Computer-aided Detection Research Articles

Effect of computer aided detection system on esophageal neoplasm diagnosis in varied levels of endoscopists

A computer-aided detection (CAD) system for early esophagus carcinoma identification during endoscopy with narrow-band imaging (NBI) was evaluated in a large-scale, prospective, tandem, randomized controlled trial to assess its effectiveness. The study was registered at the Chinese Clinical Trial Registry (ChiCTR2100050654, 2021/09/01). Involving 3400 patients were randomly assigned to either routine (routine-first) or CAD-assisted (CAD-first) NBI endoscopy, followed by the other procedure, with targeted biopsies taken at the end of the second examination. The primary outcome was the diagnosis of 1 or more neoplastic lesion of esophagus during the first examination. The CAD-first group demonstrated a significantly higher neoplastic lesion detection rate (3.12%) compared to the routine-first group (1.59%) with a relative detection ratio of 1.96 (P = 0.0047). Subgroup analysis revealed a higher detection rate in junior endoscopists using CAD-first, while no significant difference was observed for senior endoscopists. The CAD system significantly improved esophageal neoplasm detection, particularly benefiting junior endoscopists.

Facilitating Radiograph Interpretation: Refined Generative Models for Precise Bone Suppression in Chest X-rays.

Chest X-ray (CXR) is crucial for diagnosing lung diseases, especially lung nodules. Recent studies indicate that bones, such as ribs and clavicles, obscure 82 to 95% of undiagnosed lung cancers. The development of computer-aided detection (CAD) systems with automated bone suppression is vital to improve detection rates and support early clinical decision-making. Current bone suppression methods face challenges: they often depend on manual subtraction of bone-only images from CXRs, leading to inefficiency and poor generalization; there is significant information loss in data compression within deep convolutional end-to-end architectures; and a balance between model efficiency and accuracy has not been sufficiently achieved in existing research. We introduce a novel end-to-end architecture, the mask-guided model, to address these challenges. Leveraging the Pix2Pix framework, our model enhances computational efficiency by reducing parameter count by 92.5%. It features a rib mask-guided module with a mask encoder and cross-attention mechanism, which provides spatial constraints, reduces information loss during encoder compression, and preserves non-relevant areas. An ablation study evaluates the impact of various factors. The model undergoes initial training on digitally reconstructed radiographs (DRRs) derived from CT projections for bone suppression and is fine-tuned on the JSRT dataset to accelerate convergence. The mask-guided model surpasses previous state-of-the-art methods, showing superior bone suppression performance in terms of structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), and processing speed. It achieves an SSIM of 0.99 ± 0.002 and a PSNR of 36.14 ± 1.13 on the JSRT dataset. This study underscores the proposed model's effectiveness compared to existing methods, showcasing its capability to reduce model size and increase accuracy. This makes it well-suited for deployment in affordable, low-power hardware devices across various clinical settings.

Automatic cerebral microbleeds detection from MR images via multi-channel and multi-scale CNNs.

Automatic cerebral microbleeds detection from MR images via multi-channel and multi-scale CNNs.

The development and ex vivo evaluation of a computer-aided quality control system for Barrett's esophagus endoscopy.

BACKGROUND : Timely detection of neoplasia in Barrett's esophagus (BE) remains challenging. While computer-aided detection (CADe) systems have been developed to assist endoscopists, their effectiveness depends heavily on the quality of the endoscopic procedure. This study introduces a novel computer-aided quality (CAQ) system for BE, evaluating its stand-alone performance and integration with a CADe system. METHOD : The CAQ system was developed using 7,463 images from 359 BE patients. It assesses objective quality parameters (e. g., blurriness, illumination) and subjective parameters (mucosal cleanliness, esophageal expansion) and can exclude low-quality images when integrated with a CADe system.To evaluate CAQ stand-alone performance, the Endoscopic Image Quality test set, consisting of 647 images from 51 BE patients across 8 hospitals, was labeled for objective and subjective quality. To assess the benefit of the CAQ system as a preprocessing filter of a CADe system, the Barrett CADe test set was developed. It consisted of 956 video frames from 62 neoplastic patients and 557 frames from 35 non-dysplastic patients, in 12 Barrett referral centers. RESULTS : As stand-alone tool, the CAQ system achieved Cohen's Kappa scores of 0.73, 0.91, and 0.89 for objective quality, mucosal cleanliness, and esophageal expansion, comparable to inter-annotator scores of 0.73, 0.93, and 0.83. As preprocessing filter, the CAQ system improved CADe sensitivity from 82 % to 90 % and AUC from 87 % to 91 %, while maintaining specificity at 75 %. CONCLUSION : This study presents the first CAQ system for automated quality control in BE. The system effectively distinguishes poorly from well-visualized mucosa and enhances neoplasia detection when integrated with CADe.

Artificial intelligence for breast cancer screening in mammography (AI-STREAM): preliminary analysis of a prospective multicenter cohort study.

Artificial intelligence (AI) improves the accuracy of mammography screening, but prospective evidence, particularly in a single-read setting, remains limited. This study compares the diagnostic accuracy of breast radiologists with and without AI-based computer-aided detection (AI-CAD) for screening mammograms in a real-world, single-read setting. A prospective multicenter cohort study is conducted within South Korea's national breast cancer screening program for women. The primary outcomes are screen-detected breast cancer within one year, with a focus on cancer detection rates (CDRs) and recall rates (RRs) of radiologists. A total of 24,543 women are included in the final cohort, with 140 (0.57%) screen-detected breast cancers. The CDR is significantly higher by 13.8% for breast radiologists using AI-CAD (n = 140 [5.70‰]) compared to those without AI (n = 123 [5.01‰]; p < 0.001), with no significant difference in RRs (p = 0.564). These preliminary results show a significant improvement in CDRs without affecting RRs in a radiologist's standard single-reading setting (ClinicalTrials.gov: NCT05024591).

Clinical evaluation of computer-aided digital x-ray detection of pulmonary tuberculosis during community-based screening or active case-finding: a case-control study.

Clinical evaluation of computer-aided digital x-ray detection of pulmonary tuberculosis during community-based screening or active case-finding: a case-control study.

Spectrum of errors in nodule detection and characterization using machine learning: A pictorial essay.

Spectrum of errors in nodule detection and characterization using machine learning: A pictorial essay.

Human-AI Interaction in the ScreenTrustCAD Trial: Recall Proportion and Positive Predictive Value Related to Screening Mammograms Flagged by AI CAD versus a Human Reader.

Background The ScreenTrustCAD trial was a prospective study that evaluated the cancer detection rates for combinations of artificial intelligence (AI) computer-aided detection (CAD) and two radiologists. The results raised concerns about the tendency of radiologists to agree with AI CAD too much (when AI CAD made an erroneous flagging) or too little (when AI CAD made a correct flagging). Purpose To evaluate differences in recall proportion and positive predictive value (PPV) related to which reader flagged the mammogram for consensus discussion: AI CAD and/or radiologists. Materials and Methods Participants were enrolled from April 2021 to June 2022, and each examination was interpreted by three independent readers: two radiologists and AI CAD, after which positive findings were forwarded to the consensus discussion. For each combination of readers flagging an examination, the proportion recalled and the PPV were calculated by dividing the number of pathologic evaluation-verified cancers by the number of positive examinations. Results The study included 54 991 women (median age, 55 years [IQR, 46-65 years]), among whom 5489 were flagged for consensus discussion and 1348 were recalled. For examinations flagged by one reader, the proportion recalled after flagging by one radiologist was larger (14.2% [263 of 1858]) compared with flagging by AI CAD (4.6% [86 of 1886]) (P < .001), whereas the PPV of breast cancer was lower (3.4% [nine of 263] vs 22% [19 of 86]) (P < .001). For examinations flagged by two readers, the proportion recalled after flagging by two radiologists was larger (57.2% [360 of 629]) compared with flagging by AI CAD and one radiologist (38.6% [244 of 632]) (P < .001), whereas the PPV was lower (2.5% [nine of 360] vs 25.0% [61 of 244]) (P < .001). For examinations flagged by all three readers, the proportion recalled was 82.6% (400 of 484) and the PPV was 34.2 (137 of 400). Conclusion A larger proportion of participants were recalled after initial flagging by radiologists compared with those flagged by AI CAD, with a lower proportion of cancer. ClinicalTrials.gov Identifier: NCT04778670 © RSNA, 2025 See also the editorial by Grimm in this issue.

From data to artificial intelligence: evaluating the readiness of gastrointestinal endoscopy datasets.

The incorporation of artificial intelligence (AI) into gastrointestinal (GI) endoscopy represents a promising advancement in gastroenterology. With over 40 published randomized controlled trials and numerous ongoing clinical trials, gastroenterology leads other medical disciplines in AI research. Computer-aided detection algorithms for identifying colorectal polyps have achieved regulatory approval and are in routine clinical use, while other AI applications for GI endoscopy are in advanced development stages. Near-term opportunities include the potential for computer-aided diagnosis to replace conventional histopathology for diagnosing small colon polyps and increased AI automation in capsule endoscopy. Despite significant development in research settings, the generalizability and robustness of AI models in real clinical practice remain inconsistent. The GI field lags behind other medical disciplines in the breadth of novel AI algorithms, with only 13 out of 882 Food and Drug Administration (FDA)-approved AI models focussed on GI endoscopy as of June 2024. Additionally, existing GI endoscopy image databases are disproportionately focussed on colon polyps, lacking representation of the diversity of other endoscopic findings. High-quality datasets, encompassing a wide range of patient demographics, endoscopic equipment types, and disease states, are crucial for developing effective AI models for GI endoscopy. This article reviews the current state of GI endoscopy datasets, barriers to progress, including dataset size, data diversity, annotation quality, and ethical issues in data collection and usage, and future needs for advancing AI in GI endoscopy.

A deep architecture based on attention mechanisms for effective end-to-end detection of early and mature malaria parasites in a realistic scenario.

A deep architecture based on attention mechanisms for effective end-to-end detection of early and mature malaria parasites in a realistic scenario.

Impact of standard enhancement settings of endoscopy systems on performance of endoscopic artificial intelligence systems.

Artificial intelligence (AI) systems in endoscopy are predominantly developed and tested using high-quality imagery from expert centers. However, their performance may be different when applied in clinical practice, partly due to the diversity in post-processing enhancement settings used in endoscopy units. We evaluated the impact of post-processing enhancement settings on AI performance and tested specific data augmentation strategies to mitigate performance loss. We used a computer-aided detection (CADe) system for Barrett's neoplasia (6223 images, 906 patients) and a computer-aided diagnosis (CADx) system for colorectal polyps (3288 images, 969 patients), both trained on datasets acquired with Olympus equipment and with limited variability in enhancement settings. The CAD systems were then tested across a wide range of test sets, which comprised the same images, but displayed with different enhancement settings. Both CAD systems were then retrained using image enhancement-based data augmentation. The performance of the adjusted CAD systems was evaluated on the same test sets. Both systems displayed substantial performance variability over a range of enhancement settings (CADe: 83 %-92 % sensitivity, 84 %-91 % specificity; CADx: 78 %-85 % sensitivity, 45 %-63 % specificity). After retraining, variability in sensitivity and specificity was reduced to 2 % (P < 0.001) and 1 %, respectively (P = 0.003) for CADe, and 2 % (P = 0.03) and 8 %, respectively (P = 0.19) for CADx. The performance of endoscopic AI systems can vary substantially depending on post-processing enhancement settings of the endoscopy unit. Specific data augmentation can mitigate this performance loss.

AI-powered prostate cancer detection: a multi-centre, multi-scanner validation study.

Multi-centre, multi-vendor validation of artificial intelligence (AI) software to detect clinically significant prostate cancer (PCa) using multiparametric magnetic resonance imaging (MRI) is lacking. We compared a new AI solution, validated on a separate dataset from different UK hospitals, to the original multidisciplinary team (MDT)-supported radiologist's interpretations. A Conformité Européenne (CE)-marked deep-learning (DL) computer-aided detection (CAD) medical device (Pi) was trained to detect Gleason Grade Group (GG) ≥ 2 cancer using retrospective data from the PROSTATEx dataset and five UK hospitals (793 patients). Our separate validation dataset was on six machines from two manufacturers across six sites (252 patients). Data included in the study were from MRI scans performed between August 2018 to October 2022. Patients with a negative MRI who did not undergo biopsy were assumed to be negative (90.4% had prostate-specific antigen density < 0.15 ng/mL2). ROC analysis was used to compare radiologists who used a 5-category suspicion score. GG ≥ 2 prevalence in the validation set was 31%. Evaluated per patient, Pi was non-inferior to radiologists (considering a 10% performance difference as acceptable), with an area under the curve (AUC) of 0.91 vs. 0.95. At the predetermined risk threshold of 3.5, the AI software's sensitivity was 95% and specificity 67%, while radiologists at Prostate Imaging-Reporting and Data Systems/Likert ≥ 3 identified GG ≥ 2 with a sensitivity of 99% and specificity of 73%. AI performed well per-site (AUC ≥ 0.83) at the patient-level independent of scanner age and field strength. Real-world data testing suggests that Pi matches the performance of MDT-supported radiologists in GG ≥ 2 PCa detection and generalises to multiple sites, scanner vendors, and models. QuestionThe performance of artificial intelligence-based medical tools for prostate MRI has yet to be evaluated on multi-centre, multi-vendor data to assess generalisability. FindingsA dedicated AI medical tool matches the performance of multidisciplinary team-supported radiologists in prostate cancer detection and generalises to multiple sites and scanners. Clinical relevanceThis software has the potential to support the MRI process for biopsy decision-making and target identification, but future prospective studies, where lesions identified by artificial intelligence are biopsied separately, are needed.

Computer-Aided Decision Support and 3D Models in Pancreatic Cancer Surgery: A Pilot Study.

Background: Preoperative planning of patients diagnosed with pancreatic head cancer is difficult and requires specific expertise. This pilot study assesses the added value of three-dimensional (3D) patient models and computer-aided detection (CAD) algorithms in determining the resectability of pancreatic head tumors. Methods: This study included 14 hepatopancreatobiliary experts from eight hospitals. The participants assessed three radiologically resectable and three radiologically borderline resectable cases in a simulated setting via crossover design. Groups were divided in controls (using a CT scan), a 3D group (using a CT scan and 3D models), and a CAD group (using a CT scan, 3D and CAD). For the perceived fulfillment of preoperative needs, the quality and confidence of clinical decision-making were evaluated. Results: A higher perceived ability to determine degrees and the length of tumor-vessel contact was reported in the CAD group compared to controls (p = 0.022 and p = 0.003, respectively). Lower degrees of tumor-vessel contact were predicted for radiologically borderline resectable tumors in the CAD group compared to controls (p = 0.037). Higher confidence levels were observed in predicting the need for vascular resection in the 3D group compared to controls (p = 0.033) for all cases combined. Conclusions: "CAD (including 3D) improved experts' perceived ability to accurately assess vessel involvement and supports the development of evolving techniques that may enhance the diagnosis and treatment of pancreatic cancer".

Enhanced glioma tumor detection and segmentation using modified deep learning with edge fusion and frequency features

Computer-aided automatic brain tumor detection is crucial for timely diagnosis and treatment, especially in regions with limited access to medical expertise. However, existing methods often overlook edge pixel information during tumor segmentation, leading to reduced boundary accuracy, and achieve high performance primarily on highly enhanced images, making them less effective for enhancement-lagging clinical data. To address these gaps, this study proposes the Edge Incorporative Fusion (EIF) algorithm, which enhances edge-pixel contrast in MRI images, combined with the Gabor Transform (GaT) for spatial-frequency domain conversion to improve detection accuracy. These innovations integrated into a Modified Deep Learning (MDL) architecture that reduces detection time by optimizing the internal layers while maintaining superior classification performance. The EIF-MDL system developed using Python programming language and implemented in Jupyter Notebook for flexibility and reproducibility. The system was evaluated on the PLCO and NU datasets, achieving sensitivity of 98.58%, specificity of 99.09%, accuracy of 99.1%, and a Dice similarity coefficient of 98.96%, outperforming state-of-the-art methods. This robust system’s ability to excel across both enhancement-lagging and highly enhanced images highlights its potential for accurate and timely glioma diagnosis, particularly in resource-constrained healthcare environments.

A multi-task self-supervised approach for mass detection in automated breast ultrasound using double attention recurrent residual U-Net.

A multi-task self-supervised approach for mass detection in automated breast ultrasound using double attention recurrent residual U-Net.

Cross-Attention Adaptive Feature Pyramid Network with Uncertainty Boundary Modeling for Mass Detection in Digital Breast Tomosynthesis.

Computer-aided detection (CADe) of masses in digital breast tomosynthesis (DBT) is crucial for early breast cancer diagnosis. However, the variability in the size and morphology of breast masses and their resemblance to surrounding tissues present significant challenges. Current CNN-based CADe methods, particularly those that use Feature Pyramid Networks (FPN), often fail to integrate multi-scale information effectively and struggle to handle dense glandular tissue with high-density or iso-density mass lesions due to the unidirectional integration and progressive attenuation of features, leading to high false positive rates. Additionally, the commonly indistinct boundaries of breast masses introduce uncertainty in boundary localization, which makes traditional Dirac boundary modeling insufficient for precise boundary regression. To address these issues, we propose the CU-Net network, which efficiently fuses multi-scale features and accurately models blurred boundaries. Specifically, the CU-Net introduces the Cross-Attention Adaptive Feature Pyramid Network (CA-FPN), which enhances the effectiveness and accuracy of feature interactions through a cross-attention mechanism to capture global correlations across multi-scale feature maps. Simultaneously, the Breast Density Perceptual Module (BDPM) incorporates breast density information to weight intermediate features, thereby improving the network's focus on dense breast regions susceptible to false positives. For blurred mass boundaries, we introduce Uncertainty Boundary Modeling (UBM) to model the positional distribution function of predicted bounding boxes for masses with uncertain boundaries. In comparative experiments on an in-house clinical DBT dataset and the BCS-DBT dataset, the proposed method achieved sensitivities of 89.68% and 72.73% at 2 false positives per DBT volume (FPs/DBT), respectively, significantly outperforming existing state-of-the-art detection methods. This method offers clinicians rapid, accurate, and objective diagnostic assistance, demonstrating substantial potential for clinical application.

Artificial intelligence in colonoscopy - where are we now in 2024?

Colonoscopy has a crucial role in reducing colorectal cancer incidence and mortality. Different artificial intelligence (AI) systems were developed to further improve its quality assurance (computer-aided quality improvement, CAQ), lesion detection (computer-aided detection, CADe) and lesion characterization (computer-aided characterization, CADx). There were studies investigating the roles of these AI systems in different domains of standard colonoscopies. In this state-of-the-art narrative review, we summarize the current evidence, discuss existing limitations, as well as explore the future directions of AI in colonoscopy. CAQ enhances colonoscopy quality through real-time feedback and quality monitoring systems, but the studies have inconsistent results due to small training datasets and varied methodologies. CADe increases adenoma detection rate and reduces adenoma missed rates but there are concerns about false positives, unnecessary polypectomies, potential de-skilling of endoscopists, and cost-effectiveness. CADx systems have mixed results and accuracies in differentiating polyp types, its use is further hindered by inadequate representation of sessile serrated lesions and a lack of rigorous trials comparing it with standard colonoscopy. Despite the emerging evidence of AI-assisted colonoscopy, its potential drawbacks and limitations may hinder the further implementation in real-world clinical practice. Long-term data on clinical efficacy, cost effectiveness, liability and data sharing are the key areas to be addressed.

PolyDeep Advance 1: Clinical Validation of a Computer-Aided Detection System for Colorectal Polyp Detection with a Second Observer Design.

Background: PolyDeep is a computer-aided detection and characterization system that has demonstrated a high diagnostic yield for in vitro detection of colorectal polyps. Our objective is to compare the diagnostic performance of expert endoscopists and PolyDeep for colorectal polyp detection. Methods: PolyDeep Advance 1 (NCT05514301) is an unicentric diagnostic test study with a second observer design. Endoscopists performed colonoscopy blinded to PolyDeep's detection results. The main endpoint was the sensitivity for colorectal polyp (adenoma, serrated or hyperplastic lesion) detection. The secondary endpoints were the diagnostic performance for diminutive lesions (≤5 mm), neoplasia (adenoma, serrated lesion) and adenoma detection. Results: We included 205 patients (55.1% male, 63.0 ± 6.2 years of age) referred to colonoscopy (positive faecal immunochemical occult blood test = 60.5%, surveillance colonoscopy = 39.5%). We excluded eight patients due to incomplete colonoscopy. Endoscopists detected 384 lesions, of which 39 were not detected by PolyDeep. In contrast, PolyDeep predicted 410 possible additional lesions, 26 of these predictions confirmed by endoscopists as lesions, resulting in a potential 6.8% detection increase with respect to the 384 lesions detected by the endoscopists. In total, 410 lesions were detected, 20 were not retrieved, five were colorectal adenocarcinoma, 343 were colorectal polyps (231 adenomas, 39 serrated and 73 hyperplastic polyps), 42 were normal mucosa and 289 were ≤5 mm. We did not find statistically significant differences between endoscopists and PolyDeep for colorectal polyp detection (Sensitivity = 94.2%, 91.5%, p = 0.2; Specificity = 9.5%, 14.3%, p = 0.7), diminutive lesions (Sensitivity = 92.3%, 89.5%, p = 0.4; Specificity = 9.8%, 14.6%, p = 0.7), neoplasia (Sensitivity = 95.2%, 92.9%, p = 0.3; Specificity = 9.6%, 13.9%, p = 0.4) and adenoma detection (Sensitivity = 94.4%, 92.6%, p = 0.5; Specificity = 7.2%, 11.8%, p = 0.2). Conclusions: Expert endoscopists and PolyDeep have similar diagnostic performance for colorectal polyp detection.

AI-Driven Skin Cancer Detection: A Deep Learning Perspective

The rising incidence of skin cancer necessitates the development of accurate and efficient diagnostic methods. This study introduces a novel approach for detecting skin cancer using an encoding-decoding technique with convolutional neural networks (CNNs) to enhance feature extraction from skin lesion images. Advanced CNN architectures, including DenseNet201, VGG16, and Xception, are utilized to classify skin lesions into seven categories. A thorough evaluation on a large dataset confirms the effectiveness of the proposed method in accurately identifying various types of skin cancer. Furthermore, a comparative study of multiple CNN models provides key insights into their relative strengths and limitations for diagnostic purposes. This research contributes to improving computer-aided skin cancer detection, paving the way for more reliable and accessible screening solutions.

Variability in computer-aided detection effect on adenoma detection rate in randomized controlled trials: A meta-regression analysis.

Variability in computer-aided detection effect on adenoma detection rate in randomized controlled trials: A meta-regression analysis.