Many advances and technological developments in medical image acquisition systems have resulted in making the early signs of cancer more apparent on mammograms, chest radiographs, and CT images. However, the benefit of a medical imaging examination depends on both the quality of the medical images and the ability of the radiologist to interpret those images. During the past approximately 25 years, technological developments in computerized image analysis have attempted to improve the interpretation stage of the medical imaging examination.

For example, while mammography is the best screening method for the early detection of breast cancer, radiologists do miss lesions on mammograms. Misses of lesions on radiographic images may be due to the presence of quantum mottle or overlapping normal structures, or they may be due to radiologists’ insufficient search patterns and lapses in perception. The variability in image interpretation among different radiologists also has been well documented. 1,2 In addition, in screening programs (such as mammography for breast cancer screening), the detection of an abnormality is a tedious task as most cases are normal, leading to misses due to simple oversight. Use of output, however, from a computerized analysis of an image by a radiologist may help in the detection or diagnostic tasks, and potentially improve the overall interpretation of medical images and the subsequent patient care.

Development and implementation of computer-aided detection and diagnosis (CAD) involve the application of computer technology to the process of medical image interpretation. 3 Radiologists can use the output from a computerized analysis of medical images as a “second opinion” in detecting lesions and in making diagnostic decisions. It should be noted that the final diagnosis is made by the clinician, eg, the radiologist. Thus, the computer output needs to be at a sufficient level of sensitivity and specificity, and in addition, the output must be displayed in a user-friendly format for effective and efficient use by the radiologist.

One of the earliest published investigations into the computerized analysis of medical images was reported in 1967 by Winsberg et al, 4 and included a computerized method that compared mammographic density patterns in various areas within an individual breast and between right and left breasts. With advances in computer vision, artificial intelligence, and computer technology, along with recognized medical screening needs and the availability of large databases of cases, the field of computer-aided diagnosis has grown substantially since the mid 1980s, with many comprehensive reviews written. 3,5-12

Computerized detection of lesions involves having the computer locate suspicious regions, leaving the subsequent classification (eg, probability of malignancy status) of the lesion and patient management decisions to the radiologist. In such situations, the computer is acting as a second reader, like a spell-checker, in the cancer screening process. Once a possible abnormality is detected, its characteristics must be evaluated by the radiologist in order to estimate a likelihood of malignancy and to yield a decision on patient management. Thus, algorithms for computerized characterization and diagnosis are also being developed.

CAD AND BREAST CANCER

Currently, computerized image analysis techniques are being developed to detect, to characterize, and potentially to diagnose lesions on breast images. Such methods may take as input the digital image data and then output clinically relevant information, including locations of suspect lesions, characterizations of suspect lesions, estimates of the probabilities of malignancy of suspect lesions, and/or numerical indices of breast cancer risk. Most CAD methods in mammography are being developed for the detection or characterization of either mass lesions or clustered microcalcifications, which are two of the primary signs of potential breast cancer. Radiographic mass lesions can be detected and/or characterized by using the mathematical descriptors of various features including radiating patterns of density (spiculation), margin sharpness, circumscribed configurations, shape, bilateral asymmetries, local textural changes, and temporal stability, as summarized elsewhere. 3 Mathematical descriptors of calcifications are based on the radiographic presentation of individual calcification (eg, shape, area, brightness), 3,13-16 the variation of individual features within a cluster, 17 the spatial distribution of calcifications within a cluster, 17 and the knowledge that clinically significant microcalcifications are clustered. 18

Figure 1. Display interface for a multimodality workstation that displays computer outputs in numerical, pictorial, and graphical modes. Lesions outlined in red are malignant and those outlined in green are benign. Such images are retrieved automatically from an online atlas. Courtesy of Maryellen L. Giger, PhD, University of Chicago.

Computer-detection systems output the location of suspect abnormalities to the radiologist by marks on display monitors, hard-copy film, or paper. Note that in the computer output for the detection of potential lesions, both actual lesions and false-positives are indicated. Over the years, investigators have worked to increase detection sensitivity while decreasing the number of false marks per image. Investigators have also studied the necessary sensitivity and false mark rate for computerized detection methods in order to yield improvement in radiologists’ performance levels. 19 Sensitivity that is too low or the presence of too many false marks will yield computer output that may be detrimental to radiologist performance.

Figure 2. Thoracic CT image illustrating the computer detection of a nodule and a false-positive that corresponds to crossing blood vessels. Courtesy of Samuel A. Armato III, PhD, University of Chicago.

Computerized detection methods have been evaluated both in terms of their own performance as well as in terms of the performance of the human using the computer-detection output as an aid. Observer studies have shown that radiologists’ performances do increase when a computerized detection aid is used. 20-23 Research progress has resulted in commercial computer-aided detection systems, with some approved by the FDA and with many insurance carriers (including Medicare) providing coverage for such technology. Currently, prospective clinical studies of computer-aided detection systems are being performed in order to determine their additive value in mammographic interpretations in breast cancer screening programs. In one such study, 24 radiologists interpreted 12,860 mammographic cases using a commercial CAD system and showed a 19.5% increase in cancer detection along with a 18.5% increase in recall rate. In another study, researchers investigated the effect of introducing CAD into a radiology practice with initial results indicating no change in cancer detection rate or recall rate. 25 Additional prospective studies are necessary to validate the clinical usefulness of CAD.

Once a possible abnormality is detected, its characteristics must be evaluated by the radiologist in order to estimate a likelihood of malignancy and to yield a decision on patient management. Characteristics of the lesion may be evaluated further by multiple imaging techniques including special view mammography, ultrasound, and MRI in order to improve the positive predictive value for biopsy recommendations. Computer-extracted features (mathematical descriptors) can be obtained from standard mammographic views (CC and MLO) as well as special view mammograms. Similar to radiologist performance, it has been demonstrated that computer performance improved in diagnosing lesions on special view mammograms as compared to that on standard views. 26

In laboratory observer studies, computerized diagnostic methods have been shown to aid radiologists in the task of distinguishing between malignant and benign lesions. 27-29 Use of a computer diagnostic aid therefore has the potential to increase sensitivity, specificity, or both in the work-up of breast lesions. In addition, use of computer output is expected to reduce the variability between radiologists’ interpretations. 30

Computer-aided diagnosis methods in breast ultrasound are being explored by various researchers. 31-38 Mass lesions visible at ultrasound can be classified by computer using a variety of mathematical descriptors of texture, margin, and shape criteria. 31 Computerized analysis of magnetic resonance images has multiple potential benefits due to the significant variability in the assessment of MRI lesions by radiologists and by the lack of standard imaging protocols for breast MRI. Computer analysis of lesions on MRI images includes morphological features, temporal features, or combinations. 39-41 While some features encountered in breast imaging, such as margin sharpness, can be utilized across modalities, others are particular to the imaging modality, such as the computer characterization of the posterior acoustic behavior on ultrasound and the inhomogeneity of contrast uptake in breast MRI.

Radiologists interpret cases rather than individual images, thus the computerized analysis of breast images can be case-based as opposed to image-based. Computer analysis of multiple views or multiple modalities, however, requires effective and efficient displays in order to communicate the multiple images and output to the radiologist. Research has been performed to develop display interfaces, which would better present the computer output to the radiologist (see Figure 1 earlier in this article). 42-45 The output of a computer-aided diagnostic system can be presented in terms of numerical values related to the likelihood of malignancy, by displaying similar images of known diagnoses, or by a graphical representation of the unknown lesion relative to all lesions in a known database (an online atlas). 44,45 Searching within an online image atlas can be performed based on individual features or on a likelihood of malignancy.

CAD AND LUNG CANCER

An active area of research in computerized image analysis for lung cancer is in the use of computers to aid in the detection of lung nodules on thoracic CT. Even though the use of CT, as opposed to single-projection chest radiography, has helped remove much of the camouflaging presence of overlapping structures, the detection of lung nodules is still confounded by the presence of blood vessels. In addition, radiologists’ readings of CT images are hindered by the vast number of slices (images) generated by a CT scan with each requiring careful interpretation. This makes lesion detection a burdensome task that is further complicated by fatigue and distractions. It is expected that a computerized scheme for the detection of lung nodules on CT images should help radiologists focus their attention on regions suspect of being cancerous. This seems particularly advantageous in potential lung cancer screening procedures in which most cases will be normal.

Various investigators have been developing different methods for the computerized detection of lung nodules on thoracic CT images (see Figure 2). 46-59 Currently, methods include both two-dimensional and three-dimensional analyses for the delineation of lung volume, the segmentation of pulmonary structures indicative of an abnormality, and the extraction of features (ie, mathematical descriptors of lung nodules). In order to reduce false-positive detections (ie, computer-indicated locations that in fact do not correspond to a lesion), investigators have used artificial neural networks to merge computer-extracted features in order to distinguish nodules from normal structures such as blood vessels. Information such as continuity between slices can also be used to distinguish the tube-like structure of blood vessels from the approximate spherical nature of nodules.

In the development of computerized nodule detection methods, investigators have trained with cancerous lung nodules, all lung nodules, and/or any “actionable regions,” due to the limited number of cases in available databases. In addition, database collection is also hindered by the rapid growth of the field of CT imaging in which helical CT systems have evolved from single-slice to 16-slice to now clinically used 40-slice scanners. In addition, the choice of reconstruction algorithm for a CT scanner will affect the physical image properties and visual appearance of the transaxial images obtained from the CT scanner sinogram data. 57 Thus, it is expected that variations in reconstruction algorithm may also affect the performance of computerized analyses. Investigators are therefore beginning to examine the robustness of computerized lung nodule detection schemes across reconstruction algorithms and scanner type.

Once a nodule or other actionable region is detected, the radiologist is required to assess the likelihood that it is cancerous. Thus, computerized methods are being developed that include characterization of the suspect region. Most likely, computerized detection systems will include an indication of location as well as a means to indicate the probability of malignancy.

CAD AND COLON CANCER

While colon cancer is one of the leading causes of cancer deaths, many can be avoided if precursor colonic polyps are detected and removed. CT colonography (virtual colonoscopy) is being examined as a potential screening device (as an alternative to conventional colonoscopy) for the early detection of colonic polyps. Radiologist interpretation of a CT colonography examination can be quite time-consuming due to the potentially large number of axial CT images (400-700 slices). In addition, the overload of image data for interpretation may result in oversight errors. Thus, computerized image analysis techniques are being developed to aid in the interpretation of CT colonography images.

Various investigators are developing such computer algorithms for the visual presentation of CT colonography images for human interpretation60-64 and for computerized image analysis 65-74 to aid in the detection of colonic polyps. It is important to note that in the combined interpretation process, the presentation of the three-dimensional image data set as well as the computer output is critical for accurate interpretation, ease of use, and reduction in interpretation times. It is felt that appropriate displays and image analysis methods will help the process of bringing CT colonography to routine clinical use.

Computerized image analysis of CT colonography images has various stages, including segmentation of the colonic wall, detection of polyp candidates, reduction of false-positive detections, and display of final detection output. In the computerized detection of the polyps, geometric features of each voxel of the segmented colon are computed (Yoshida paper). The geometric model for structures in the lumen of the colon involves the calculation of a shape index in order to help distinguish polyps (which are caplike structures) from folds (which are elongated, ridge-like structures). 66,69 Subsequent mathematical descriptors are employed to eliminate false-positive detections that are caused, for example, by the presence of stool in the colon during imaging. 72,73

CT colonoscopy is emerging as a potential cancer screening tool with developments in both the image acquisition protocols and computerized image interpretation aids. It appears that both will be needed for ultimate clinical use.

SUMMARY

Limitations in the human eye-brain visual system, the presence of overlapping structures in images, and the vast number of normal cases in screening programs provide motivation for the use of computer techniques that have the potential to improve detection and diagnostic performance, and ultimately, patient care. The ultimate success of computerized image analyses in the interpretation of medical images depends on both the ability of the computer systems that extract and characterize suspect lesions and the ability of the radiologist to incorporate the computer output into decision-making. The clinical usefulness of computer-aided detection for screening mammography is being tested by actual prospective clinical usage of commercial systems. In addition, various observer studies for computer-aided diagnosis indicate the promising role of computer interpretation aids in diagnostic workup. For thoracic CT and CT colonography for the early detection of lung cancer and colon cancer, respectively, effective and efficient displays are also needed to help the radiologist incorporate the computer output into the three-dimensional image data.

The development of computerized techniques for aiding in the interpretation of medical images is progressing rapidly, and ultimate incorporation into routine clinical care is expected in the near future. Prospective clinical trials, however, will be needed.

ACKNOWLEGDMENTS

The author is grateful for the many fruitful discussions she has had with the faculty and research staff in the Department of Radiology, University of Chicago.

M.L. Giger is a shareholder in a computer aided detection company. It is the University of Chicago conflict-of-interest policy that investigators disclose publicly actual or potential significant financial interests that may appear to be affected by their research activities.

Maryellen L. Giger, PhD, is professor of radiology, University of Chicago.

References:

  1. Beam C, Layde PM, Sullivan D. Variability in the interpretation of screening mammograms by US radiologists. Arch Intern Med. 1996;156:209-213.
  2. Elmore JG, Wells CK, Lee CH, et al. Variability in radiologists’ interpretations of mammograms. New Engl J Med. 1994; 331:1492.
  3. Giger ML, Huo Z, Kupinski MA, Vyborny CJ. Computer-aided diagnosis in mammography. In: Sonka M, Fitzpatrick MJ, eds. Handbook of Medical Imaging. Vol 2. Medical Imaging Processing and Analysis. SPIE; 2000:915-1004.
  4. Winsberg F, Elkin FM, Macy J, Bordaz V, Weymouth W. Detection of radiographic abnormalities in mammograms by means of optical scanning and computer analysis. Radiology. 1967;89:211-215.
  5. Giger ML. Computer-aided diagnosis. In: Haus AG, Yaffe MJ, eds. Technical Aspects of Mammography. Oakbrook, Ill: RSNA Publications; 1993:283.
  6. Vyborny CJ, Giger ML. Computer vision and artificial intelligence in mammography. AJR Am J Roentgenol.1994;162:699.
  7. Vyborny CJ, Giger ML, Nishikawa RM. Computer-aided detection and diagnosis. Radiol Clin North Am. 2000;38:725-740.
  8. Doi K, MacMahon H, Giger ML, Hoffmann K, eds. Computer-Aided Diagnosis in Medical Imaging. Proceedings of the 1st International Workshop on Computer-Aided Diagnosis. New York: Elsevier; 1999.
  9. Jiang Y. Classification of breast lesions in mammograms. In: Bankman IN, ed. Handbook of Medical Imaging. Processing and Analysis. San Diego: Academic Press; 2000:341-358.
  10. Giger ML. Computerized image analysis in breast cancer detection and diagnosis. Seminars in Breast Disease. 2002;5:199-210.
  11. Giger ML. Computer-aided diagnosis in breast imaging. Proceedings of the 7th International Workshop on Digital Mammography, Chapel Hill, 2004. In press.
  12. Giger ML, MacMahon H. Image processing and computer-aided diagnosis. In: Greenes RA, Bauman R, eds. Radiologic Clinics of North America. Philadelphia: Saunders Publishing Co; 1996:565-596.
  13. Chan HP, Doi K, Galhotra S, et al. Image feature analysis and computer-aided diagnosis in digital radiography. I. Automated detection of microcalcifications in mammography. Med Phys. 1987;14:538.
  14. Chang YH, Zheng B, Good WF. Identification of clustered microcalcifications on digitized mammograms using morphology and topography-based computer-aided detection schemes: a preliminary experiment. Invest Radiol. 1998;33:746.
  15. Ibrahim N, Fujita H, Hara T, et al. Automated detection of clustered microcalcifications on mammograms: CAD system application to MIAS database. Phys Med Biol. 1997;42:2577.
  16. Karssemeijer N. A stochastic method for automated detection of microcalcifications in digital mammograms. Information Processing in Medical Imaging. New York: Springer-Verlag; 1991:227.
  17. Jiang Y, Nishikawa RM, Wolverton DE, et al. Automated feature analysis and classification of malignant and benign clustered microcalcifications. Radiology. 1996;198:671.
  18. Nishikawa RM, Giger ML, Doi K, Vyborny CJ, Schmidt RA. Computer-aided detection of clustered microcalcifications: an improved method for grouping detected signals. Med Phys. 1993;20:1661-1666.
  19. Zheng B, Ganott MA, Britton CA, et al. Soft-copy mammographic readings with different computer-assisted detection cuing environments: preliminary findings. Radiology. 2001;221:633-640.
  20. Kegelmeyer WP Jr, Pruneda JM, Bourland PD, et al. Computer-aided mammographic screening for spiculated lesions. Radiology. 1994;191:331.
  21. Chan HP, Doi K, Vyborny CJ, et al. Improvements in radiologist’s detection of clustered microcalcifications on mammograms: the potential of computer-aided diagnosis. Invest Radiol. 1990;25:1102.
  22. Nawano S, Murakami K, Moriyama N, et al. Computer-aided diagnosis in full digital mammography. Invest Radiol. 1999; 34:310-316.
  23. Thurfjell E, Gelig Thurfjell M, Egge E, et al. Sensitivity and specificity of computer-assisted breast cancer detection in mammography screening. Acta Radiol. 1988;39:384-388.
  24. Freer TW, Ulissey MJ. Screening mammography with computer-aided detection. Prospective study of 12,860 patients in a community breast center. Radiology. 2001;222:781-786.
  25. Gur D, Sumkin JH, Rockette HE, et al. Changes in breast cancer detection and mammography recall rates after the introduction of a computer-aided detection system. J Natl Cancer Inst. 2004;96:185-190.
  26. Huo Z, Giger ML, Vyborny CJ. Computerized analysis of multiple-mammographic views: potential usefulness of special view mammograms in computer-aided diagnosis. IEEE Trans Med Imaging. 2001;20:1285-1292.
  27. Jiang Y, Nishikawa RM, Schmidt RA, et al. Improving breast cancer diagnosis with computer-aided diagnosis. Acad Radiol. 1999;6:22.
  28. Huo Z, Giger ML, Vyborny CJ, Metz CE. Effectiveness of CAD in the diagnosis of breast cancer: an observer study on an independent database of mammograms Radiology. 2002;224:560-568.
  29. Chan HP, Sahiner B, Helvie MA, et al. Improvement of radiologists’ characterization of mammographic masses by using computer-aided diagnosis: an ROC study. Radiology. 1999;212:817-827.
  30. Jiang Y, Nishikawa RM, Schmidt RA, et al. Potential of computer-aided diagnosis to reduce variability in radiologists’ interpretations of mammograms depicting microcalcifications. Radiology. 2001; 220:787-794.
  31. Stavros AT, Thickman D, Rapp CL, et al. Solid breast nodules: use of sonography to distinguish between benign and malignant lesions. Radiology. 1995;196:123.
  32. Giger ML, Al-Hallaq H, Huo Z, et al. Computerized analysis of lesions in us images of the breast. Acad Radiol. 1999; 6:665-674.
  33. Garra BS, Krasner BH, Horii SC, Ascher S, Mun SK, Zeman RK. Improving the distinction between benign and malignant breast lesions: the value of sonographic texture analysis. Ultrason Imaging 1993; 15:267-285.
  34. Chen DR, Chang RF, Huang YL. Computer-aided diagnosis applied to us of solid breast nodules by using neural networks. Radiology. 1999;213:407–412.
  35. Golub RM, Parsons RE, Sigel B, et al. Differentiation of breast tumors by ultrasonic tissue characterization. J Ultrasound Med. 1993;12:601–608.
  36. Horsch K, Giger ML, Venta LA, Vyborny CJ. Computerized diagnosis of breast lesions on ultrasound. Med Phys. 2002;29:157-164.
  37. Sahiner B, LeCarpentier GL, Chan HP, et al. Computerized characterization of breast masses using three-dimensional ultrasound images. Proc SPIE. 1998;3338: 301-312.
  38. Horsch K, Giger ML, Vyborny CJ, Huo Z, Venta LA. Performance of CAD in the interpretation of lesions on breast sonography. Acad Radiol. 2004;11:272-280.
  39. Gilhuijs KGA, Giger ML, Bick U. Automated analysis of breast lesions in three dimensions using dynamic magnetic resonance imaging. Med Phys. 1998;25:1647.
  40. Penn AI, Bolinger L, Schnall MD, Loew MH. Discrimination of MR images of breast masses using fractal interpolation function models. Acad Radiol. 1999;6:156-163.
  41. Chen W, Giger ML, Lan L, Bick U. Computerized interpretation of breast MRI: investigation of enhancement-variance dynamics. Med Phys. 2004;31:1076-1082.
  42. Swett HA, Fisher PR, Cohn AI, Miller PI, Mutalik PG. Expert system controlled image display. Radiology. 1989;172:487-493.
  43. Sklansky J, Tao EY, Ornes C, Disher AC. A visualized mammographic database in computer-aided diagnosis. In: Computer-Aided Diagnosis in Medical Imaging. Amsterdam: Elsevier; 1999:215.
  44. Giger ML, Huo Z, Lan L, Vyborny CJ. Intelligent search workstation for computer-aided diagnosis. Proceedings of Computer Assisted Radiology and Surgery. 2000:822-827.
  45. Giger ML, Z Huo, CJ Vyborny, et al. Results of an observer study with an intelligent mammographic workstation for CAD. In: Peitgen HO, ed. Digital Mammography IWDM 2002. Berlin: Springer; 2003:297-303.
  46. Giger ML, Bae KT, MacMahon H. Computerized detection of pulmonary nodules in computed tomography images. Invest Radiol. 1994;29:459-465.
  47. Fiebich M, Wietholt C, Renger BC, et al. Automatic detection of pulmonary nodules in low-dose screening thoracic CT examinations. Proc SPIE. 1999;3661: 1434-1439.
  48. Lou SL, Chang CL, Lin KP, Chen TS. Object-based deformation technique for 3-D CT lung nodule detection. Proc SPIE. 1999;3661:1544-1552.
  49. Satoh H, Ukai Y, Niki N, et al. Computer aided diagnosis system for lung cancer based on retrospective helical CT image. Proc SPIE. 1999;3661:1324-1335.
  50. Taguchi H, Kawata Y, Niki N, et al. Helical CT images using curved surface morphology analysis. Proc SPIE. 1999; 3661:1307-1314.
  51. Armato SG, Giger ML, MacMahon H. Automated detection of lung nodules in CT scans: preliminary results. Med Phys. 2001;28:1552–1561.
  52. Brown MS, McNitt-Gray MF, Goldin JG, Suh RD, Sayre JW, Aberle DR. Patient-specific models for lung nodule detection and surveillance in CT images. IEEE Trans Med Imaging. 2001;20:1242-1250.
  53. Fan L, Novak CL, Qian J, Kohl G, Naidich DP. Automatic detection of lung nodules from multi-slice low-dose CT images. Proc SPIE. 2001;4322:1828-1835.
  54. Gurcan MN, Sahiner B, Petrick N, et al. Lung nodule detection on thoracic computed tomography images: preliminary evaluation of a computer-aided diagnosis system. Med Phys. 2002;29:2552-2558.
  55. Ko JP, Betke M. Chest CT: automated nodule detection and assessment of change over time-preliminary experience. Radiology. 2001;218:267-273.
  56. Lee Y, Hara T, Fujita H, Itoh S, Ishigaki T. Automated detection of pulmonary nodules in helical CT images based on an improved template-matching technique. IEEE Trans Med Imaging. 2001;20:595-604.
  57. Armato SG, Li F, Giger ML, MacMahon H, Sone S, Doi K. Lung cancer: performance of automated lung nodules detection applied to cancers missed in a CT screening program. Radiology. 2002;225:685-692.
  58. Armato SG, Altman MB, Wilkie J, et al. Automated lung nodule classification following automated nodule detection on CT: a serial approach. Med Phys. 2003;30:1188-1197.
  59. Armato SG, Altman MB, LaRiviere PJ. Automated detection of lung nodules in CT scans: effect of image reconstruction algorithm. Med Phys. 2003;30:461-472.
  60. Hara AK, Johnson CD, Reed JE, et al. Detection of colorectal polyps with CT colography: initial assessment of sensitivity and specificity. Radiology. 1997;205:59-65
  61. Dachman AH, Kuniyoshi JK, Boyle CM, et al. CT colonography with three-dimensional problem solving for detection of colonic polyps. AJR Am J Roentgenol. 1998;171:989-995.
  62. Royster AP, Fenlon HM, Clarke PD, Nunes DP, Ferrucci JT. CT colonoscopy of colorectal neoplasms: two-dimensional and three-dimensional virtual-reality techniques with colonoscopic correlation. AJR Am J Roentgenol. 1997;169:1237-1242.
  63. McFarland EG, Brink JA, Pilgram TK, et al. Spiral CT colonography: reader agreement and diagnostic performance with two- and three-dimensional image-display techniques. Radiology. 2001;218:375-383.
  64. Macari M, Megibow AJ. Pitfalls of using three-dimensional CT colonography with two-dimensional imaging correlation. AJR Am J Roentgenol. 2001;176:137-143. 65.
  65. Summers RM, Johnson CD, Pusanik LM, Malley JD, Youssef AM, Reed JE. Automated polyp detection at CT colonography: feasibility assessment in a human population. Radiology. 2001;219:51-59.
  66. Yoshida H, Masutani Y, MacEneaney P, Rubin DT, Dachman AH. Computerized detection of colonic polyps at CT colonography on the basis of volumetric features: pilot study. Radiology. 2002;222:327-336.
  67. Yoshida H, Nappi J. Three-dimensional computer-aided diagnosis scheme for detection of colonic polyps. IEEE Trans Med Imaging. 2001;20:1261-1274.
  68. Kiss G, Van Cleynenbreugel J, Thomeer M, Suetens P, Marchal G. Computer-aided diagnosis in virtual colonography via combination of surface normal and sphere fitting methods. Eur Radiol. 2002;12:77-81.
  69. Yoshida H, Nappi J, MacEneaney P, Rubin DT, Dachman AH. Computer-aided diagnosis scheme for detection of polyps at CT colonography. Radiographics. 2002;22:963-979.
  70. Nappi J, Dachman, MacEneaney P, Yoshida H. Automated knowledge-guided segmentation of colonic walls for computerized detection of polyps in CT colonography. J Comput Assist Tomogr. 2002;26:493-504.
  71. Yoshida H, Nappi J. Three-dimensional computer-aided diagnosis scheme for detection of colonic polyps. IEEE Trans Med Imaging. 2001;20:1261-1274.
  72. Nappi J, Yoshida H. Automated detection of polyps with CT colonography: evaluation of volumetric features for reduction of false-positive findings. Acad Radiol. 2002;9:386-397.
  73. Nappi J, Yoshida H. Feature-guided analysis for reduction of false positives in CAD of polyps for computed tomographic colonography. Med Phys. 2003;30:1592-1601.
  74. Nappi JJ, Frimmel H, Dachman AH, Yoshida H. Computerized detection of colorectal masses in CT colonography based on fuzzy merging and wall-thickening analysis. Med Phys. 2004;31:860-872.