomography (IVOCT), 168– 171 intravascular ultrasound (IVUS) computer-aided systems, 167– 168 features, 165– 167 overview, 145– 146 SPECT and PET imaging description, 151 image analysis, 152– 155 planar image analysis, 152 reviewed methods, 155– 156 ultrasound imaging Doppler ultrasound, 157– 158 feature selection, ranking and classification, 159– 160 image intensity-based features, 158– 159 M-mode echocardiography, 156– 157 segmentation, 158 two-dimensional imaging, 157 ultrasound image analysis, 158 vascular ultrasound, 160 X-ray imaging calcium score, 146– 147 cardiac function, 149 pericardial/epicardial fat, 147 plaque quantification, 149 reviewed methods, 150 stenosis detection and measurement, 147– 149 Cardiovascular magnetic resonance imaging (CMRI) carotid plaque quantification, 164– 165 extracellular volume (ECV) mappings, 162– 163 flow assessment, 163– 164 left ventricular function assessment, 161– 162 overview, 160– 161 reviewed methods, 165 tagged, 163 T1 relaxation time, 162– 163 Carotid plaque quantification, 164– 165 CBIR see Computer-based image retrieval (CBIR) system CBMIR see Content-based medical image retrieval (CBMIR) CDISC see Clinical data interchange standards consortium (CDISC) Index Circularity, 15 Haralick’ s, 15 Classification performance, 446– 447 Classifier tools, 442– 443 Clinical data interchange standards consortium (CDISC), 496 Clinical evaluation, of SSM, 400– 409 anatomical landmark selection, 401– 402 building augmented scapula, 405– 406 intra- and inter-observer reliability, 402– 404 local and global validity of, 406 purpose, 400– 401 results, 406– 409 Clinical Trial Center Aachen (CTC-A), 497 Clinical trial management system (CTMS), 493– 494 CMRI see Cardiovascular magnetic resonance imaging (CMRI) Cole-Cole model, 454 Color features, 327 Color space transformations, 434– 435 Computer-aided analysis of retinal fundus images detection of retinal vasculature, 59 feature selection and pattern classification, 63– 64 openness of MTA, 59– 60 thickness of vessels, 60 tortuosity of vessels, 61 Computer-Aided Classification (CAC) system and focal liver lesions (FLLs) classification module, 313– 314 data collection protocols, 306 dataset description, 305– 306 overview, 304– 305 PCA-NN based hierarchical classifier design, 315– 319 PCA-NN based multi-class classifier design, 308– 315 ROI extraction protocols, 306– 308 overview, 303– 304 Computer-aided detection of masses, 258– 259 of microcalcifications, 250– 258 description, 250– 251 enhancement by regularized reconstruction, 251– 253 by joint DBT-PPJ approach, 254– 256 in 2D planar projection (PPJ) image, 253– 254 in 2D projection views, 256– 258 overview, 249– 250 Computer-aided diagnosis (CAD) automation of data flow, 502 of data processing, 502– 503 of workflow, 502 and data augmentation of chest CT images, 193– 194 effect on CAD training, 197– 198 evaluation of volumetry algorithms, 198– 199 511 image space lesion insertion using Poisson image editing, 190– 193 of mammograms, 194– 197 overview, 187– 190 dental imaging arteriosclerosis screening, 115– 117 CAD with intraoral radiographs, 104– 105 dental panoramic radiographs (DPRs), 105– 106 maxillary sinuses, 117– 119 osteoporosis screening, 106– 111 overview, 103– 104 periodontal diseases, 119– 122 texture analysis, 111– 115 diabetic macular edema (DME) automated denoising, 88– 91 automated segmentation, 91– 94 overview, 85– 87 sub-retinal layer thickness estimation, 98– 99 sub-retinal surface segmentation error, 96– 98 in ECG recording, 501– 502 of gastrointestinal (GI) diseases and feature coding, 288– 290 and feature extraction, 291– 292 image acquisition and experimental setup, 295 and key points extraction, 290– 291 and LLC histogram, 293, 295– 296 and max pooling strategy, 293 overview, 285– 287 and polyp recognition, 287– 288, 296– 298 and Probabilistic Latent Semantic Analysis (PLSA) model, 290, 293– 295 and visual words calculation, 292 of plus disease, 58– 59 of spinal abnormalities description, 271– 273 and intervertebral disc disease, 278– 281 magnetic resonance imaging (MRI), 270– 271 overview, 269– 270 and subchondral bone marrow abnormalities, 277 vertebral body fracture, 273– 277 in wound imaging, 499– 501 Computer-based image retrieval (CBIR) system color features, 327 overview, 326– 327 texture features, 327– 329 Content-based medical image retrieval (CBMIR) architecture of, 469– 470 distance function module, 472 evaluating similarity with diversity new information experiment, 484– 486 performance experiment, 486 feature extraction module global image features, 470– 471 local image features, 471– 472 and group near-duplicate elements, 482– 484 and Higiia 512 knowledge database, 475– 476 overview, 473– 475 potential and limitations, 476– 478 similarity queries and perceptual parameters, 478– 481 overview, 467– 468 perceptual parameters (PP), 472– 473 similarity queries, 473, 481– 482 Contrast Source Inversion (CSI) algorithm, 455 Cortical bone properties, 337 Cortical porosity, 356 Cross-validation and pattern classification, 233– 234 Cross-validation, and digital pathology, 443 CTC-A see Clinical Trial Center Aachen (CTC-A) CTMS see Clinical trial management system (CTMS) CVD see Cardiovascular disease (CVD) Data augmentation of chest CT images, 193– 194 effect on CAD training, 197– 198 evaluation of volumetry algorithms, 198– 199 image space lesion insertion using Poisson image editing, 190– 193 of mammograms, 194– 197 overview, 187– 190 DBT see Digital breast tomosynthesis (DBT) Debye model, 454 Denoising, 435 Dental imaging arteriosclerosis screening, 115– 117 CAD with intraoral radiographs, 104– 105 dental panoramic radiographs (DPRs), 105– 106 maxillary sinuses, 117– 119 osteoporosis screening, 106– 111 overview, 103– 104 periodontal diseases, 119– 122 texture analysis, 111– 115 Dental panoramic radiographs (DPRs), 105– 106 Depth estimation, of burns classification stage, 133 interpretation of coordinates, 135– 137 material, 137 multidimensional scaling analysis (MDS), 134– 135 psychophysical experiment, 134 results, 133– 134, 137 segmentation stage, 132– 133 Dermatological ulcers, CAD of computer-based image retrieval (CBIR) system color features, 327 overview, 326– 327 texture features, 327– 329 protocols for clinical photography equipment, 329 disposable rulers, 330 image resolution, 329– 330 images acquisition, 329 Index image subject, 330 surgical drape, 330 results, 330 wound healing, 323– 324 histopathological analysis, 326 rate, 324 tissues analysis by image, 324– 326 Diabetic macular edema (DME) automated denoising, 88– 91 performance of, 95– 96 automated segmentation, 91– 94 overview, 85– 87 sub-retinal layer thickness estimation, 98– 99 sub-retinal surface segmentation error, 96– 98 Diabetic retinopathy (DR) hemorrhage detection methods, 40 microaneurysm detection methods, 40– 46 OD segmentation, 47– 48 optic cup segmentation, 48– 49 Dielectric properties, of tissues, 452– 454 Diffuse lung diseases airways and lung parenchyma in CT airway tree analysis, 211– 212 airway tree labeling, 210– 211 airway tree segmentation, 209– 210 lung lobe segmentation, 213 lung parenchyma analysis, 214 lung segmentation, 212– 213 emphysema and airway diseases, 204– 207 interstitial lung diseases, 207– 208 overview, 203– 204 Digital breast tomosynthesis (DBT) characteristics of, 243– 249 computer-aided detection in of masses, 258– 259 of microcalcifications, 250– 258 overview, 249– 250 overview, 241– 243 Digital pathology, 432– 433 see also Pathology and breast cancer histopathology diagnostic tool color space transformations, 434– 435 motivation of approach, 433– 434 and classification performance, 446– 447 and classifier tools, 442– 443 and cross-validation, 443 and disease-inspired feature design interstitial distance, 440 nucleus homogeneity feature, 441– 442 triangulation edge lengths, 441 and feature quality, 445– 446 and image database, 443 and nuclei segmentation, 445 image and thresholding for, 438– 439 object-size assertion, 439 and preprocessing denoising, 435 stain deconvolution, 435– 437 Index and stain deconvolution, 444– 445 and validation metrics, 443– 444 Disease-inspired feature design interstitial distance, 440 nucleus homogeneity feature, 441– 442 triangulation edge lengths, 441 Distance function module, 472 DME see Diabetic macular edema (DME) Doppler ultrasound, 157– 158 Double-ring filter, 31– 32 DPRs see Dental panoramic radiographs (DPRs) DR see Diabetic retinopathy (DR) Dual-energy X-ray absorptiometry (DXA) clinical data, 343– 344 CT compared to, 348 description, 341– 342 measurement error, 342– 343 radiation dose, 342 DXA see Dual-energy X-ray absorptiometry (DXA) EBV see Endobronchial valves (EBV) ECAE see Extra corporeal air error (ECAE) ECRF see Electronic case report form (eCRF) ECV see Extracellular volume (ECV) mappings EDCS see Electronic data capture system (EDCS) Edge-based PVA modeling, 10 Electronic case report form (eCRF), 492 Electronic data capture system (EDCS), 494– 495 Elongatedness, 16 Emphysema and airway diseases, 204– 207 Endobronchial valves (EBV), 206 Evaluation metrics, 16– 17 Exploratory noise analysis, 17– 18 Extracellular volume (ECV) mappings, 162– 163 Extra corporeal air error (ECAE), 209 Fast healthcare interoperability resources (FHIR), 496 FDTD see Finite-difference-time-domain (FDTD) simulations FEA see Finite element analysis (FEA) Feature coding, 288– 290 Feature extraction, 291– 292 and pattern classification, 230– 233 Feature extraction module global image features, 470– 471 local image features, 471– 472 Feature quality, and digital pathology, 445– 446 Feature selection, 159– 160 FHIR see Fast healthcare interoperability resources (FHIR) Finite-difference-time-domain (FDTD) simulations, 457 Finite element analysis (FEA), 352– 355 FLAIR see Fluid attenuation inversion recovery (FLAIR) MRI FLLs see Focal liver lesions (FLLs) Flow assessment, 163– 164 513 Fluid attenuation inversion recovery (FLAIR) MRI, 4– 5 Focal liver lesions (FLLs) classification module, 313– 314 data collection protocols, 306 dataset description, 305– 306 overview, 304– 305 PCA-NN based hierarchical classifier design, 315– 319 PCA-NN based multi-class classifier design, 308– 315 ROI extraction protocols, 306– 308 Forced vital capacity (FVC), 207 Fracture risk assessment, 356 Full-width-at-half-maximum (FWHM), 211 Fuzzy edge model, 10– 11 FVC see Forced vital capacity (FVC) FWHM see Full-width-at-half-maximum (FWHM) Gastrointestinal (GI) diseases, CAD of and feature coding, 288– 290 and feature extraction, 291– 292 image acquisition and experimental setup, 295 and key points extraction, 290– 291 and LLC histogram, 293, 295– 296 and max pooling strategy, 293 overview, 285– 287 and polyp recognition, 287– 288, 296– 298 experiment results of, 295 and Probabilistic Latent Semantic Analysis (PLSA) model, 290, 293– 295 and visual words calculation, 292 GGO see Ground glass opacities (GGO) GI see Gastrointestinal (GI) diseases, CAD of Glaucoma large cupping detection, 46– 49 nerve fiber layer defect (NLFD), 49– 51 Global edge description, 11– 12 Global image features, 470– 471 Global shape metrics, 15– 16 Google Web Toolkit (GWT), 497 Graphical user interfaces (GUI), 497 Ground glass opacities (GGO), 207 GUI see Graphical user interfaces (GUI) GWT see Google Web Toolkit (GWT) Haralick’ s circularity, 15 HC see Honeycombing (HC) H&E see Hematoxylin and eosin (H&E) stains Healthcare information system (HIS), 495– 496 Heart imaging, 461 Hematoxylin and eosin (H&E) stains, 429 Hemorrhage detection methods, 40 High-order local autocorrelation (HLAC), 33– 35 High resolution computed tomography (HRCT), 204 Higiia knowledge database, 475– 476 514 overview, 473– 475 potential and limitations, 476– 478 similarity queries and perceptual parameters, 478– 481 Hippocampi and image-based bone models, 372 HIS see Healthcare information system (HIS) HLAC see High-order local autocorrelation (HLAC) Honeycombing (HC), 207 HRCT see High resolution computed tomography (HRCT) HR-pQCT for bone and joint imaging accuracy and precision, 352 age-related bone changes, 357 cortical porosity, 356 finite element analysis (FEA), 352– 355 fracture risk assessment, 356 image analysis and measures, 351– 352 patient positioning and image acquisition, 350 registration, 351 scanner hardware and specifications, 348– 350 segmentation, 350– 351 trabecular level, 354– 355 whole bone level, 354 Humerus and image-based bone models, 371– 372 and PPCA, 397– 398 Hypertensive retinopathy arterio-venous nicking (AVN), 39– 40 artery-vein ratio (AVR), 36– 39 overview, 35– 36 Idiopathic pulmonary fibrosis (IPF), 207 ILD see Interstitial lung diseases (ILD) Image acquisition for bone imaging, 350 for gastrointestinal (GI) diseases, 295 Image analysis and measures, 351– 352 Image-based bone models, and SSM hippocampi data, 372 scapula and humerus data, 371– 372 segmentation, 372– 374 Image database, and digital pathology, 443 Image data management, 499 Image intensity-based features, 158– 159 Immunohistochemical stains, 428– 429 Inter-observer reliability, 402– 404 Interpretation of coordinates, 135– 137 Interstitial distance, 440 Interstitial lung diseases (ILD), 207– 208 Intervertebral disc disease, 278– 281 Intra-observer reliability, 402– 404 Intraoral radiographs, 104– 105 Intravascular optical coherence tomography (IVOCT), 168– 171 Intravascular ultrasound (IVUS) computer-aided systems, 167– 168 features, 165– 167 Index IPF see Idiopathic pulmonary fibrosis (IPF) IVOCT see Intravascular optical coherence tomography (IVOCT) IVUS see Intravascular ultrasound (IVUS) Java R Interface (JRI), 499 Java server pages (JSP), 497 Joint imaging, 460– 461 HR-pQCT for accuracy and precision, 352 age-related bone changes, 357 cortical porosity, 356 finite element analysis (FEA), 352– 355 fracture risk assessment, 356 image analysis and measures, 351– 352 patient positioning and image acquisition, 350 registration, 351 scanner hardware and specifications, 348– 350 segmentation, 350– 351 trabecular level, 354– 355 whole bone level, 354 JRI see Java R Interface (JRI) JSP see Java server pages (JSP) Key points extraction, 290– 291 Laboratory information system (LIS), 495 Large cupping detection, 46– 49 Left main bronchus (LMB), 210 Left ventricular function assessment, 161– 162 Levernberg-Marquardt method, 456 LIS see Laboratory information system (LIS) LLC histogram, 293, 295– 296 LMB see Left main bronchus (LMB) LMPR see Log-Magnitude and Phase Reconstruction (LMPR) Local image features, 471– 472 Log-Magnitude and Phase Reconstruction (LMPR), 456 Lung lobe segmentation, 213 Lung parenchyma analysis, 214 Lung segmentation, 212– 213 Magnetic resonance imaging (MRI), 3– 4 acquisition noise, 5– 6 fluid attenuation inversion recovery (FLAIR), 4– 5 intensity inhomogeneity, intensity non-standardness, noise in parallel MRI, partial volume averaging (PVA), scanning parameters, 6– 7 single-coil noise models, 5– 6 for spinal abnormalities, 270– 271 vertebral body fracture based on, 275– 277 and white matter lesions (WML), 3– 4 Mammograms, 223 data augmentation of, 194– 197 Index Matched filtering, 31 Maxillary sinuses, 117– 119 Max pooling strategy, 293 MDS see Multidimensional scaling analysis (MDS) Mean lung density (MLD), 214 Mean-virtual shape, 383– 384 deforming, 386– 388 rescaling, 396 stability, 390– 392 Measurement error in DXA, 342– 343 in pQCT, 346 in QCT, 345– 346 Medical registries, 495 Microaneurysm detection methods, 40– 46 Microcalcifications detection, 250– 258 description, 250– 251 enhancement by regularized reconstruction, 251– 253 by joint DBT-PPJ approach, 254– 256 in 2D planar projection (PPJ) image, 253– 254 in 2D projection views, 256– 258 Micro-computed tomography (μ CT), 347– 348 Microwave imaging algorithms qualitative imaging, 456– 457 quantitative imaging, 455– 456 and bone imaging, 460 and brain imaging, 460 brain tumor detection, 457– 460 and breast imaging, 460 challenges, 461– 462 dielectric properties of tissues, 452– 454 and heart imaging, 461 and miscellaneous biomedical body imaging, 461 overview, 451– 452 setup, 454 soft tissues and joint imaging, 460– 461 MLD see Mean lung density (MLD) M-mode echocardiography, 156– 157 Modified Gradient (MGM) algorithm, 455 Morphological image processing, 31– 32 Multidimensional scaling analysis (MDS), 134– 135 Nerve fiber layer defect (NLFD), 49– 51 New information experiment, 484– 486 Newton-Kontorovich method, 456 Newton-type algorithms, 455 NLFD see Nerve fiber layer defect (NLFD) Noise analysis exploratory, 17– 18 testing for common distributions, testing for spatial correlation, 8– 9 testing for stationarity, Non-rigid registration assessing registration quality, 389 coherence point drift (CPD), 380– 382 multiple iterations of, 388– 389 515 correspondence quality, 389– 390 finding correspondence across sample, 384– 386 mean-virtual shape, 383– 384 deforming, 386– 388 stability, 390– 392 testing for values, 382– 383 Non-specific interstitial pneumonia (NSIP), 207 NSIP see Non-specific interstitial pneumonia (NSIP) Nuclei segmentation, 445 image and thresholding for, 438– 439 object-size assertion, 439 Nucleus homogeneity feature, 441– 442 Object-size assertion, 439 OC see OpenClinica (OC) ODM see Operational data model (ODM) OD segmentation, 47– 48 Office of Rare Diseases Research (ORDR), 495 OpenClinica (OC), 497– 498 Operational data model (ODM), 496 Optic cup segmentation, 48– 49 ORDR see Office of Rare Diseases Research (ORDR) Osteoporosis, bone, 338– 339 diagnosis, 339– 340 and loss, 338 Osteoporosis screening, 106– 111 PACS see Picture archiving and communication system (PACS) Pathology see also Digital pathology challenges in, 429– 430 hematoxylin and eosin (H&E) stains, 429 immunohistochemical stains, 428– 429 overview, 428 quality analysis, 430– 432 Pattern classification and cross-validation, 233– 234 and feature extraction, 230– 233 PCA see Principal component analysis (PCA) Performance experiment, 486 Pericardial/epicardial fat, 147 Periodontal diseases, 119– 122 Peripheral quantitative computed tomography (pQCT) clinical utility of, 346– 347 measurement errors, 346 Picture archiving and communication system (PACS), 467, 492 Planar image analysis, 152 Plaque quantification, 149 PLSA see Probabilistic Latent Semantic Analysis (PLSA) model Plus disease, 58 CAD of, 58– 59 Polyp recognition, 287– 288, 296– 298 experiment results of, 295 Positron emission tomography (PET) imaging description, 151 516 image analysis, 152– 155 planar image analysis, 152 reviewed methods, 155– 156 PPCA see Probabilistic principal component analysis (PPCA) PQCT see Peripheral quantitative computed tomography (pQCT) Preprocessing, and digital pathology denoising, 435 stain deconvolution, 435– 437 Principal component analysis (PCA), 394– 395 -NN based hierarchical classifier design, 315– 319 -NN based multi-class classifier design, 308– 315 Probabilistic Latent Semantic Analysis (PLSA) model, 290, 293– 295 Probabilistic principal component analysis (PPCA) humerus, 397– 398 methods, 395– 396 scapula, 396– 397 with size correction, 396 without-size correction, 396 Procrustes superposition, 393– 394 Protocols for CAD dermatological ulcers clinical photography equipment, 329 disposable rulers, 330 image resolution, 329– 330 images acquisition, 329 image subject, 330 surgical drape, 330 system interconnection, 499 Psychophysical experiment, 134 PVA quantification edge-based modeling, 10 estimating α , 12– 13 fuzzy edge model, 10– 11 global edge description, 11– 12 overview, 9– 10 segmentation, 13 QCT see Quantitative computed tomography (QCT) Qualitative imaging algorithms, 456– 457 Quality analysis, of pathology, 430– 432 Quantitative computed tomography (QCT) dose in, 345 measurement error, 345– 346 overview, 344– 345 Quantitative imaging algorithms, 455– 456 Radiation dose, for DXA, 342 Radiology information system (RIS), 495 RDR Framework, 498 Registration basics of, 374– 375 for bone imaging, 351 non-rigid assessing registration quality, 389 Index coherence point drift (CPD), 380– 382, 388– 389 correspondence quality, 389– 390 finding correspondence across sample, 384– 386 mean-virtual shape, 383– 384, 386– 388, 390– 392 testing for values, 382– 383 rigid centroid size and initialization, 378 data, 378 intrinsic consensus shape, 378– 380 iterative closest point (ICP), 376 iterative median closest point (IMCP), 376– 378 robust iterative closest point (ICPR), 376 Representational state transfer (REST), 499 REST see Representational state transfer (REST) Retinal fundus images blood vessel segmentation artificial neural network (ANN), 33– 35 double-ring filter and black top-hat transformation, 31– 32 high-order local autocorrelation (HLAC), 33– 35 matched filtering, 31 morphological image processing, 31– 32 detection of retinal vasculature, 59 diabetic retinopathy (DR) hemorrhage detection methods, 40 microaneurysm detection methods, 40– 46 OD segmentation, 47– 48 optic cup segmentation, 48– 49 feature selection and pattern classification, 63– 64 glaucoma large cupping detection, 46– 49 nerve fiber layer defect (NLFD), 49– 51 hypertensive retinopathy arterio-venous nicking (AVN), 39– 40 artery-vein ratio (AVR), 36– 39 overview, 35– 36 openness of MTA, 59– 60 overview, 29– 30 thickness of vessels, 60 three-dimensional eye devices, 51– 52 tortuosity of vessels, 61 Retinal vasculature, detection of, 59 Retinopathy of prematurity (ROP) computer-aided analysis of retinal fundus images detection of retinal vasculature, 59 feature selection and pattern classification, 63– 64 openness of MTA, 59– 60 thickness of vessels, 60 tortuosity of vessels, 61 overview, 57– 58 plus disease, 58 CAD of, 58– 59 results, 69– 73 retinal vasculature detection of retinal vessels, 64 measurement of arcade angle, 65 Index parabolic modeling of MTA, 65 segmentation of detected vessels, 64 thickness of MTA, 65– 68 tortuosity of vessels, 68– 69 telemedicine for ROP In calgary (TROPIC), 61– 63 Right main bronchus (RMB), 210 Rigid registration centroid size and initialization, 378 data, 378 intrinsic consensus shape humerus, 379– 380 scapula, 378– 379 iterative closest point (ICP), 376 iterative median closest point (IMCP), 376– 377 software, 378 robust iterative closest point (ICPR), 376 RIS see Radiology information system (RIS) RMB see Right main bronchus (RMB) ROP see Retinopathy of prematurity (ROP) Scapula augmented, 405– 406 and image-based bone models, 371– 372 and PPCA, 396– 397 Secure file transfer protocol (SFTP), 499 Segmentation automated, 91– 94 of bone imaging, 350– 351 depth estimation, of burns, 132– 133 OD, 47– 48 optic cup, 48– 49 PVA quantification, 13 and SSM, 372– 374 ultrasound imaging, for CVD, 158 SFTP see Secure file transfer protocol (SFTP) Shape analysis boundary-based techniques, 14– 15 global shape metrics, 15– 16 overview, 13– 14 Shape characterization, 22– 25 Shape model treatment principal component analysis (PCA), 394– 395 procrustes superposition, 393– 394 shape space and tangent space, 394 Shape space and tangent space, 394 Signal data management, 499 Simple object access protocol (SOAP), 499 Single photon emission computed tomography (SPECT) imaging description, 151 image analysis, 152– 155 planar image analysis, 152 reviewed methods, 155– 156 Single sign on (SSO), 496 Single walled carbon nanotubes (SWCT), 459 Skin wound healing, 323– 324 histopathological analysis, 326 517 rate, 324 tissues analysis by image, 324– 326 SOAP see Simple object access protocol (SOAP) Soft tissues, and joint imaging, 460– 461 Solidity, 16 Spinal abnormalities, CAD of description, 271– 273 and intervertebral disc disease, 278– 281 magnetic resonance imaging (MRI), 270– 271 overview, 269– 270 and subchondral bone marrow abnormalities, 277 vertebral body fracture based on CT, 274– 275 based on MRI, 275– 277 based on radiography, 273– 274 SSM see Statistical shape modeling (SSM) SSO see Single sign on (SSO) Stain deconvolution, 435– 437, 444– 445 Standardization, 18 and brain extraction, Statistical shape modeling (SSM) applications, 409– 410 clinical evaluation, 400– 409 anatomical landmark selection, 401– 402 building augmented scapula, 405– 406 intra- and inter-observer reliability, 402– 404 local and global validity of, 406 purpose, 400– 401 results, 406– 409 and image-based bone models hippocampi data, 372 scapula and humerus data, 371– 372 segmentation, 372– 374 overview, 369– 371 probabilistic principal component analysis (PPCA) humerus, 397– 398 methods, 395– 396 scapula, 396– 397 with size correction, 396 without-size correction, 396 and registration (see also Non-rigid registration; Rigid registration) basics of, 374– 375 rigid versus non-rigid, 375 shape model treatment principal component analysis (PCA), 394– 395 procrustes superposition, 393– 394 shape space and tangent space, 394 theoretical evaluation, 399– 400 Stenosis detection and measurement, 147– 149 Study management tool, 497 Subchondral bone marrow abnormalities, 277 Sub-retinal layer thickness estimation, 98– 99 Sub-retinal surface segmentation error, 96– 98 SWCT see Single walled carbon nanotubes (SWCT) System architecture, 498– 499 System interconnection, 496 518 image and signal data management, 499 protocols, 499 system architecture, 498– 499 Tabá r masking procedures, 224– 226 TAE see Tracheal air error (TAE) Tagged CMRI, 163 Tangent space and shape space, 394 Telemedicine, for burns burn depth estimation classification stage, 133 interpretation of coordinates, 135– 137 material, 137 multidimensional scaling analysis (MDS), 134– 135 psychophysical experiment, 134 results, 133– 134, 137 segmentation stage, 132– 133 description, 131– 132 overview, 129– 130 surface area estimation, 138– 140 Texture analysis, 111– 115 Texture features, 327– 329 Three-dimensional eye devices, 51– 52 Tikhonov regularization, 456 Tissues, dielectric properties of, 452– 454 Trabecular bone properties, 337 Trabecular level, 354– 355 Tracheal air error (TAE), 209 T1 relaxation time, 162– 163 Triangulation edge lengths, 441 Two-dimensional imaging, 157 2D planar projection (PPJ) image, 253– 254 2D projection views, 256– 258 UIP see Usual interstitial pneumonia (UIP) Ultrasound image analysis, 158 Ultrasound imaging, for CVD Doppler ultrasound, 157– 158 feature selection, ranking and classification, 159– 160 image intensity-based features, 158– 159 M-mode echocardiography, 156– 157 segmentation, 158 two-dimensional imaging, 157 ultrasound image analysis, 158 vascular ultrasound, 160 Ultra-wideband (UWB) signal, 456 Usual interstitial pneumonia (UIP), 207 UWB see Ultra-wideband (UWB) signal Validation metrics, 443– 444 Vascular ultrasound, 160 Vertebral body fracture based on CT, 274– 275 based on MRI, 275– 277 Index based on radiography, 273– 274 Visual words calculation, 292 Volumetry algorithms, 198– 199 White matter lesions (WML) evaluation metrics, 16– 17 experimental data, 17 exploratory noise analysis, 17– 18 and MRI, 3– 4 acquisition noise, 5– 6 fluid attenuation inversion recovery (FLAIR), 4– 5 intensity inhomogeneity, intensity non-standardness, noise in parallel MRI, partial volume averaging (PVA), scanning parameters, 6– 7 single-coil noise models, 5– 6 noise analysis testing for common distributions, testing for spatial correlation, 8– 9 testing for stationarity, overview, 2– 3 and PVA quantification edge-based modeling, 10 estimating α , 12– 13 fuzzy edge model, 10– 11 global edge description, 11– 12 overview, 9– 10 segmentation, 13 segmentation evaluation, 18– 22 shape analysis boundary-based techniques, 14– 15 global shape metrics, 15– 16 overview, 13– 14 shape characterization, 22– 25 standardization, 18 and brain extraction, Whole bone level, 354 Whole bone properties, 337– 338 WML see White matter lesions (WML) Wound healing, skin, 323– 324 histopathological analysis, 326 rate, 324 tissues analysis by image, 324– 326 X-ray imaging, of CVD calcium score, 146– 147 cardiac function, 149 pericardial/epicardial fat, 147 plaque quantification, 149 reviewed methods, 150 stenosis detection and measurement, 147– 149 YACTA see “ Yet Another CT Analyzer” (YACTA) “ Yet Another CT Analyzer” (YACTA), 208 ... mammography JAMA-Journal of the American Medical Association 20 14;311 (24 ): 24 99– 507 26 2 Medical Image Analysis and Informatics: Computer-Aided Diagnosis and Therapy 13 Lang K, Andersson I, Rosso... improvements in sensitivity and lead time of radiological signs Clinical Radiology 20 03;58 (2) : 128 – 32 26 8 Medical Image Analysis and Informatics: Computer-Aided Diagnosis and Therapy 115 Rangayyan... depends on the availability of and advances in MRI equipment and computer analysis capacity 26 9 27 0 Medical Image Analysis and Informatics: Computer-Aided Diagnosis and Therapy Anterior Body Facet