Computer-aided Detection Performance Research Articles

Performance of Computer-Aided Detection and Quality of Bowel Preparation: A Comprehensive Analysis of Colonoscopy Outcomes

BackgroundArtificial intelligence (AI) has emerged as a promising tool for detecting and characterizing colorectal polyps during colonoscopy, offering potential enhancements in traditional colonoscopy procedures to improve outcomes in patients with inadequate bowel preparation.AimsThis study aimed to assess the impact of an AI tool on computer-aided detection (CADe) assistance during colonoscopy in this population.MethodsThis case–control study utilized propensity score matching (PSM) for age, sex, race, and colonoscopy indication to analyze a database of patients who underwent colonoscopy at a single tertiary referral center between 2017 and 2023. Patients were excluded if the procedure was incomplete or aborted owing to poor preparation. The patients were categorized based on the use of AI during colonoscopy. Data on patient demographics and colonoscopy performance metrics were collected. Univariate and multivariate logistic regression models were used to compare the groups.ResultsAfter PSM patients with adequately prepped colonoscopies (n = 1466), the likelihood of detecting hyperplastic polyps (OR = 2.0, 95%CI 1.7–2.5, p < 0.001), adenomas (OR = 1.47, 95%CI 1.19–1.81, p < 0.001), and sessile serrated polyps (OR = 1.90, 95%CI 1.20–3.03, p = 0.007) significantly increased with the inclusion of CADe. In inadequately prepped patients (n = 160), CADe exhibited a more pronounced impact on the polyp detection rate (OR = 4.34, 95%CI 1.6–6.16, p = 0.049) and adenomas (OR = 2.9, 95%CI 2.20–8.57, p < 0.001), with a marginal increase in withdrawal and procedure times.ConclusionThis study highlights the significant improvement in detecting diminutive polyps (< 5 mm) and sessile polyps using CADe, although notably, this benefit was only seen in patients with adequate bowel preparation. In conclusion, the integration of AI in colonoscopy, driven by artificial intelligence, promises to significantly enhance lesion detection and diagnosis, revolutionize the procedure’s effectiveness, and improve patient outcomes.

Incorrect Computer Aided Detection (CAD) marks lead to early quitting: A potential mechanism for poor CAD performance in clinical practice

Incorrect Computer Aided Detection (CAD) marks lead to early quitting: A potential mechanism for poor CAD performance in clinical practice

Application of computer-aided detection (CAD) software to automatically detect nodules under SDCT and LDCT scans with different parameters

PurposeTo explore the application of computer-aided detection (CAD) software on automatically detecting nodules under standard-dose CT (SDCT) and low-dose CT (LDCT) scans with different parameters including definition modes and blending levels of adaptive statistical iterative reconstruction (ASIR), whose influence was important to optimize radiology workflow serving for clinical work. Materials and methods117 patients underwent SDCT and LDCT scans. The comprehensive performance of CAD in detect pulmonary nodules including under different ASIR blending levels (0%, 60%, and 80%) and high-definition (HD) or non-HD modes were assessed. The true positive (TP) rate, false positive (FP) rate and the sensitivity were recorded. ResultsThe stand-alone sensitivity of CAD system was 78.03% (515/660) in SDCT images and 70.15% (456/650) on LDCT images (p < 0.05). The sensitivity of CAD system to pulmonary nodules under non-HD mode was higher than that under HD mode. The detectability of nodules in images reconstructed with 60% and 80% ASIR was found significantly superior to that with 0% ASIR (p < 0.001). The overall sensitivity of CAD system on LDCT images reconstructed with 60% ASIR under HD mode was greater than that with 0% ASIR (p < 0.05), but lower than that with 80% ASIR. However, under non-HD mode, CAD demonstrated a comparable performance on LDCT images reconstructed with 60% ASIR to those reconstructed with 80% ASIR. ConclusionUsing the CAD system to detect pulmonary nodules on LDCT images with appropriate levels of ASIR could maintain high diagnostic sensitivity while reducing the radiation dose, which is useful to optimize the radiology workflow.

Strengths and Weaknesses of an Artificial Intelligence Polyp Detection Program as Assessed by a High-Detecting Endoscopist

Strengths and Weaknesses of an Artificial Intelligence Polyp Detection Program as Assessed by a High-Detecting Endoscopist

Reforming the workers’ compensation process for occupational lung disease among miners in South Africa: an efficiency study of claims assessment

The large burden of silicosis and tuberculosis (TB) in the South African mining industry, coupled with an under-resourcing of the compensation agencies responsible for certifying occupational lung disease, have resulted in serious backlogs. This work aimed to measure the efficiency gains from triaging occupational lung disease claims using claim type, years of mining exposure and computer aided detection (CAD) to save on scarce medical adjudicators. During 2020, the compensation authority started to triage claims for TB and those of miners with < 10years of service to two-person panels instead of the four-person panel plus radiologist used previously. Efficiency gain was calculated in medical person-units saved and reduction in delays. Different service thresholds predictive of silicosis were simulated, as well as the impact of pre-classification of chest X-rays with CAD using different combinations of sensitivity and specificity. The new triage system saved 20.3% in person-time units and reduced delays by 10-20days. Without CAD the greatest efficiency gain (28%) was projected from dispensing with a mining service threshold and routing all non-TB claims to the small panels at the outset. Simulation of four different CAD sensitivity/specificity combinations yielded efficiency gains of 18.2-36.1%, with 31.1% judged the most realistic. Use of sensitivity of close to 100% would not be feasible because of the very low resulting specificity. Pre-adjudication triage of claims at the compensation agency is capable of saving a substantial proportion of adjudicator time and reducing certification delays. Additional efficiency gains are achievable by referring all claims to small panels to begin with and improvement of CAD performance for this ex-miner population.

Deep Learning for Detection of Pulmonary Metastasis on Chest Radiographs.

Background A computer-aided detection (CAD) system may help surveillance for pulmonary metastasis at chest radiography in situations where there is limited access to CT. Purpose To evaluate whether a deep learning (DL)-based CAD system can improve diagnostic yield for newly visible lung metastasis on chest radiographs in patients with cancer. Materials and Methods A regulatory-approved CAD system for lung nodules was implemented to interpret chest radiographs from patients referred by the medical oncology department in clinical practice. In this retrospective diagnostic cohort study, chest radiographs interpreted with assistance from a CAD system after the implementation (January to April 2019, CAD-assisted interpretation group) and those interpreted before the implementation (September to December 2018, conventional interpretation group) of the CAD system were consecutively included. The diagnostic yield (frequency of true-positive detections) and false-referral rate (frequency of false-positive detections) of formal reports of chest radiographs for newly visible lung metastasis were compared between the two groups using generalized estimating equations. Propensity score matching was performed between the two groups for age, sex, and primary cancer. Results A total of 2916 chest radiographs from 1521 patients (1546 men, 1370 women; mean age, 62 years) and 5681 chest radiographs from 3456 patients (2941 men, 2740 women; mean age, 62 years) were analyzed in the CAD-assisted interpretation and conventional interpretation groups, respectively. The diagnostic yield for newly visible metastasis was higher in the CAD-assisted interpretation group (0.86%, 25 of 2916 [95% CI: 0.58, 1.3] vs 0.32%, 18 of 568 [95% CI: 0.20, 0.50%]; P = .004). The false-referral rate in the CAD-assisted interpretation group (0.34%, 10 of 2916 [95% CI: 0.19, 0.64]) was not inferior to that in the conventional interpretation group (0.25%, 14 of 5681 [95% CI: 0.15, 0.42]) at the noninferiority margin of 0.5% (95% CI of difference: -0.15, 0.35). Conclusion A deep learning-based computer-aided detection system improved the diagnostic yield for newly visible metastasis on chest radiographs in patients with cancer with a similar false-referral rate. © RSNA, 2021 Online supplemental material is available for this article.

Performance changes due to differences in training data for cerebral aneurysm detection in head MR angiography images

The performance of computer-aided detection (CAD) software depends on the quality and quantity of the dataset used for machine learning. If the data characteristics in development and practical use are different, the performance of CAD software degrades. In this study, we investigated changes in detection performance due to differences in training data for cerebral aneurysm detection software in head magnetic resonance angiography images. We utilized three types of CAD software for cerebral aneurysm detection in MRA images, which were based on 3D local intensity structure analysis, graph-based features, and convolutional neural network. For each type of CAD software, we compared three types of training pattern, which were two types of training using single-site data and one type of training using multisite data. We also carried out internal and external evaluations. In training using single-site data, the performance of CAD software largely and unpredictably fluctuated when the training dataset was changed. Training using multisite data did not show the lowest performance among the three training patterns for any CAD software and dataset. The training of cerebral aneurysm detection software using data collected from multiple sites is desirable to ensure the stable performance of the software.

Artificial intelligence in the diagnosis of gastric precancerous conditions by image-enhanced endoscopy: a multicenter, diagnostic study (with video)

Gastric precancerous conditions, including gastric atrophy (GA) and intestinal metaplasia (IM), play an important role in the development of gastric cancer. Image-enhanced endoscopy (IEE) shows great potential in diagnosing gastric precancerous conditions and adenocarcinoma. In this study, a deep convolutional neural network system, named ENDOANGEL, was constructed to detect gastric precancerous conditions by IEE. Endoscopic images were retrospectively obtained from 5 hospitals in China for the development, validation, and internal and external test of the system. Prospective consecutive patients receiving IEE were enrolled from January 13, 2020 to October 29, 2020 in Renmin Hospital of Wuhan University to assess in real time the applicability of the proposed computer-aided detection (CADe) system in clinical practice, and the performance of CADe was compared with that of endoscopists. Six thousand two hundred fifty endoscopic images from 760 patients and 98 video clips from 77 individuals undergoing IEE were enrolled in this study. The diagnostic accuracy of GA was .901 (95% confidence interval [CI], .883-.917) in the internal test set, .864 (95% CI, .842-.884) in the multicenter external test set, and .878 (95% CI, .796-.935) in the prospective video test set. The diagnostic accuracy of IM was .908 (95% CI, .889-.924) in the internal test set, .859 (95% CI, .837-.880) in the multicenter external test set, and .898 (95% CI, .820-.950) in the prospective video test set. CADe achieved similar diagnostic accuracy to that of the experts for detecting GA (.869 [95% CI, .790-.927] vs .846 [95% CI, .808-.879], P= .396) and IM (.888 [95% CI, .812-.941] vs .820 [95% CI, .780-.855], P= .117) and was superior to that of nonexperts for GA (.750 [95% CI, .711-.786], P= .008) and IM (.736 [95% CI, .697-.773], P= .028). CADe achieved high diagnostic accuracy in gastric precancerous conditions, which was similar to that of experts and superior to that of nonexperts. Thus, CADe provides possibilities for a wide application in assisting in the diagnosis of gastric precancerous conditions.

Computer-aided Detection of Subsolid Nodules at Chest CT: Improved Performance with Deep Learning-based CT Section Thickness Reduction.

Background Studies on the optimal CT section thickness for detecting subsolid nodules (SSNs) with computer-aided detection (CAD) are lacking. Purpose To assess the effect of CT section thickness on CAD performance in the detection of SSNs and to investigate whether deep learning-based super-resolution algorithms for reducing CT section thickness can improve performance. Materials and Methods CT images obtained with 1-, 3-, and 5-mm-thick sections were obtained in patients who underwent surgery between March 2018 and December 2018. Patients with resected synchronous SSNs and those without SSNs (negative controls) were retrospectively evaluated. The SSNs, which ranged from 6 to 30 mm, were labeled ground-truth lesions. A deep learning-based CAD system was applied to SSN detection on CT images of each section thickness and those converted from 3- and 5-mm section thickness into 1-mm section thickness by using the super-resolution algorithm. The CAD performance on each section thickness was evaluated and compared by using the jackknife alternative free response receiver operating characteristic figure of merit. Results A total of 308 patients (mean age ± standard deviation, 62 years ± 10; 183 women) with 424 SSNs (310 part-solid and 114 nonsolid nodules) and 182 patients without SSNs (mean age, 65 years ± 10; 97 men) were evaluated. The figures of merit differed across the three section thicknesses (0.92, 0.90, and 0.89 for 1, 3, and 5 mm, respectively; P = .04) and between 1- and 5-mm sections (P = .04). The figures of merit varied for nonsolid nodules (0.78, 0.72, and 0.66 for 1, 3, and 5 mm, respectively; P < .001) but not for part-solid nodules (range, 0.93-0.94; P = .76). The super-resolution algorithm improved CAD sensitivity on 3- and 5-mm-thick sections (P = .02 for 3 mm, P < .001 for 5 mm). Conclusion Computer-aided detection (CAD) of subsolid nodules performed better at 1-mm section thickness CT than at 3- and 5-mm section thickness CT, particularly with nonsolid nodules. Application of a super-resolution algorithm improved the sensitivity of CAD at 3- and 5-mm section thickness CT. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Goo in this issue.

Optimal matrix size of chest radiographs for computer-aided detection on lung nodule or mass with deep learning

To investigate the optimal input matrix size for deep learning-based computer-aided detection (CAD) of nodules and masses on chest radiographs. We retrospectively collected 2088 abnormal (nodule/mass) and 352 normal chest radiographs from two institutions. Three thoracic radiologists drew 2758 abnormalities regions. A total of 1736 abnormal chest radiographs were used for training and tuning convolutional neural networks (CNNs). The remaining 352 abnormal and 352 normal chest radiographs were used as a test set. Two CNNs (Mask R-CNN and RetinaNet) were selected to validate the effects of the squared different matrix size of chest radiograph (256, 448, 896, 1344, and 1792). For comparison, figure of merit (FOM) of jackknife free-response receiver operating curve and sensitivity were obtained. In Mask R-CNN, matrix size 896 and 1344 achieved significantly higher FOM (0.869 and 0.856, respectively) for detecting abnormalities than 256, 448, and 1792 (0.667-0.820) (p < 0.05). In RetinaNet, matrix size 896 was significantly higher FOM (0.906) than others (0.329-0.832) (p < 0.05). For sensitivity of abnormalities, there was a tendency to increase sensitivity when lesion size increases. For small nodules (< 10mm), the sensitivities were 0.418 and 0.409, whereas the sensitivities were 0.937 and 0.956 for masses. Matrix size 896 and 1344 in Mask R-CNN and matrix size 896 in RetinaNet showed significantly higher sensitivity than others (p < 0.05). Matrix size 896 had the highest performance for various sizes of abnormalities using different CNNs. The optimal matrix size of chest radiograph could improve CAD performance without additional training data. • Input matrix size significantly affected the performance of a deep learning-based CAD for detection of nodules or masses on chest radiographs. • The matrix size 896 showed the best performance in two different CNN detection models. • The optimal matrix size of chest radiographs could enhance CAD performance without additional training data.

Short-term Reproducibility of Pulmonary Nodule and Mass Detection in Chest Radiographs: Comparison among Radiologists and Four Different Computer-Aided Detections with Convolutional Neural Net

To investigate the reproducibility of computer-aided detection (CAD) for detection of pulmonary nodules and masses for consecutive chest radiographies (CXRs) of the same patient within a short-term period. A total of 944 CXRs (Chest PA) with nodules and masses, recorded between January 2010 and November 2016 at the Asan Medical Center, were obtained. In all, 1092 regions of interest for the nodules and mass were delineated using an in-house software. All CXRs were randomly split into 6:2:2 sets for training, development, and validation. Furthermore, paired follow-up CXRs (n = 121) acquired within one week in the validation set, in which expert thoracic radiologists confirmed no changes, were used to evaluate the reproducibility of CAD by two radiologists (R1 and R2). The reproducibility comparison of four different convolutional neural net algorithms and two chest radiologists (with 13- and 14-years’ experience) was conducted. Model performances were evaluated by figure-of-merit (FOM) analysis of the jackknife free-response receiver operating curve and reproducibility rates were evaluated in terms of percent positive agreement (PPA) and Chamberlain’s percent positive agreement (CPPA). Reproducibility analysis of the four CADs and R1 and R2 showed variations in the PPA and CPPA. Model performance of YOLO (You Only Look Once) v2 based eDenseYOLO showed a higher FOM (0.89; 0.85–0.93) than RetinaNet (0.89; 0.85–0.93) and atrous spatial pyramid pooling U-Net (0.85; 0.80–0.89). eDenseYOLO showed higher PPAs (97.87%) and CPPAs (95.80%) than Mask R-CNN, RetinaNet, ASSP U-Net, R1, and R2 (PPA: 96.52%, 94.23%, 95.04%, 96.55%, and 94.98%; CPPA: 93.18%, 89.09%, 90.57%, 93.33%, and 90.43%). There were moderate variations in the reproducibility of CAD with different algorithms, which likely indicates that measurement of reproducibility is necessary for evaluating CAD performance in actual clinical environments.

Volume Visualization for Improving CT Lung Nodule Detection.

Inspired by the outstanding performance of deep convolutional neural networks (CNNs), nowadays modern computer-aided detection (CAD) systems for CT lung nodules generally delve into 2D or 3D CNNs directly without considering traditional image preprocessing techniques. However, detection of large pulmonary nodules and masses are computationally challenging, especially for 3D CNNs. In this paper, we examine the possibility of using volume visualized CT thin-slab images with 2D CNNs to reduce computation complexity and improve CAD performance. We tested 4 types of images: original 2D CT, 2D projection of thin slabs, mixture by arranging original and projection in different color channels, and mixture by the pixelwise maximum intensity of original CT and projection. We evaluated these images on a dataset of 30 CT scans with 30 different-sized nodules and masses on GoogLeNet via a transfer learning and cross validation paradigm. We found that projection visualization alone had a better or equal area-under curve score for all the different-sized nodules and masses. However, mixture by the maximum of CT and projection demonstrated a preferred performance with a true positive rate of 0.8 and a false positive rate of 0.046 in detecting large nodules and masses.

Computer-Aided Detection of Pulmonary Nodules in Computed Tomography Using ClearReadCT.

This study evaluates the accuracy of a computer-aided detection (CAD) application for pulmonary nodular lesions (PNL) in computed tomography (CT) scans, the ClearReadCT (Riverain Technologies). The study was retrospective for 106 biopsied PNLs from 100 patients. Seventy-five scans were Contrast-Enhanced (CECT) and 25 received no enhancer (NECT). Axial reconstructions in soft-tissue and lung kernel were applied at three different slice thicknesses, 0.75mm (CECT/NECT n = 25/6), 1.5mm (n = 18/9) and 3.0mm (n = 43/18). We questioned the effect of (1) enhancer, (2) kernel and (3) slice thickness on the CAD performance. Our main findings are: (1) Vessel suppression is effective and specific in both NECT and CECT. (2) Contrast enhancement significantly increased the CAD sensitivity from 60% in NECT to 80% in CECT, P = 0.025 Fischer's exact test. (3) The CAD sensitivity was 84% in 3mm slices compared to 68% in 0.75mm slices, P > 0.2 Fischer's exact test. (4) Small lesions of low attenuation were detected with higher sensitivity. (5) Lung kernel reconstructions increased the false positive rate without affecting the sensitivity (P > 0.05 McNemar's test). In conclusion, ClearReadCT showed an optimized sensitivity of 84% and a positive predictive value of 67% in enhanced lung scans with thick, soft kernel reconstructions. NECT, thin slices and lung kernel reconstruction were associated with inferior performance.

Performance analysis of a computer-aided detection system for lung nodules in CT at different slice thicknesses.

.We study the performance of a computer-aided detection (CAD) system for lung nodules in computed tomography (CT) as a function of slice thickness. In addition, we propose and compare three different training methodologies for utilizing nonhomogeneous thickness training data (i.e., composed of cases with different slice thicknesses). These methods are (1) aggregate training using the entire suite of data at their native thickness, (2) homogeneous subset training that uses only the subset of training data that matches each testing case, and (3) resampling all training and testing cases to a common thickness. We believe this study has important implications for how CT is acquired, processed, and stored. We make use of 192 CT cases acquired at a thickness of 1.25 mm and 283 cases at 2.5 mm. These data are from the publicly available Lung Nodule Analysis 2016 dataset. In our study, CAD performance at 2.5 mm is comparable with that at 1.25 mm and is much better than at higher thicknesses. Also, resampling all training and testing cases to 2.5 mm provides the best performance among the three training methods compared in terms of accuracy, memory consumption, and computational time.

Collaboration between interactive three-dimensional visualization and computer aided detection of pulmonary embolism on computed tomography pulmonary angiography views.

This paper explores the benefits of interactive three-dimensional (3D) visualization of stimuli using a computer aided detection (CAD) system of pulmonary embolisms (PEs), on computed tomography pulmonary angiography (CTPA) views. We designed a new CAD method that prompts the PE sites on CTPA views; we then utilized two interactive approaches of 3D visualization to assess CAD performance. This collaboration allows different methodologies to be used to assess PE-CAD performance, by comparison with the common method in which radiologists are prompted with CAD stimuli directly on the CTPA views. Both 3D approaches are based on the voxel size of the CTPA examination, and consider the acquisition settings. A set of ten retrospective CTPA cases were collected, with different acquisition parameters, in terms of voxel size and spatial resolution. 3D visualization CAD performance was examined by an experienced radiologist. Both 3D visualization methods proved to have a constructive impact on improving CAD performance. The rate of true positive (TP) responses increased by 27%; while the rate of false positive (FP) responses dropped by 31%. We concluded that evaluation of PE-CAD performance utilizing interactive 3D visualization could increase or ascertain the correct rate of TP stimuli, as well as noticeably reduce FP responses.

Global optimal hybrid geometric active contour for automated lung segmentation on CT images

Lung segmentation on thoracic CT images plays an important role in early detection, diagnosis and 3D visualization of lung cancer. The segmentation accuracy, stability, and efficiency of serial CT scans have a significant impact on the performance of computer-aided detection. This paper proposes a global optimal hybrid geometric active contour model for automated lung segmentation on CT images. Firstly, the combination of global region and edge information leads to high segmentation accuracy in lung regions with weak boundaries or narrow bands. Secondly, due to the global optimality of energy functional, the proposed model is robust to the initial position of level set function and requires fewer iterations. Thus, the stability and efficiency of lung segmentation on serial CT slices can be greatly improved by taking advantage of the information between adjacent slices. In addition, to achieve the whole process of automated segmentation for lung cancer, two assistant algorithms based on prior shape and anatomical knowledge are proposed. The algorithms not only automatically separate the left and right lungs, but also include juxta-pleural tumors into the segmentation result. The proposed method was quantitatively validated on subjects from the publicly available LIDC-IDRI and our own data sets. Exhaustive experimental results demonstrate the superiority and competency of our method, especially compared with the typical edge-based geometric active contour model.

The effect of radiation dose reduction on computer-aided detection (CAD) performance in a low-dose lung cancer screening population.

Lung cancer screening with low-dose CT has recently been approved for reimbursement, heralding the arrival of such screening services worldwide. Computer-aided detection (CAD) tools offer the potential to assist radiologists in detecting nodules in these screening exams. In lung screening, as in all CT exams, there is interest in further reducing radiation dose. However, the effects of continued dose reduction on CAD performance are not fully understood. In this work, we investigated the effect of reducing radiation dose on CAD lung nodule detection performance in a screening population. The raw projection data files were collected from 481 patients who underwent low-dose screening CT exams at our institution as part of the National Lung Screening Trial (NLST). All scans were performed on a multidetector scanner (Sensation 64, Siemens Healthcare, Forchheim Germany) according to the NLST protocol, which called for a fixed tube current scan of 25 effective mAs for standard-sized patients and 40 effective mAs for larger patients. The raw projection data were input to a reduced-dose simulation software to create simulated reduced-dose scans corresponding to 50% and 25% of the original protocols. All raw data files were reconstructed at the scanner with 1 mm slice thickness and B50 kernel. The lungs were segmented semi-automatically, and all images and segmentations were input to an in-house CAD algorithm trained on higher dose scans (75-300 mAs). CAD findings were compared to a reference standard generated by an experienced reader. Nodule- and patient-level sensitivities were calculated along with false positives per scan, all of which were evaluated in terms of the relative change with respect to dose. Nodules were subdivided based on size and solidity into categories analogous to the LungRADS assessment categories, and sub-analyses were performed. From the 481 patients in this study, 82 had at least one nodule (prevalence of 17%) and 399 did not (83%). A total of 118 nodules were identified. Twenty-seven nodules (23%) corresponded to LungRADS category 4 based on size and composition, while 18 (15%) corresponded to LungRADS category 3 and 73 (61%) corresponded to LungRADS category 2. For solid nodules ≥8 mm, patient-level median sensitivities were 100% at all three dose levels, and mean sensitivities were 72%, 63%, and 63% at original, 50%, and 25% dose, respectively. Overall mean patient-level sensitivities for nodules ranging from 3 to 45 mm were 38%, 37%, and 38% at original, 50%, and 25% dose due to the prevalence of smaller nodules and nonsolid nodules in our reference standard. The mean false-positive rates were 3, 5, and 13 per case. CAD sensitivity decreased very slightly for larger nodules as dose was reduced, indicating that reducing the dose to 50% of original levels may be investigated further for use in CT screening. However, the effect of dose was small relative to the effect of the nodule size and solidity characteristics. The number of false positives per scan increased substantially at 25% dose, illustrating the importance of tuning CAD algorithms to very challenging, high-noise screening exams.

A method for evaluating the performance of computer-aided detection of pulmonary nodules in lung cancer CT screening: detection limit for nodule size and density.

We propose the application of virtual nodules to evaluate the performance of computer-aided detection (CAD) of lung nodules in cancer screening using low-dose CT. The virtual nodules were generated based on the spatial resolution measured for a CT system used in an institution providing cancer screening and were fused into clinical lung images obtained at that institution, allowing site specificity. First, we validated virtual nodules as an alternative to artificial nodules inserted into a phantom. In addition, we compared the results of CAD analysis between the real nodules (n = 6) and the corresponding virtual nodules. Subsequently, virtual nodules of various sizes and contrasts between nodule density and background density (ΔCT) were inserted into clinical images (n = 10) and submitted for CAD analysis. In the validation study, 46 of 48 virtual nodules had the same CAD results as artificial nodules (kappa coefficient = 0.913). Real nodules and the corresponding virtual nodules showed the same CAD results. The detection limits of the tested CAD system were determined in terms of size and density of peripheral lung nodules; we demonstrated that a nodule with a 5-mm diameter was detected when the nodule had a ΔCT > 220 HU. Virtual nodules are effective in evaluating CAD performance using site-specific scan/reconstruction conditions. Advances in knowledge: Virtual nodules can be an effective means of evaluating site-specific CAD performance. The methodology for guiding the detection limit for nodule size/density might be a useful evaluation strategy.

COMPUTER-AIDED DETECTION OF BRAIN TUMORS USING MORPHOLOGICAL RECONSTRUCTION

Computer aided detection (CAD) systems helps the detection of abnormalities in medical images using advanced image processing and pattern recognition techniques. CAD has advantages in accelerating decision-making and reducing the human error in detection process. In this study, a CAD system is developed which is based on morphological reconstruction and classification methods with the use of morphological features of the regions of interest to detect brain tumors from brain magnetic resonance (MR) images. The CAD system consists of four stages: the preprocessing, the segmentation, region of interest specification and tumor detection stages. The system is evaluated on REMBRANDT dataset with 497 MR image slices of 10 patients. In the classification stage the performance of CAD has achieved accuracy of 93.36% with Decision Tree Algorithm, 94.89% with Artificial Neural Network (Multilayer Perceptron), 96.93% with K-Nearest Neighbour Algorithm and 96.93% with Meta-Learner (Decorate) Algorithm. These results show that the proposed technique is effective and promising for detecting tumors in brain MR images and enhances the classification process to be more accurate. The using morphological reconstruction method is useful and adaptive than the methods used in other CAD applications.

Heterogeneous data analysis: Online learning for medical-image-based diagnosis

Heterogeneous Data Analysis (HDA) is proposed to address a learning problem of medical image databases of Computed Tomographic Colonography (CTC). The databases are generated from clinical CTC images using a Computer-aided Detection (CAD) system, the goal of which is to aid radiologists' interpretation of CTC images by providing highly accurate, machine-based detection of colonic polyps. We aim to achieve a high detection accuracy in CAD in a clinically realistic context, in which additional CTC cases of new patients are added regularly to an existing database. In this context, the CAD performance can be improved by exploiting the heterogeneity information that is brought into the database through the addition of diverse and disparate patient populations. In the HDA, several quantitative criteria of data compatibility are proposed for efficient management of these online images. After an initial supervised offline learning phase, the proposed online learning method decides whether the online data are heterogeneous or homogeneous. Our previously developed Principal Composite Kernel Feature Analysis (PC-KFA) is applied to the online data, managed with HDA, for iterative construction of a linear subspace of a high-dimensional feature space by maximizing the variance of the non-linearly transformed samples. The experimental results showed that significant improvements in the data compatibility were obtained when the online PC-KFA was used, based on an accuracy measure for long-term sequential online datasets. The computational time is reduced by more than 93% in online training compared with that of offline training.