(Radiographics. 2000;20:1169-1177.)
© RSNA, 2000
Improvement in Detection of Pulmonary Nodules: Digital Image Processing and Computer-aided Diagnosis1
Heber MacMahon, MD
1 From the Department of Radiology, University of Chicago, 5841 S Maryland Ave, Chicago, IL 60637. From the Opening Plenary Session at the 1999 RSNA scientific assembly. Received March 7, 2000; revision requested April 5; revision received and accepted April 27. Address correspondence to the author (e-mail: macm@midway.uchicago.edu).
Index Terms: Computers, diagnostic aid • Lung, nodule, 60.281 • Lung neoplasms, diagnosis, 60.30 • Radiography, digital, 60.1215 • Subtraction, digital, 60.1215 • Subtraction, dual-energy, 60.1299
Introduction
Visual detection of pulmonary nodules on screen-film radiographs is notoriously unreliable. Muhm and colleagues (1) reported that 90% of peripheral nodules and 75% of perihilar nodules detected during a lung cancer screening program were visible in retrospect on earlier radiographs. Austin and colleagues (2) reviewed 27 cases of potentially resectable lung cancer that were missed on chest radiographs. The average diameter of the missed lesions was 1.6 cm, and many were substantially larger. Recent results of the Early Lung Cancer Action Project confirmed the poor sensitivity of traditional analog radiography compared with computed tomography (CT) in the detection of pulmonary nodules (3). Of those cancers detected with CT in the first 1,000 patients, only one in four was detected on chest radiographs, though all but one were greater than 5 mm in diameter.
Digital radiography has the potential to improve the accuracy of nodule detection by virtue of its greater latitude, improved local contrast, and more consistent image quality in comparison with screen-film systems. Specific techniques, such as dual-energy subtraction, temporal subtraction, and automated nodule detection schemes, can improve diagnostic accuracy further, especially regarding nodules in the potentially resectable range of 6–2 cm in diameter.
Dual-Energy Subtraction Chest Radiography
Dual-energy subtraction radiography exploits the fact that structures containing calcium selectively attenuate the lower energy photons in the x-ray beam. By subtracting the ribs and other bones from the image and displaying them separately, this technique can improve diagnostic accuracy for nodules and other opacities. Two fundamentally different approaches have been used. One involves use of sequential x-ray exposures in rapid succession, at different kilovolt peak settings. This method has advantages, although the inevitable delay between the first and second exposure can introduce misregistration artifacts. The second approach involves use of a single exposure, which is recorded by two receptors separated by a filter. It is this single-exposure technique that is embodied in the first commercial device to employ energy subtraction for chest radiography (FCR 9501 ES; Fuji Medical Systems, Stanford, Conn).
The kilovolt-peak, or kVp, setting selected for an x-ray exposure represents the maximum or peak kilovoltage. The x-ray beam that traverses the patient actually consists of a wide spectrum of energies. Energy subtraction radiography takes advantage of selective attenuation of the lower components of the x-ray spectrum by high-atomic-energy materials, including calcium, mainly because of photoelectric reactions. At higher photon energies, the difference in x-ray attenuation between bone and soft tissues is almost entirely due to the difference in the number of Compton reactions. Therefore, bone contrast is maximized with a low-kilovolt-peak technique. A high-kilovolt-peak technique has been favored for conventional chest radiography, as it renders the bones relatively transparent.
In the case of the dual-energy chest unit, a single exposure is performed with standard technique. Instead of a single storage phosphor plate, two plates are used with a copper filter sandwiched between them (Fig 1). The full energy spectrum of the primary beam is recorded on the first plate in the usual manner. The first imaging plate also serves as a filter, preferentially absorbing lower energy components of the beam. Radiation that passes through the first plate undergoes further low-energy filtration (beam hardening) by the copper filter before it encounters the second plate. The image recorded by the second plate, therefore, consists mainly of the high-energy components of the beam. Because this is a high-kilovolt-peak image, the contrast between bone and calcium is markedly reduced compared with that on the image recorded on the first plate. The image acquired on the second plate also differs from that acquired on the first plate in that it is produced by fewer x-ray photons owing to substantial attenuation of the beam by the first plate and the copper filter. To compensate for this difference in radiation flux, the second image signal is boosted electronically. Pronounced quantum mottle, characteristic of a low-exposure image, is smoothed by means of an iterative processing procedure. The pixel values of the soft-tissue components of the images are equilibrated, and a weighted subtraction of the second from the first image is performed, which yields a difference image composed largely of low-energy-attenuation components such as calcium and bone. To produce the soft-tissue image, the average values of the bone components in the two images are equilibrated, and the images are subtracted to yield the soft-tissue components (manufacturer's recommendations, Fuji Medical Systems). In all, three posteroanterior images are produced: the standard image, the soft-tissue image, and the bone image (Fig 2).
View larger version (14K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1. Schematic represents dual-energy subtraction chest radiographic systems. Two storage phosphor plates, with an intervening copper filter, are used to record the image from a single exposure. The image on the second plate, which is derived from high-energy photons, is amplified and subtracted from the image on the first plate to produce separate soft-tissue and bone images.
|
|
View larger version (141K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2a. (a) Posteroanterior digital chest radiograph obtained with dual-energy subtraction chest unit. (b) Soft-tissue image. (c) Bone image shows small calcified granuloma (arrow).
|
|
View larger version (133K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2b. (a) Posteroanterior digital chest radiograph obtained with dual-energy subtraction chest unit. (b) Soft-tissue image. (c) Bone image shows small calcified granuloma (arrow).
|
|
View larger version (139K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2c. (a) Posteroanterior digital chest radiograph obtained with dual-energy subtraction chest unit. (b) Soft-tissue image. (c) Bone image shows small calcified granuloma (arrow).
|
|
When primary interpretation is performed with soft copy, it is necessary to develop a display that incorporates the dual-energy subtraction images with the standard posteroanterior and lateral images and facilitates rapid comparison with previous examinations, which may also include dual-energy subtraction images. We have implemented a workstation display that "stacks" dual-energy subtraction images behind the standard posteroanterior image for both current and previous examinations. The standard images are reviewed first, in the usual manner. Then, by clicking on a button on the tool bar, the dual-energy subtraction images are displayed. (This feature has proved to be a substantial incentive for using the workstation in preference to film for routine interpretation.)
Because dual-energy subtraction radiography separates the calcium and soft-tissue components of the thorax, it has obvious potential to improve the accuracy of detection of nodules and other focal opacities. Dual-energy subtraction images have been shown to be significantly superior in the detection of noncalcified pulmonary nodules compared with either screen-film or standard digital radiographs (4,5). At my institution, we have found that the dual-energy subtraction images have certain distinct advantages. The soft-tissue dual-energy subtraction images clearly improve detectability of any focal soft-tissue opacity, such as a nodule, that is partly or completely obscured by overlying bones (Fig 3). In some cases, the soft-tissue image can help characterize a lesion by revealing its margins more completely than the standard image does.
View larger version (107K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3a. (a) Posteroanterior digital chest radiograph shows poorly defined opacity (arrow) in the left middle part of the lung that is partly obscured by overlying structures. (b) Soft-tissue image clearly shows a nodule (primary lung carcinoma).
|
|
View larger version (78K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3b. (a) Posteroanterior digital chest radiograph shows poorly defined opacity (arrow) in the left middle part of the lung that is partly obscured by overlying structures. (b) Soft-tissue image clearly shows a nodule (primary lung carcinoma).
|
|
The bone image can help confirm the presence of calcification in benign pulmonary nodules or hilar lymph nodes. In most cases, the bone image serves to confirm the presence of calcification that was suspected on the basis of the standard image. However, we have seen examples in which the standard image revealed a nodule with no suggestion of calcification, and the bone image showed unequivocal calcification, which indicates a benign origin. Although small amounts of dystrophic calcification can be detected with thin-section CT in as much as 10% of lung cancers, we have not detected any examples with dual-energy subtraction radiography, and we suspect that most cases would contain insufficient quantities of calcium for detection with dual-energy subtraction radiography. Rib abnormalities that can mimic lung nodules, such as sclerotic metastases or bone islands, are also more clearly visible and more accurately characterized on dual-energy subtraction bone images. Calcified pleural plaques, which may be caused by asbestos exposure, can mimic soft-tissue abnormalities and can be mistaken for pulmonary nodules or consolidation when viewed en face. In such cases, the bone image clearly reveals the calcific nature of the abnormality (Fig 4).
View larger version (97K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4a. (a) Posteroanterior digital radiograph shows nodular opacity (arrow) lateral to the right hilum. (b) Bone image shows the calcific nature of the opacity (calcified pleural plaque).
|
|
View larger version (85K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4b. (a) Posteroanterior digital radiograph shows nodular opacity (arrow) lateral to the right hilum. (b) Bone image shows the calcific nature of the opacity (calcified pleural plaque).
|
|
Although energy subtraction performed at current screen-film dose levels provides useful diagnostic information, a moderately higher dose produces a substantial improvement in the quality of such images. For the posteroanterior view, we use 110 kV with a 10:1 grid and 8–10 mAs (range, 4–16 mAs) for an average-size patient. This technique would correspond approximately to a 200-speed screen-film system (we rate the Insight HC system [Eastman-Kodak, Rochester, NY] at a speed of 320). Because the majority of the total radiation dose to the patient is contributed by the lateral projection and because energy subtraction is not normally performed in lateral radiography, we use a relatively lower exposure for the lateral image. Although energy subtraction is not currently available with any of the commercial offerings that incorporate the newer flat-panel detector technologies, these devices have the potential for allowing high-quality dual-energy subtraction imaging.
In an observer test designed to evaluate the effect of dual-energy subtraction radiography on diagnostic accuracy, six attending radiologists and six radiology residents interpreted 46 chest radiographs, including 25 images of noncalcified nodules. Every observer had better accuracy with dual-energy subtraction radiography, and the difference in diagnostic accuracy was statistically significant for both attending radiologists and residents as a group (6).
Temporal Subtraction
Digital radiography allows various types of image enhancement to be performed, but techniques that improve the visibility of abnormal findings also tend to emphasize certain features of normal anatomy. In an ideal situation, the visibility of pathologic findings would be improved selectively, whereas normal anatomic structures would be suppressed. In the case of a patient with a previous chest radiograph, an opportunity exists to enhance selectively areas of interval change, including regions with new or altered pathologic conditions, by using the previous radiographs as a subtraction mask. The temporal subtraction technique involves automated two-dimensional warping and registration of a previous chest radiograph with a current chest radiograph to produce a difference image in which unchanged areas appear uniformly gray, while new opacities appear as isolated dark foci that stand out from the uniform background (7). Although the quality of the temporal subtraction image is affected by variations in patient positioning, this limitation can be partially overcome by the geometric warping that is performed (8).
One of the unique advantages of temporal subtraction is that it can highlight areas of subtle change that may not appear obviously abnormal when viewed in isolation. The ability of temporal subtraction to improve detection of a broad range of abnormalities, such as nodules, infiltrative opacities, and even alterations in local pulmonary perfusion, is an important advantage (Figs 5–7).
View larger version (146K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5a. (a) Posteroanterior baseline chest radiograph obtained in an asymptomatic patient who had previously undergone left upper lobectomy for lung cancer. (b) Chest radiograph obtained several years later shows a nodule in the right upper lobe that is obscured by the clavicle and was overlooked at the first reading. (c) Temporal subtraction image obtained by subtracting the previous (a) from the current (b) radiograph clearly shows the new nodule, which appears dark because of inversion of the gray scale of the current radiograph during the subtraction process.
|
|
View larger version (140K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5b. (a) Posteroanterior baseline chest radiograph obtained in an asymptomatic patient who had previously undergone left upper lobectomy for lung cancer. (b) Chest radiograph obtained several years later shows a nodule in the right upper lobe that is obscured by the clavicle and was overlooked at the first reading. (c) Temporal subtraction image obtained by subtracting the previous (a) from the current (b) radiograph clearly shows the new nodule, which appears dark because of inversion of the gray scale of the current radiograph during the subtraction process.
|
|
View larger version (155K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5c. (a) Posteroanterior baseline chest radiograph obtained in an asymptomatic patient who had previously undergone left upper lobectomy for lung cancer. (b) Chest radiograph obtained several years later shows a nodule in the right upper lobe that is obscured by the clavicle and was overlooked at the first reading. (c) Temporal subtraction image obtained by subtracting the previous (a) from the current (b) radiograph clearly shows the new nodule, which appears dark because of inversion of the gray scale of the current radiograph during the subtraction process.
|
|
View larger version (159K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6a. (a) Posteroanterior baseline chest radiograph shows a faint subpleural opacity (arrow) in the lower left lung and normal hila. (b) Chest radiograph obtained more than 2 years later shows a prominent left hilum and increased peripheral opacity. (c) Temporal subtraction image shows increased opacity in the left hilum secondary to lymphadenopathy and radiolucency throughout the left lower lobe due to altered perfusion. Normal right hemithorax appears uniform. (d) CT scan obtained several weeks after b was obtained helps confirm the presence of a hilar tumor mass.
|
|
View larger version (147K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6b. (a) Posteroanterior baseline chest radiograph shows a faint subpleural opacity (arrow) in the lower left lung and normal hila. (b) Chest radiograph obtained more than 2 years later shows a prominent left hilum and increased peripheral opacity. (c) Temporal subtraction image shows increased opacity in the left hilum secondary to lymphadenopathy and radiolucency throughout the left lower lobe due to altered perfusion. Normal right hemithorax appears uniform. (d) CT scan obtained several weeks after b was obtained helps confirm the presence of a hilar tumor mass.
|
|
View larger version (146K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6c. (a) Posteroanterior baseline chest radiograph shows a faint subpleural opacity (arrow) in the lower left lung and normal hila. (b) Chest radiograph obtained more than 2 years later shows a prominent left hilum and increased peripheral opacity. (c) Temporal subtraction image shows increased opacity in the left hilum secondary to lymphadenopathy and radiolucency throughout the left lower lobe due to altered perfusion. Normal right hemithorax appears uniform. (d) CT scan obtained several weeks after b was obtained helps confirm the presence of a hilar tumor mass.
|
|
View larger version (148K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6d. (a) Posteroanterior baseline chest radiograph shows a faint subpleural opacity (arrow) in the lower left lung and normal hila. (b) Chest radiograph obtained more than 2 years later shows a prominent left hilum and increased peripheral opacity. (c) Temporal subtraction image shows increased opacity in the left hilum secondary to lymphadenopathy and radiolucency throughout the left lower lobe due to altered perfusion. Normal right hemithorax appears uniform. (d) CT scan obtained several weeks after b was obtained helps confirm the presence of a hilar tumor mass.
|
|
View larger version (108K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7a. (a) Posteroanterior baseline chest radiograph. (b) Radiograph obtained 20 months later shows poorly defined opacity (arrow) inferior to the right hilum. (c) Temporal subtraction image enhances the increased opacity, indicating a significant interval change. (d) CT scan helps confirm the presence of a primary lung carcinoma (arrow) in the posterior right lower lobe.
|
|
View larger version (102K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7b. (a) Posteroanterior baseline chest radiograph. (b) Radiograph obtained 20 months later shows poorly defined opacity (arrow) inferior to the right hilum. (c) Temporal subtraction image enhances the increased opacity, indicating a significant interval change. (d) CT scan helps confirm the presence of a primary lung carcinoma (arrow) in the posterior right lower lobe.
|
|
View larger version (131K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7c. (a) Posteroanterior baseline chest radiograph. (b) Radiograph obtained 20 months later shows poorly defined opacity (arrow) inferior to the right hilum. (c) Temporal subtraction image enhances the increased opacity, indicating a significant interval change. (d) CT scan helps confirm the presence of a primary lung carcinoma (arrow) in the posterior right lower lobe.
|
|
View larger version (106K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7d. (a) Posteroanterior baseline chest radiograph. (b) Radiograph obtained 20 months later shows poorly defined opacity (arrow) inferior to the right hilum. (c) Temporal subtraction image enhances the increased opacity, indicating a significant interval change. (d) CT scan helps confirm the presence of a primary lung carcinoma (arrow) in the posterior right lower lobe.
|
|
In a controlled experiment, 11 observers interpreted 50 chest radiographs that included 29 lungs with new opacities, including nodules. Previous radiographs were provided for comparison in each case. Findings in the same cases were interpreted on a different occasion with the benefit of temporal subtraction images in addition to the current and previous radiographs. The observers as a group showed a substantial and highly significant improvement in diagnostic accuracy when temporal subtraction images were used. Additionally, the average interpretation time was reduced by 19% when temporal subtraction images were used (9).
In a similar observer test, 32 radiologists and 12 radiology residents interpreted 20 chest radiographs that showed 13 lungs with cancer and 27 without. All groups showed a highly significant improvement in accuracy when temporal subtraction images were used, and residents were significantly more accurate with temporal subtraction images than were chest radiologists without temporal subtraction images. Even chest radiologists, who had the lowest average error rate with unaided detection, showed a significant improvement with temporal subtraction images (10).
Because temporal subtraction enhances any type of pathologic change, its diagnostic benefits are complementary to other computer-aided diagnostic techniques, such as energy subtraction and computer-aided nodule detection. Temporal subtraction is currently being investigated in the setting of a lung cancer screening program in Japan (11).
Automated Lung Nodule Detection
It is evident from the reports of Muhm et al (1) and Austin et al (2) that many potentially curable cancers that are overlooked by radiologists are clearly visible in retrospect. Failure of a radiologist to focus on the abnormality has been shown to be a contributing cause in such cases (12). The purpose of computer-aided diagnosis in nodule detection is to direct the radiologist's attention by identifying and indicating suspect focal opacities on a radiograph that may represent cancer (Fig 8). A nodule detection program developed at the University of Chicago consists of five basic steps (13). The initial step involves a filtering operation that enhances nodular opacities on the image. A second filter is applied to the original image to suppress nodular opacities. From these two images, a difference image is produced on which the complex anatomic background is minimized while nodular structures are maximized. Next, multiple gray-level thresholds are determined on the basis of the histogram of the difference image to identify "nodule candidates." The nodule candidates are then classified according to the threshold levels at which the nodule candidates are identified. Various image features are extracted from the difference image and the original image by means of region-growing techniques and edge-gradient analysis to separate spurious from real nodules. The image features include contrast, effective diameter, degree of circularity or irregularity, and rates of change of the effective diameter and degrees of circularity and irregularity as gray-level thresholds are varied. Finally, a rule-based analysis is applied to reduce the number of false-positive results.
View larger version (125K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8. Automated nodule detection. Digital radiograph shows results of a computer-aided diagnostic program that indicates two suspect areas (arrows). One area (right upper lobe) contains an actual nodule, and the other contains a typical false-positive result due to normal anatomic structures.
|
|
In a large observer test performed at an annual meeting of the Radiological Society of North America (RSNA), 20 abnormal chest radiographs containing a nodule and 20 normal chest radiographs were included. One hundred forty-six observers interpreted the cases on a computer workstation, first without and then with the benefit of the nodule detection program. The results were evaluated by means of receiver operating characteristic analysis to determine the effect of the nodule detection program on diagnostic accuracy (14).
Detection accuracy achieved by individuals, on average, correlated with their experience and degree of specialization. Thus, chest radiologists achieved the highest accuracy and were followed by other radiologists, radiology residents, and nonradiologists. Detection accuracy improved significantly for each group when they used the computer-aided diagnostic program. As with dual-energy subtraction and temporal subtraction, radiology residents with the nodule detection program had slightly (though not significantly) better accuracy than did chest radiologists without the computer-aided diagnostic program.
This nodule detection program has an average sensitivity of 70%–80%, with an average of one to two false-positive findings per radiograph. Although some obvious nodules are still missed with the program and false-positive findings are numerous, even experienced chest radiologists benefited from its use in the observer test. Although accuracy of the computer program alone is less than that of most radiologists, the errors of the program tend to differ from those of human observers. Consequently, use of the nodule detection program can increase the accuracy of even experienced radiologists.
There have been concerns that the false-positive results generated by such programs might lead to overdiagnosis. The results of this observer test, as well as clinical experience with a mammography computer-aided diagnostic program, suggest that radiologists can readily dismiss the majority of computer-generated false-positive findings and that specificity does not usually suffer (15).
Clinical Implementation of Computer-aided Diagnosis
A computer-aided diagnostic system for mammography has been developed on the basis of film digitization with a combination of hard copy interpretation and soft copy computer-aided diagnostic display (16). Although this approach may have certain niche applications for the chest (ie, screening programs for cancer or industrial lung disease), it seems unlikely that a film-based system would be widely accepted. Particularly when so many hospitals are converting to digital image acquisition, it would be logical to implement chest computer-aided diagnosis in an all-digital workstation environment. In a picture archiving and communication system, or PACS, chest computer-aided diagnostic programs could be applied to images as soon as they were acquired. The results could be stored as overlays that would be available for immediate display at the workstation.
A simple and intuitive user interface is essential for widespread acceptance of image processing and computer-aided diagnosis in a high-volume examination such as chest radiography. Such an interface would allow images to be evaluated initially in the traditional way, prior to rapid display of dual-energy subtraction, temporal subtraction, or nodule detection results. As advanced image processing systems develop, it is likely that a greater degree of integration will occur between the various focused schemes that currently exist.
Dual-energy subtraction is already available and in routine use at many sites. Temporal subtraction and nodule detection by means of computer-assisted diagnosis are still limited to research applications, although they are likely to become available for clinical use commercially during the next few years. As interpretation migrates from the traditional view box to the computer workstation, these diagnostic tools will become more widely accepted by practicing radiologists. While such "enhanced" digital radiography systems cannot be expected to approach the sensitivity of CT for small nodules, they should substantially improve detection accuracy compared with that of traditional screen-film systems.
References
-
Muhm JR, Miller WE, Fontana RS, et al. Lung cancer detected during a screening program using 4-month chest radiographs. Radiology 1983; 148:609-615.[Abstract/Free Full Text]
-
Austin JHM, Romney BM, Goldsmith LS. Missed bronchogenic carcinoma: radiographic findings in 27 patients with a potentially resectable lesion evident in retrospect. Radiology 1992; 182:115-122.[Abstract/Free Full Text]
-
Henschke CI, McCauley DI, Yankelevitz DF, et al. Early lung cancer action project: overall design and findings from baseline screening. Lancet 1999; 354:99-105.[Medline]
-
Hartman TE. Dual-energy radiography. Semin Roentgenol 1997; 32:45-49.[Medline]
-
Kido S, Ikezoe J, Naito H, et al. Clinical evaluation of pulmonary nodules with single-exposure dual-energy subtraction chest radiography with an iterative noise-reduction algorithm. Radiology 1995; 194:407-412.[Abstract/Free Full Text]
-
MacMahon H, Cannon W, Engelmann RM, Carlin M, Doi K. Dual-energy subtraction computed chest radiography: comparison of diagnostic accuracy with conventional computed radiography (abstr). Radiology 1998; 209(P):544.
-
Kano A, Doi K, MacMahon H, Hassell D, Giger M. Digital image subtraction of temporally sequential chest images for detection of interval change. Med Phys 1994; 21:453-461.[Medline]
-
Ishida T, Ashizawa K, Engelmann R, Katsuragawa S, MacMahon H, Doi K. Application of temporal subtraction for detection of interval change in chest radiographs: improvement of subtraction images using automated initial image matching. J Digital Imaging 1999; 12:77-86.[Medline]
-
Difazio MC, MacMahon H, Xu XW, et al. Digital chest radiography: effect of temporal subtraction images on detection accuracy. Radiology 1997; 202:447-452.[Abstract/Free Full Text]
-
MacMahon H, Engelmann RM, Roe C, Behlen FM, Hoffmann KR, Doi K. Use of temporal subtraction for detection of early lung cancer: results of observer test (abstr). Radiology 1998; 209(P):543.[Abstract/Free Full Text]
-
Katsuragawa S, Sasaki Y, MacMahon H, Ishida T, Doi K. Application of temporal subtraction to screening chest radiographs with a mobile computed radiography system. In: Doi K, MacMahon H, Giger ML, Hoffmann KR, eds. Computer-aided diagnosis in medical imaging. Amsterdam, the Netherlands: Elsevier, 1999; 51-56.
-
Kundel HL, Nodine CF, Carmody D. Visual scanning, pattern recognition and decision-making in pulmonary nodule detection. Invest Radiol 1978; 13:175-181.[Medline]
-
Kobayashi T, Xu XW, MacMahon H, Metz CE, Doi K. Effect of computer-aided diagnosis scheme on radiologists' performance in detection of lung nodules on radiographs. Radiology 1996; 199:843-848.[Abstract/Free Full Text]
-
MacMahon H, Engelmann R, Behlen FM, et al. Computer-aided diagnosis of pulmonary nodules: results of a large-scale observer test. Radiology 1999; 213:723-726.[Abstract/Free Full Text]
-
Doi T, Hasegawa A, Hunt B, et al. Clinical results with the R2 ImageChecker mammographic CAD system. In: Doi K, MacMahon H, Giger ML, Hoffmann KR, eds. Computer-aided diagnosis in medical imaging. Amsterdam, the Netherlands: Elsevier, 1999; 201-207.
-
Nishikawa RM, Giger ML, Wolverton DE, Schmidt RA, Doi K. Prospective testing of a clinical CAD workstation for the detection of breast lesions on mammograms. In: Doi K, MacMahon H, Giger ML, Hoffmann KR, eds. Computer-aided diagnosis in medical imaging. Amsterdam, the Netherlands: Elsevier, 1999; 209-214.
This article has been cited by other articles:
|
|
|
|
 
F. Girvin and J. P. Ko
Pulmonary Nodules: Detection, Assessment, and CAD
Am. J. Roentgenol.,
October 1, 2008;
191(4):
1057 - 1069.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
M. K. Gould, J. Fletcher, M. D. Iannettoni, W. R. Lynch, D. E. Midthun, D. P. Naidich, and D. E. Ost
Evaluation of Patients With Pulmonary Nodules: When Is It Lung Cancer?: ACCP Evidence-Based Clinical Practice Guidelines (2nd Edition)
Chest,
September 1, 2007;
132(3_suppl):
108S - 130S.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
J. E. Kuhlman, J. Collins, G. N. Brooks, D. R. Yandow, and L. S. Broderick
Dual-Energy Subtraction Chest Radiography: What to Look for beyond Calcified Nodules
RadioGraphics,
January 1, 2006;
26(1):
79 - 92.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
G. D. Rubin, J. K. Lyo, D. S. Paik, A. J. Sherbondy, L. C. Chow, A. N. Leung, R. Mindelzun, P. K. Schraedley-Desmond, S. E. Zinck, D. P. Naidich, et al.
Pulmonary Nodules on Multi-Detector Row CT Scans: Performance Comparison of Radiologists and Computer-aided Detection
Radiology,
January 1, 2005;
234(1):
274 - 283.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
C. Schaefer-Prokop and M. Prokop
New imaging techniques in the treatment guidelines for lung cancer
Eur. Respir. J.,
February 1, 2002;
19(35_suppl):
71S - 83s.
[Abstract]
[Full Text]
[PDF]
|
|
|
