Feng Li

University of Chicago, Chicago, IL, United States

Are you Feng Li?

Claim your profile

Publications (98)139.95 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: OBJECTIVE: To evaluate radiologists' ability to detect focal pneumonia by use of standard chest radiographs alone compared with standard plus bone-suppressed chest radiographs. METHODS: Standard chest radiographs in 36 patients with 46 focal airspace opacities due to pneumonia (10 patients had bilateral opacities) and 20 patients without focal opacities were included in an observer study. A bone suppression image processing system was applied to the 56 radiographs to create corresponding bone suppression images. In the observer study, eight observers, including six attending radiologists and two radiology residents, indicated their confidence level regarding the presence of a focal opacity compatible with pneumonia for each lung, first by use of standard images, then with the addition of bone suppression images. Receiver operating characteristic (ROC) analysis was used to evaluate the observers' performance. RESULTS: The mean value of the area under the ROC curve (AUC) for eight observers was significantly improved from 0.844 with use of standard images alone to 0.880 with standard plus bone suppression images (P < 0.001) based on 46 positive lungs and 66 negative lungs. CONCLUSION: Use of bone suppression images improved radiologists' performance for detection of focal pneumonia on chest radiographs. KEY POINTS : • Bone suppression image processing can be applied to conventional digital radiography systems. • Bone suppression imaging (BSI) produces images that appear similar to dual-energy soft tissue images. • BSI improves the conspicuity of focal lung disease by minimizing bone opacity. • BSI can improve the accuracy of radiologists in detecting focal pneumonia.
    European Radiology 07/2012; · 4.34 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Managing and supervising the complex imaging examinations performed for clinical research in an academic medical center can be a daunting task. Coordinating with both radiology and research staff to ensure that the necessary imaging is performed, analyzed, and delivered in accordance with the research protocol is nontrivial. The purpose of this communication is to report on the establishment of a new Human Imaging Research Office (HIRO) at our institution that provides a dedicated infrastructure to assist with these issues and improve collaborations between radiology and research staff. The HIRO was created with three primary responsibilities: 1) coordinate the acquisition of images for clinical research per the study protocol, 2) facilitate reliable and consistent assessment of disease response for clinical research, and 3) manage and distribute clinical research images in a compliant manner. The HIRO currently provides assistance for 191 clinical research studies from 14 sections and departments within our medical center and performs quality assessment of image-based measurements for six clinical research studies. The HIRO has fulfilled 1806 requests for medical images, delivering 81,712 imaging examinations (more than 44.1 million images) and related reports to investigators for research purposes. The ultimate goal of the HIRO is to increase the level of satisfaction and interaction among investigators, research subjects, radiologists, and other imaging professionals. Clinical research studies that use the HIRO benefit from a more efficient and accurate imaging process. The HIRO model could be adopted by other academic medical centers to support their clinical research activities; the details of implementation may differ among institutions, but the need to support imaging in clinical research through a dedicated, centralized initiative should apply to most academic medical centers.
    Academic radiology 04/2012; 19(6):762-71. · 2.09 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE To evaluate tumor enhancement for malignant pleural mesothelioma (MPM) on contrast-enhanced computed tomography (CT) scans in clinical trials. METHOD AND MATERIALS A modified RECIST protocol was applied to 16 patients with MPM who participated in a clinical trial. Each patient had between 2 and 18 CT scans for a total of 94 scans. Among the 94 CT scans, 14 baseline and 64 follow-up scans were contrast enhanced. For each scan, a radiologist obtained the mean Hounsfield Unit (HU) value of the tumor in the single CT section demonstrating the largest tumor area. The mean HU value of the descending aorta was also measured in the same section. Among the 64 enhanced follow-up scans, the 48 scans that generated the same RECIST response category between the clinical measurement and a subsequent verification measurement were used in this study. RECIST measurements were used to classify each tumor response as either partial response (PR), progressive disease (PD), or stable disease (SD). The mean tumor HU value and normalized enhancement ratio (mean tumor HU divided by mean aorta HU) in each follow-up scan were compared with the corresponding baseline scan measurements. RESULTS For the 48 follow-up scans, mean differences in the normalized enhancement ratio were -0.14 for the 21 tumors with a RECIST response category of PR, -0.11 for the 16 SD tumors, and 0.02 for the 11 PD tumors. P-values for normalized enhancement ratio between PR and PD and between SD and PD were 0.004 and 0.01, respectively. Mean differences in the mean tumor HU values were -15 for PR tumors, -14 for SD tumors, and 18 for PD tumors; however, these differences were not statistically significant. CONCLUSION Normalized tumor enhancement is related to tumor response in mesothelioma patients. CLINICAL RELEVANCE/APPLICATION Normalized tumor enhancement could be utilized to improve response assessment in clinical trials.
    Radiological Society of North America 2011 Scientific Assembly and Annual Meeting; 12/2011
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE To compare protocol-specific and clinical radiologists’ measurements based on the Response Evaluation Criteria in Solid Tumors (RECIST) for malignant pleural mesothelioma (MPM) in clinical trials. METHOD AND MATERIALS A modified RECIST tumor response assessment protocol was applied to 16 patients with MPM who participated in a clinical trial. Each patient had between 2 and 18 computed tomography (CT) scans for a total of 94 scans. This protocol included (1) transverse 3-mm CT scans, (2) the acquisition of tumor thickness measurements perpendicular to the chest wall, and (3) two measurement positions on each of three CT sections separated by at least 10 mm (six linear measurements total). RECIST measurements were used to classify each tumor response as either partial response (PR), progressive disease (PD), or stable disease (SD). All 16 baseline and 78 follow-up CT scans were interpreted and measured by clinical radiologists during the standard clinical workflow. The clinical radiologists were aware of the RECIST protocol but were not specifically trained in its application. Subsequently, a trained research radiologist retrospectively (and without knowledge of the clinical interpretation) acquired tumor measurements in strict adherence to the protocol guidelines. Differences between the protocol-specific and clinical tumor measurements were calculated. RESULTS Four chest radiologists provided clinical measurements for 52 of the CT scans, and 27 other radiologists or residents measured the other 42 scans. Among the 16 baseline CT scans, only three scans were measured clinically at the prescribed total of six tumor locations (two positions on each of three sections); the remaining 13 scans had been measured at between 3-8 positions on 3-4 sections. Mean summed tumor size was 43 mm (95% CI: 34-75 mm) by protocol and 44 (95% CI; 23-86 mm) by clinical measurement (P=0.94). Based on these measurements, the RECIST response category for the protocol-specific and clinical interpretations was the same in 58 (74%) of the 78 follow-up scans but different for 20 (26%) scans. CONCLUSION Strict adherence to protocol tumor measurement criteria yields tumor response categories that may differ from those obtained when measurements are performed during routine clinical workflow. CLINICAL RELEVANCE/APPLICATION Tumor measurements must be made in a consistent and reliable manner to achieve meaningful results in clinical trials.
    Radiological Society of North America 2011 Scientific Assembly and Annual Meeting; 11/2011
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE To evaluate a computer-aided detection (CAD) system for detecting nodules on chest radiographs (CXRs), and to compare its performance with that of previous versions. METHOD AND MATERIALS A new CAD system (Riverain Medical, OnGuard 5.2) that incorporates bone suppression technology, was applied to posteroanterior (PA) CXRs of 83 adult patients. The 83 radiographs included 78 primary lung cancers and 10 solitary metastatic nodules (5 patients with 2 nodules in different lungs). The mean size of the 88 nodules was 17.8 mm (range, 7-30 mm), including 50 nodules between 7-19 mm (small) and 38 nodules between 20-30 mm (large). Nodule subtlety was rated independently by two experienced radiologists on a 1-10 scale (from “extremely subtle” to “extremely obvious”), including 16 nodules with mean ratings of 1-1.5 (extremely subtle), 60 from 2-4.5 (very subtle/subtle), and 12 from 5-7 (more obvious). RESULTS CAD marked 66 of 88 nodules (sensitivity: 75%) with an average of 0.5 FP marks per image. Sensitivity was 70% for small (7-19 mm) and 82% for larger (20-30 mm) nodules. For very subtle, subtle, and more obvious nodules, the sensitivities were 25%, 83% and 100% respectively. The sensitivity was 52% for clinically unreported malignant nodules. Among the 43 FP marks, 47% were caused by benign opacities such as scars, and the rest were due to calcified cartilage, or summation opacities. These results represent a substantial improvement over those achieved on the same database with the first FDA approved version of this CAD system, which had a sensitivity of 53% with an average of 5.8 false positives per radiograph. CONCLUSION Nodule detection CAD on chest radiographs can now achieve high sensitivity with greatly improved specificity compared with earlier versions. When those cancers that were graded as extremely subtle or very subtle were excluded, CAD sensitivity was 86%, with an average of 0.5 false positive marks per CXR. CLINICAL RELEVANCE/APPLICATION CAD for chest radiographs has substantially improved and may provide critical protection against common types of potentially serious oversight errors.
    Radiological Society of North America 2011 Scientific Assembly and Annual Meeting; 11/2011
  • [Show abstract] [Hide abstract]
    ABSTRACT: To determine whether use of bone suppression (BS) imaging, used together with a standard radiograph, could improve radiologists' performance for detection of small lung cancers compared with use of standard chest radiographs alone and whether BS imaging would provide accuracy equivalent to that of dual-energy subtraction (DES) radiography. Institutional review board approval was obtained. The requirement for informed consent was waived. The study was HIPAA compliant. Standard and DES chest radiographs of 50 patients with 55 confirmed primary nodular cancers (mean diameter, 20 mm) as well as 30 patients without cancers were included in the observer study. A new BS imaging processing system that can suppress the conspicuity of bones was applied to the standard radiographs to create corresponding BS images. Ten observers, including six experienced radiologists and four radiology residents, indicated their confidence levels regarding the presence or absence of a lung cancer for each lung, first by using a standard image, then a BS image, and finally DES soft-tissue and bone images. Receiver operating characteristic (ROC) analysis was used to evaluate observer performance. The average area under the ROC curve (AUC) for all observers was significantly improved from 0.807 to 0.867 with BS imaging and to 0.916 with DES (both P < .001). The average AUC for the six experienced radiologists was significantly improved from 0.846 with standard images to 0.894 with BS images (P < .001) and from 0.894 to 0.945 with DES images (P = .001). Use of BS imaging together with a standard radiograph can improve radiologists' accuracy for detection of small lung cancers on chest radiographs. Further improvements can be achieved by use of DES radiography but with the requirement for special equipment and a potential small increase in radiation dose.
    Radiology 09/2011; 261(3):937-49. · 6.34 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: To determine whether liver function correlating with indocyanine green (ICG) clearance could be estimated quantitatively from gadoxetate disodium-enhanced magnetic resonance (MR) images. This retrospective study was approved by the institutional review board, and the requirement for informed consent was waived. Twenty-three consecutive patients who underwent an ICG clearance test and gadoxetate disodium-enhanced MR imaging with the same parameters as were used for a preoperative examination were chosen. The hepatocellular uptake index (HUI) from liver volume (V(L))and mean signal intensity of the liver on contrast-enhanced T1-weighted images with fat suppression (L(20)) and mean signal intensity of the spleen on contrast-enhanced T1-weighted images with fat suppression (S(20)) on 3D gradient-echo T1-weighted images with fat suppression obtained at 20 minutes after gadoxetate disodium (0.025 mmol per kilogram of body weight) administration was determined with the following equation: V(L)[(L(20)/S(20)) - 1]. The correlation of the plasma disappearance rate of ICG (ICG-PDR) and various factors derived from MR imaging, including HUI, iron and fat deposition in the liver and spleen, and spleen volume (V(S)), were evaluated with stepwise multiple regression analysis. The difference between the ratio of the remnant HUI to the HUI of the total liver (rHUI/HUI) and ratio of the liver remnant V(L) to the total V(L) (rV(L)/V(L)) was evaluated in four patients who had segmental heterogeneity of liver function. HUI and V(S) were the factors significantly correlated with ICG-PDR (R = 0.87). The mean value and its 95% confidence interval were 0.18 and 0.01 to 0.34, respectively, for the following calculation: (rHUI/HUI) - (rV(L)/V(L)). The liver function correlating with ICG-PDR can be estimated quantitatively from the signal intensities and the volumes of the liver and spleen on gadoxetate disodium-enhanced MR images, which may improve the estimation of segmental liver function.
    Radiology 06/2011; 260(3):727-33. · 6.34 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The purpose of this article is to evaluate radiologists' ability to detect subtle nodules by use of standard chest radiographs alone compared with bone suppression imaging used together with standard radiographs. The cases used in this observer study comprised radiographs of 72 patients with a subtle nodule and 79 patients without nodules taken from the Japanese Society of Radiological Technology nodule database. A new image-processing system was applied to the 151 radiographs to create corresponding bone suppression images. Two image reading sets were used with an independent test method. The first reading included half of the patients (a randomly selected subset A) showing only the standard image and the remaining half (subset B) showing the standard image plus bone suppression images. The second reading entailed the same subsets; however, subset A was accompanied by bone suppression images, whereas subset B was shown with only the standard image. The two image sets were read by three experienced radiologists, with an interval of more than 2 weeks between the sessions. Receiver operating characteristic (ROC) curves, with and without localization, were obtained to evaluate the observers' performance. The mean value of the area under the ROC curve for the three observers was significantly improved, from 0.840 with standard radiographs alone to 0.863 with additional bone suppression images (p = 0.01). The area under the localization ROC curve was also improved with bone suppression imaging. The use of bone suppression images improved radiologists' performance in the detection of subtle nodules on chest radiographs.
    American Journal of Roentgenology 05/2011; 196(5):W535-41. · 2.90 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE A reliable method for quantitative evaluation of liver function with anatomic details in humans has not been established to date. Gadoxetate disodium is a targeted MR contrast agent that can combine two features for extracellular and hepatocyte-specific agents. Our objective in this study was to evaluate liver function quantitatively by use of a newly developed computerized method on gadoxetate disodium-enhanced MR images. METHOD AND MATERIALS The plasma disappearance rate of indocyanine green (ICG-PDR) and 3D-GRE T1-weighted images with fat suppression obtained at 20 min after gadoxetate disodium administration were obtained from 16 patients. The manual outlines of the whole liver and spleen were obtained to measure the liver volume (V) and the signal intensity of the liver (L20) and spleen (S20) from gadoxetate disodium-enhanced MR images for comparison with a computerized method using local thresholding technique. The time for drawing outlines, and the correlation coefficient between ICG-PDR and various feature values including V(L20 - S20)/S20 were compared for the manual and the computerized methods. The informed consent requirement was waived, and this retrospective study was approved by the Institutional Review Board. RESULTS A significantly high correlation was observed between ICG-PDR and V(L20 – S20)/S20 by the computerized method (r = 0.897, 95%CI = 0.897 - 0.900) and by the manual method (r = 0.899, 95%CI = 0.898 - 0.900), whereas the correlation coefficient between ICG-PDR and L20 was 0.813 (95%CI = 0.808 - 0.818) by the computerized method and 0.833 (95%CI = 0.828 - 0.839) by the manual method. The mean time for drawing outlines was significantly shorter by the computerized method (3.62 min, 95%CI = 2.07 - 5.16) than by the manual method (14.3 min, 95%CI = 12.7 - 15.9). CONCLUSION The liver function corresponding to ICG-PDR can be estimated quantitatively and efficiently by use of the computerized method from the signal intensity of the liver on gadoxetate disodium-enhanced MR images with appropriate corrections for the liver volume and the extracellular contrast effect of gadoxetate disodium approximated by the signal intensity of the spleen. CLINICAL RELEVANCE/APPLICATION The liver function corresponding to ICG-PDR can be estimated quantitatively and efficiently by use of the computerized method from the gadoxetate disodium-enhanced MR imaging.
    Radiological Society of North America 2010 Scientific Assembly and Annual Meeting; 11/2010
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE Interstitial lung disease (ILD) is a common pulmonary pathology that can cause high morbidity and mortality in patients. Disease extent is one of the most important clinical indices for quantifying ILD and is defined as the ratio of (a) abnormal lung volume and (b) entire lung volume. We have developed an algorithm to accurately measure the entire volumes of lungs with severe ILD (agreement with reference lungs = 98.1%). In this study, we develop a scheme to automatically detect abnormal lung volumes in CT. METHOD AND MATERIALS We collected 31 normal CT scans and 37 abnormal CT scans with ILD at the University of Chicago Medical Center with Philips Brilliance scanners (120 - 140 kVp, 200 - 400 mAs, and 1 - 3 mm slice thickness). There were 28 cases with ground glass opacity, 28 with linear reticular, and 15 with honeycombing patterns. An expert chest radiologist manually delineated the abnormal lung volumes as reference standards. We first used image density to segment normal lung areas and texture information to segment abnormal lung areas. The airways were identified and tracked from lung apices to bases, and were removed from lungs to improve the accuracy of lung segmentation. We then automatically created many volumes of interest (VOIs) over the segmented lung regions and tried to classify these VOIs into normal and abnormal classes. We developed 5 new run-length matrix features to accurately characterize VOIs. We also employed 2 image density features and 12 texture features to represent additional information in VOIs. We systematically investigated the effect of various parameters in detecting abnormal VOIs, including different combinations of features, different statistical classification algorithms, and size and dimension of VOIs. A leave-one-case-out validation scheme was employed to evaluate the performance of our detection scheme. RESULTS The sensitivity and specificity for detecting abnormal VOIs were 86% and 90%, respectively, and the area under receiver operating characteristic curve was 0.92 when we used linear discriminant analysis with 12 features and 64×64×64 matrix size for VOIs. CONCLUSION Our scheme achieved a high performance in detecting abnormal lung volumes with ILD. CLINICAL RELEVANCE/APPLICATION Accurate detection of abnormal lung volumes with ILD is an important technique for quantifying disease extent of ILD in clinical practice.
    Radiological Society of North America 2010 Scientific Assembly and Annual Meeting; 11/2010
  • Jiahui Wang, Feng Li, Qiang Li
    [Show abstract] [Hide abstract]
    ABSTRACT: We developed an automated method for the segmentation of lungs with severe diffuse interstitial lung disease (DILD) in multi-detector CT. In this study, we would like to compare the performance levels of this method and a thresholdingbased segmentation method for normal lungs, moderately abnormal lungs, severely abnormal lungs, and all lungs in our database. Our database includes 31 normal cases and 45 abnormal cases with severe DILD. The outlines of lungs were manually delineated by a medical physicist and confirmed by an experienced chest radiologist. These outlines were used as reference standards for the evaluation of the segmentation results. We first employed a thresholding technique for CT value to obtain initial lungs, which contain normal and mildly abnormal lung parenchyma. We then used texture-feature images derived from co-occurrence matrix to further segment lung regions with severe DILD. The segmented lung regions with severe DILD were combined with the initial lungs to generate the final segmentation results. We also identified and removed the airways to improve the accuracy of the segmentation results. We used three metrics, i.e., overlap, volume agreement, and mean absolute distance (MAD) between automatically segmented lung and reference lung to evaluate the performance of our segmentation method and the thresholding-based segmentation method. Our segmentation method achieved a mean overlap of 96.1%, a mean volume agreement of 98.1%, and a mean MAD of 0.96 mm for the 45 abnormal cases. On the other hand the thresholding-based segmentation method achieved a mean overlap of 94.2%, a mean volume agreement of 95.8%, and a mean MAD of 1.51 mm for the 45 abnormal cases. Our new method obtained higher performance level than the thresholding-based segmentation method.
    Proc SPIE 03/2010;
  • [Show abstract] [Hide abstract]
    ABSTRACT: Indocyanine green (ICG) is widely used for its clearance test in the evaluation of liver function. Gadoxetate disodium (Gd-EOB-DTPA) is a targeted MR contrast agent partially taken up by hepatocytes. The objective of this study was to evaluate the feasibility of an estimation of the liver function corresponding to plasma disappearance rate of indocyanine green (ICG-PDR) by use of the signal intensity of the liver alone in Gd-EOB-DTPA enhanced MR imaging (EOB-MRI). We evaluated fourteen patients who had EOB-MRI and ICG clearance test within 1 month. 2D-GRE T1 weighted images were obtained at pre contrast, 3 min (equilibrium phase) and 20 min (hepatobiliary phase) after the intravenous administration of Gd-EOB-DTPA, and the mean signal intensity of the liver was measured. The correlation between ICG-PDR and many parameters derived from the signal intensity of the liver in EOB-MRI was evaluated. The correlation coefficient between ICG-PDR and many parameters derived from the signal intensity of the liver in EOBMRI was low and not significant. The estimation of the liver function corresponding to ICG-PDR by use of the signal intensity of the liver alone in EOB-MRI would not be reliable.
    Proc SPIE 01/2010; 7626(1):762604.
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE To evaluate the conspicuity of small lung nodules by use of new bone suppression chest images compared with standard chest radiographs. METHOD AND MATERIALS The cases used in this study consisted of 142 patients (62 men, 80 women; mean age, 60 years) with a solitary lung nodule on standard chest radiographs collected by the Japanese Society of Radiological Technology, including 91 malignant and 51 benign nodules. The mean size of these nodules was 17 mm (range 6-30 mm). There were 127 located in the peripheral lung, and 15 obscured by heart or diaphragm. The nodule subtlety in 5 categories (1: very subtle, n=25; 2: subtle, n=25; 3: relatively subtle, n=47; 4: obvious, n=35; 5: very obvious, n=10) had been previously rated by radiologists. A new image processing system (SoftView from Riverain Medical), which can suppress the conspicuity of bones without increasing x-ray dose, was applied to the 142 standard radiographs to create corresponding bone suppression images. The nodule contrast was defined as the difference in the pixel values between the nodule area (from the contours of nodules that were drawn by two radiologists) and the area around the nodules. The improved conspicuity of nodules on the bone suppression images was categorized and analyzed based on nodule malignancy classification, nodule location, and nodule subtlety categories. RESULTS The mean contrast was 530 on the standard image vs. 597 on the bone suppression images for all nodules, including 535 vs. 598 for the 91 malignant nodules, 522 vs. 595 for the 51 benign nodules, respectively (all P < 0.001). Regarding nodule location, the mean contrast was 537 vs. 609 for the 127 peripheral lung nodules (P < 0.001), and 471 vs. 501 for the 15 nodules obscured by heart or diaphragm (P = 0.12) for the standard and bone suppression images, respectively. When subtlety was considered, the mean contract was 593 vs. 703 for the 45 obvious nodules (categories 4 and 5) and 501 vs. 548 for the 97 subtle nodules (subtlety 1, 2 and 3) (both P < 0.001) for the standard and bone suppression images, respectively. For the 25 very subtle nodules, the mean contrast was 407 and 455, respectively (P = 0.03). CONCLUSION Bone suppression chest images can significantly improve the conspicuity of small lung nodules, even for very subtle lesions. CLINICAL RELEVANCE/APPLICATION The conspicuity of small lung nodules on chest radiographs can be improved by creating bone suppression images.
    Radiological Society of North America 2009 Scientific Assembly and Annual Meeting; 11/2009
  • [Show abstract] [Hide abstract]
    ABSTRACT: Accurate detection of diffuse lung disease is an important step for computerized diagnosis and quantification of this disease. It is also a difficult clinical task for radiologists. We developed a computerized scheme to assist radiologists in the detection of diffuse lung disease in multi-detector computed tomography (CT). Two radiologists selected 31 normal and 37 abnormal CT scans with ground glass opacity, reticular, honeycombing and nodular disease patterns based on clinical reports. The abnormal cases in our database must contain at least an abnormal area with a severity of moderate or severe level that was subjectively rated by the radiologists. Because statistical texture features may lack the power to distinguish a nodular pattern from a normal pattern, the abnormal cases that contain only a nodular pattern were excluded. The areas that included specific abnormal patterns in the selected CT images were then delineated as reference standards by an expert chest radiologist. The lungs were first segmented in each slice by use of a thresholding technique, and then divided into contiguous volumes of interest (VOIs) with a 64 x 64 x 64 matrix size. For each VOI, we determined and employed statistical texture features, such as run-length and co-occurrence matrix features, to distinguish abnormal from normal lung parenchyma. In particular, we developed new run-length texture features with clear physical meanings to considerably improve the accuracy of our detection scheme. A quadratic classifier was employed for distinguishing between normal and abnormal VOIs by the use of a leave-one-case-out validation scheme. A rule-based criterion was employed to further determine whether a case was normal or abnormal. We investigated the impact of new and conventional texture features, VOI size and the dimensionality for regions of interest on detecting diffuse lung disease. When we employed new texture features for 3D VOIs of 64 x 64 x 64 voxels, our system achieved the highest performance level: a sensitivity of 86% and a specificity of 90% for the detection of abnormal VOIs, and a sensitivity of 89% and a specificity of 90% for the detection of abnormal cases. Our computerized scheme would be useful for assisting radiologists in the diagnosis of diffuse lung disease.
    Physics in Medicine and Biology 11/2009; 54(22):6881-99. · 2.70 Impact Factor
  • Source
    Jiahui Wang, Feng Li, Qiang Li
    [Show abstract] [Hide abstract]
    ABSTRACT: Accurate segmentation of lungs with severe interstitial lung disease (ILD) in thoracic computed tomography (CT) is an important and difficult task in the development of computer-aided diagnosis (CAD) systems. Therefore, we developed in this study a texture analysis-based method for accurate segmentation of lungs with severe ILD in multidetector CT scans. Our database consisted of 76 CT scans, including 31 normal cases and 45 abnormal cases with moderate or severe ILD. The lungs in three selected slices for each CT scan were first manually delineated by a medical physicist, and then confirmed or revised by an expert chest radiologist, and they were used as the reference standard for lung segmentation. To segment the lungs, we first employed a CT value thresholding technique to obtain an initial lung estimate, including normal and mild ILD lung parenchyma. We then used texture-feature images derived from the co-occurrence matrix to further identify abnormal lung regions with severe ILD. Finally, we combined the identified abnormal lung regions with the initial lungs to generate the final lung segmentation result. The overlap rate, volume agreement, mean absolute distance (MAD), and maximum absolute distance (dmax) between the automatically segmented lungs and the reference lungs were employed to evaluate the performance of the segmentation method. Our segmentation method achieved a mean overlap rate of 96.7%, a mean volume agreement of 98.5%, a mean MAD of 0.84 mm, and a mean dmax of 10.84 mm for all the cases in our database; a mean overlap rate of 97.7%, a mean volume agreement of 99.0%, a mean MAD of 0.66 mm, and a mean dmax of 9.59 mm for the 31 normal cases; and a mean overlap rate of 96.1%, a mean volume agreement of 98.1%, a mean MAD of 0.96 mm, and a mean dmax of 11.71 mm for the 45 abnormal cases with ILD. Our lung segmentation method provided accurate segmentation results for abnormal CT scans with severe ILD and would be useful for developing CAD systems for quantification, detection, and diagnosis of ILD.
    Medical Physics 10/2009; 36(10):4592-9. · 2.91 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To evaluate the number of actual detections versus "accidental" detections by a computer-aided detection (CAD) system for small nodular lung cancers (<or=30 mm) on chest radiographs, using two different criteria for measuring performance. A Food-and-Drug-Administration-approved CAD program (version 1.0; Riverain Medical) was applied to 34 chest radiographs with a "radiologist-missed" nodular cancer and 36 radiographs with a radiologist-mentioned nodule (a newer version 3.0 was also applied to the 36-case database). The marks applied by this CAD system consisted of 5-cm-diameter circles. A strict "nodule-in-center" criterion and a generous "nodule-in-circle" criterion were compared as methods for the calculation of CAD sensitivity. The increased sensitivities by the nodule-in-circle criterion were considered as nodules detected by chance. The number of false-positive (FP) marks was also analyzed. For the 34 radiologist-missed cancers, the nodule-in-circle criterion caused eight more cancers (24%) to be detected by chance, as compared to the nodule-in-center criterion, when using the version 1.0 results. For the 36 radiologist-mentioned nodules, the nodule-in-circle criterion caused seven more lesions (19%) to be detected by chance, as compared to the nodule-in-center criterion, when using the version 1.0 results, and three more lesions (8%) to be detected by chance when using the version 3.0 results. Version 1.0 yielded a mean of six FP marks per image, while version 3.0 yielded only three FP marks per image. The specific criteria used to define true- and false-positive CAD detections can substantially influence the apparent accuracy of a CAD system.
    Journal of Digital Imaging 05/2009; 23(1):66-72. · 1.10 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The aim of this study was to investigate the subjective similarity for pairs of images with various abnormal patterns of diffuse interstitial lung disease on thin-section computed tomography by experienced radiologists to explore a basis for selecting similar images to assist radiologists' interpretation. Four major patterns (ground-glass opacity, nodular opacity, reticular opacity, and honeycombing) on thin-section computed tomographic images were identified by at least two of three radiologists. One radiologist manually selected 104 image pairs, in which the images in each pair had the same pattern and were similar in appearance. An additional 208 image pairs were randomly selected and evenly divided among the four patterns. These pairs were then rated for subjective similarity (on a continuous scale ranging from 0 = not similar at all to 1.0 = almost identical) by 12 radiologists. For radiologist-selected pairs, the mean similarity rated by the 12 radiologists was 0.72. For randomly selected pairs, the mean similarity was higher for the same pattern (0.47) than for the varying patterns (0.27) (P < .001), and among the same pattern, the mean similarity was 0.63 for ground-glass opacity, 0.58 for honeycombing, 0.45 for nodular opacity, and 0.32 for reticular opacity. The mean standard deviation for similarity ratings on all pairs given by the 12 radiologists was 0.05 (rang, 0.01-0.09). Subjective similarity ratings for pairs of abnormal images can be measured reliably and reproducibly by radiologists and will provide a basis for the selection of similar images to assist radiologists' interpretation.
    Academic radiology 04/2009; 16(4):477-85. · 2.09 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The successful development of high performance computer-aided-diagnostic systems has potential to assist radiologists in the detection and diagnosis of diffuse lung disease. We developed in this study an automated scheme for the detection of diffuse lung disease on multi-detector computed tomography (MDCT). Our database consisted of 68 CT scans, which included 31 normal and 37 abnormal cases with three kinds of abnormal patterns, i.e., ground glass opacity, reticular, and honeycombing. Two radiologists first selected the CT scans with abnormal patterns based on clinical reports. The areas that included specific abnormal patterns in the selected CT images were then delineated as reference standards by an expert chest radiologist. To detect abnormal cases with diffuse lung disease, the lungs were first segmented from the background in each slice by use of a texture analysis technique, and then divided into contiguous volumes of interest (VOIs) with a 64×64×64 matrix size. For each VOI, we calculated many statistical texture features, including the mean and standard deviation of CT values, features determined from the run length matrix, and features from the co-occurrence matrix. A quadratic classifier was employed for distinguishing between normal and abnormal VOIs by use of a leave-one-case-out validation scheme. A rule-based criterion was employed to further determine whether a case was normal or abnormal. For the detection of abnormal VOIs, our CAD system achieved a sensitivity of 86% and a specificity of 90%. For the detection of abnormal cases, it achieved a sensitivity of 89% and a specificity of 90%. This preliminary study indicates that our CAD system would be useful for the detection of diffuse lung disease.
    Proc SPIE 02/2009;
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE To investigate the subjective similarity for pairs of images with various abnormal patterns of diffuse lung diseases (DLD) on thin-section computed tomography (CT) by experienced radiologists in order to explore a basis for selecting similar images in assisting radiologists’ image interpretation. METHOD AND MATERIALS A total of 226 regions of interest (ROI) images (75 mm x 75 mm) with 4 major pattern types (ground-glass opacity [GGO], nodular opacity, reticular opacity, and honeycombing) of DLD on thin-section CT in 71 patients (mean age, 54 years; 35 men and 36 women) were identified by three radiologists. One radiologist manually selected 104 image pairs, where the images in each pair had the same pattern type and were similar in appearance. An additional 208 image pairs were randomly selected, evenly divided between the 4 pattern types. The subjective similarity (from “0: not similar at all” to “1.0: almost identical”) for each of pair was rated by 12 experienced radiologists. Similarity ratings and correlation coefficients were analyzed according to radiologist groupings and image pattern types. RESULTS The correlation coefficient of subjective similarity for all 312 pairs of images was 0.69 ± 0.05 (mean ± standard deviation) between one radiologist and any another one. For selected pairs with the same pattern types, the mean similarity was 0.72 ± 0.05 by the 12 observers. For randomized pairs, the mean similarity was 0.49 ± 0.06 for the same pattern type and 0.24 ± 0.04 for the varying pattern types (P < 0.001). For randomized pairs with the same pattern types, the mean similarity was higher for GGO (0.63 ± 0.05) and honeycombing (0.58 ± 0.05) than for reticular opacity (0.32 ± 0.05). CONCLUSION Subjective similarity ratings for pairs of abnormal images can be measured reliably and reproducibly by radiologists, and will provide an objective measure for the selection of similar images in assisting radiologists’ image interpretation. CLINICAL RELEVANCE/APPLICATION Subjective similarity measurements for pairs of abnormal images can provide an objective measure for the selection of similar images in order to assist radiologists’ image interpretation.
    Radiological Society of North America 2008 Scientific Assembly and Annual Meeting; 12/2008
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE To evaluate the performance of a CAD system for detecting small lung cancers on dual-energy subtraction (ES) chest radiographs compared to standard radiographs. METHOD AND MATERIALS A newer non-FDA approved Nodule Detection CAD system (Riverain version 4.0) was applied to the radiographs of 50 patients (mean age, 68.9 years; 27 men and 23 women) including standard posterior anterior (PA) and ES soft-tissue images. The 50 radiographs included 52 nodular lung cancers (mean size, 22.6 mm; range, 8-30 mm) that were confirmed by surgery or follow-up. The CAD sensitivity (the percentage of cancers marked) and number of false positives (FP) marked were calculated for CAD detection on both the standard chest images and ES soft-tissue images. Novel combinations of CAD marks (one combination using all marks from standard and soft-tissue images, and another combination using only marks occurring on both standard and ES images in a similar location) were used for determining the sensitivity and FP marks per patient. RESULTS The 4.0 CAD system marked 30 (sensitivity: 58%; FPs: 2.2 per image) out of 52 nodular cancers using standard chest images, and 35 (sensitivity: 67%; FPs: 1.6 per image) by using soft-tissue images. Three cancers were marked only on standard images and eight were marked only on soft-tissue images. The CAD sensitivity was 73% (38/52) with 3.34 FPs per patient using all marks for both images, and was 50% (26/52) with 0.48 FPs per patient when only considering closely located marks (between the standard and soft-tissue images). CONCLUSION CAD systems for detection of nodules on chest radiographs can achieve substantially improved performance by incorporating dual energy soft-tissue images, which provides gains in sensitivity as well as specificity for detection of small lung cancers. CLINICAL RELEVANCE/APPLICATION CAD performance for detection of small lung cancers can be improved by using dual energy soft-tissue chest radiographs.
    Radiological Society of North America 2008 Scientific Assembly and Annual Meeting; 12/2008

Publication Stats

1k Citations
139.95 Total Impact Points

Institutions

  • 2002–2011
    • University of Chicago
      • Department of Radiology
      Chicago, IL, United States
  • 2009
    • Duke University Medical Center
      • Department of Radiology
      Durham, NC, United States
  • 2008
    • Duke University
      Durham, North Carolina, United States