Experienced Radiologists Research Articles

CT-based radiomics analysis for prediction of pathological subtypes of lung adenocarcinoma

ObjectiveLung cancer is a leading cause of cancer-related deaths worldwide, and its early and accurate diagnosis is crucial for improving patient survival. This paper developed an artificial intelligence (AI) model based on machine learning combined with radiomics for predicting the pathologic type of ground glass nodules (GGN). It explored its potential and challenges in practical applications. MethodsA total of 179 GGN patients with postoperative pathologically confirmed lung adenocarcinoma from June 2022 to June 2024 were collected from a hospital, including 22 cases of atypical adenomatous hyperplasia (AAH), 61 cases of adenocarcinoma in situ (AIS), 55 cases of invasive adenocarcinoma (IAC), and 41 cases of minimally invasive adenocarcinoma (MIA). Two experienced radiologists outlined the imaging data's regions of interest (ROI). Radiomic features were extracted and selected through normalization, mutual information, Spearman correlation coefficient, recursive feature elimination, and LASSO regression. Different machine learning models were developed and the best model was determined based on classification accuracy. ResultsDifferent machine learning models were trained and tested, and a variety of deep learning models were selected for comparison, among which Random Forest had the highest accuracy of 83.3%, and the AUC value of 0.892(95%CI, 0.846-0.923) in recognizing lung adenocarcinoma types. The AUC values reached 0.92 and 0.94 respectively in diagnosing AIS and IAC. ConclusionRadiomics combined with machine learning models, such as Random Forest, outperform average physician diagnostic accuracy in identifying lung adenocarcinoma types. The model is valuable for early and precise lung adenocarcinoma diagnosis, enhancing clinical decision-making.

Diagnostic Efficacy of Multimodal Ultrasound for Mass and Non-Mass Enhancements in Dynamic Contrast-Enhanced MRI of Idiopathic Granulomatous Mastitis and Breast Cancer

PurposeIdiopathic granulomatous mastitis (IGM) poses diagnostic challenges due to its diverse clinical and radiological presentations, often mimicking malignancies. This study aimed to assess the diagnostic efficacy of multimodal ultrasound for mass and non-mass enhancements in Dynamic Contrast-Enhanced MRI (DCE-MRI) of IGM and breast cancer. MethodsA retrospective analysis involved patients confirmed histopathologically with IGM and BC. All patients underwent conventional ultrasound (C-US), ultrasound elastography (UE), contrast-enhanced ultrasound (CEUS), and DCE-MR examinations. Blind experienced radiologists assessed imaging findings. Diagnostic accuracy, sensitivity, and specificity were calculated for mass and non-mass enhancements. ResultsFor mass enhancements (ME), multimodal ultrasound demonstrated strong efficacy (AUC = 0.8651, 95% CI: 0.7431 to 0.9871), exhibiting high sensitivity (83.3%) and specificity (92.4%) in differentiating IGM from breast cancer. However, for non-mass enhancements (NME), multimodal ultrasound showed limited accuracy (AUC = 0.6306) with lower sensitivity (65.6%) and specificity (81.2%) in distinguishing between IGM and breast cancer. ConclusionMultimodal ultrasound displayed good diagnostic efficacy for mass enhancements in DCE-MRI for idiopathic granulomatous mastitis and breast cancer, while for non-mass enhancement patterns, DCE-MRI remains the most valuable radiological modality for comprehensively assessing this condition's complexities.

Assessing the Performance of Artificial Intelligence Assistance for Prostate MRI: A Two-Center Study Involving Radiologists With Different Experience Levels.

Artificial intelligence (AI) assistance may enhance radiologists' performance in detecting clinically significant prostate cancer (csPCa) on MRI. Further validation is needed for radiologists with different experiences. To assess the performance of experienced and less-experienced radiologists in detecting csPCa, with and without AI assistance. Retrospective. Nine hundred patients who underwent prostate MRI and biopsy (median age 67 years; 356 with csPCa and 544 with non-csPCa). 3-T and 1.5-T, diffusion-weighted imaging using a single-shot gradient echo-planar sequence, turbo spin echo T2-weighted image. CsPCa regions based on biopsy results served as the reference standard. Ten less-experienced (<500 prostate MRIs) and six experienced (>1000 prostate MRIs) radiologists reviewed each case twice using Prostate Imaging Reporting and Data System v2.1, with and without AI, separated by 4-week intervals. Cases were equally distributed among less-experienced radiologists, and 90 cases were randomly assigned to each experienced radiologist. Reading time and diagnostic confidence were assessed. Area under the curve (AUC), sensitivity, specificity, reading time, and diagnostic confidence were compared using the DeLong test, Chi-squared test, Fisher exact test, or Wilcoxon rank-sum test between the two sessions. A P-value <0.05 was considered significant. Adjusting threshold using Bonferroni correction was performed for multiple comparisons. For less-experienced radiologists, AI assistance significantly improved lesion-level sensitivity (0.78 vs. 0.88), sextant-level AUC (0.84 vs. 0.93), and patient-level AUC (0.84 vs. 0.89). For experienced radiologists, AI assistance only improved sextant-level AUC (0.82 vs. 0.91). AI assistance significantly reduced median reading time (250 s [interquartile range, IQR: 157, 402] vs. 130 s [IQR: 88, 209]) and increased diagnostic confidence (5 [IQR: 4, 5] vs. 5 [IQR: 4, 5]) irrespective of experience and enhanced consistency among experienced radiologists (Fleiss κ: 0.53 vs. 0.61). AI-assisted reading improves the performance of detecting csPCa on MRI, particularly for less-experienced radiologists. 3 TECHNICAL EFFICACY: Stage 2.

Ultrasound S-detect system can improve diagnostic performance of less experienced radiologists in differentiating breast masses. A retrospective dual-center study.

To compare the performance of radiologists when assisted by an S-detect system with that of radiologists or an S-detect system alone in diagnosing breast masses on US images in a dual-center setting. US images were retrospectively identified 296 breast masses (150 benign, 146 malignant) by investigators at two medical centers. Six radiologists from the two centers independently analyzed the US images and classified each mass into categories 2 to 5. The radiologists then re-reviewed the images with the assistance of the S-detect system. The diagnostic value of radiologists alone, S-detect system alone, and S-detect alone, and radiologists + S-detect were analyzed and compared. Radiologists had significantly decreased the average false negative rate (FNR) for diagnosing breast masses using S-detect system (- 10.7%) (P < 0.001) and increased the area under the receiver operating characteristic curve (AUC) from 0.743 to 0.788 (P < 0.001). Seventy-seven out of 888 US images from six radiologists in this study were changed positively (from false negative to true positive or from false positive to true negative) with the S-detect, while 39 out of 888 US images were altered negatively. Radiologists had better performance for the diagnosis of malignant breast masses on US images with an S-detect system than without. The study used a dual-center design and compared the performance of the radiologists alone, the S-detect alone, and the radiologists combined with the S-detect. The study reported an improvement in sensitivity and AUC particularly for low to intermediate-level radiologists, involved cases and radiologists from two different centers, and compared the diagnostic value of using S-detect system for masses of different sizes.

Implementation of structured radiology reporting and its associated accuracy in comparison to pancreas multi-disciplinary clinic expert radiology review.

To evaluate the feasibility of implementation of structured reporting in the setting of a high-volume pancreatic multidisciplinary clinic (PMDC) and to assess its value by comparing the accuracy of structured reports with expert imaging reviews. A single institutional prospective cohort study was conducted during March 2022 to May 2024 to understand the feasibility of implementation of structured reporting (SR) for all patients who were seen in our weekly PMDC. Descriptive and regression analyses were performed to find an association between SR and difference in vascular involvement of the primary pancreatic tumor between the radiology report and expert radiologist review (gold standard) during PMDC. Among 466 patients seen in the PMDC, 426 (91.4%) had reports generated prior to PMDC. Of this, 294 reports met the inclusion criteria. The usage of SR increased from 58.3% in Mar 2022 to 87.8% in June 2024. Majority of the reports that used SR (n = 226, 76.9%), were performed for initial staging (n = 197, 67.0%) of PC. The median years of experience of reading radiologists that used non-SR was 14 (IQR: 8-27) years, while it was 9 (IQR -9-15) years for those who used SR (p = 0.030). Of note, as compared to the radiology report, increased vascular involvement was noted in PMDC review 62.5% (20 out of 32) of the time with non-SRs, whereas increased vascular involvement during PMDC review was noted in only 36.6% (48 out of 132) of the time with SRs. On multivariable analysis, using SR lowered the odds of increase in vascular involvement during PMDC review by 0.29 times (95CIs 0.11-0.79; p = 0.015). SR is feasible and superior to the free-text reporting with respect to the accuracy of peri-pancreatic vascular involvement. While its use cannot replace the PMDC radiology review, it can nonetheless be an indispensable tool in clinical management, particularly in a non-PMDC setting.

Different Evaluation Strategies of Oncological CT Examinations with Regard to Professional Experience: A Clinical Study Using Eye-tracking.

Contrast-enhanced CT is the standard imaging technique in oncological objectives. Rates of missed pathologies depend on work experience of the respective radiologists. Thus the aim of this study is to analyze the eye movements of professionals while reading CT images in order to evaluate whether the eye-fixation patterns and search strategies of experienced radiologists could explain higher detection rates of pathologies and whether such patterns can be learned.Anonymized images of 10 patients were presented to three medical students and six radiologists with different levels of work experience. During image analysis, ocular fixation positions were recorded using an eye-tracking software tool. The CT scans were analyzed retrospectively, considering the individual course of disease with the issue of successful detection of all pathologies. Visual attention and dwell time of ocular fixation on clinically important abnormalities or areas with pathological findings, general search patterns, and time efficiency were assessed. For statistical analysis, interobserver variability and accuracy of lesion detection were evaluated taking into account individual experience.The results revealed that observer sensitivity depends on work experience due to a more systematic order of inspection and a well-known course of disease, e.g. in case of metastatic spread. The areas of missed pathologies mostly included secondary findings. Inexperienced readers changed the stratification considerably more often and required more time for reporting or detecting pathologies.Our results suggest that experienced radiological physicians reduce their amount of missed findings by looking more systematically at images and by applying a more targeted inspection of clinically important regions. · CT interpretation by radiology residents is faster and less error-prone compared to postgraduate residents. · systematic image analysis is trainable. · engrams tend to be acquired through experience. · Kluge S, Schülke C, Ites HC et al. Different Evaluation Strategies of Oncological CT Examinations with Regard to Professional Experience: A Clinical Study Using Eye-tracking. Fortschr Röntgenstr 2024; DOI 10.1055/a-2452-2180.

Evaluation of Volumetric Magnetic Resonance Imaging in Determining the Indication for Total Neoadjuvant Therapy in Rectal Cancer.

This study aims to evaluate the relationship between volumetric measurements of residual tumor via magnetic resonance imaging (MRI) and pathologic complete response (pCR) in rectal cancer patients undergoing neoadjuvant chemoradiotherapy (nCRT). Patients with locally advanced rectal cancer who had pelvic MRI for clinical staging and completed nCRT followed by radical resection were included. Two experienced radiologists measured tumor volume on MRI obtained before and after nCRT. We compared the pre- and post-CRT tumor volume and measured tumor volume reduction rates. The median value of tumor volume reduction rate in all patients was 64.7% (min-max - 81.1-98.1%). When the relationship between tumor volume and tumor regression grade (TRG) after nCRT was assessed, it was found that 18 of 21 (86%) patients with a good response (TRG 1) had a post-CRT tumor volume of ≤ 8 cm3 (p = 0.001). While 9 of 10 patients with pCR after nCRT had a tumor volume of ≤ 8 cm3, one patient had pCR despite having a tumor volume greater than 8 cm3 (p = 0.015). The correlation between post-nCRT residual tumor volume and pCR underscores the potential of volumetric MRI as a predictive tool in tailoring rectal cancer treatment. For patients with residual tumor volumes greater than 8 cm3, extending neoadjuvant chemotherapy as part of TNT may enhance the likelihood of achieving pCR. This approach advocates for a more personalized treatment strategy, potentially optimizing outcomes for rectal cancer patients.

Abstract 4135968: Radiomic features of skeletal muscle on prostate T1-Weighted MRI associated with MACE outcomes in prostate cancer patients undergoing hormonal therapy

Background: Patients with high-risk prostate cancer (PC) who undergo hormonal therapy(HT) frequently die of major adverse cardiovascular events (MACE). Skeletal muscle on abdominal T1-weighted MRI (T1W) has been shown to be associated with cardiovascular fitness. Aims: To quantify heterogeneity associated with skeletal muscle tissue composition on prostate MRI patients undergoing HT using computationally derived radiomic features and identify their association with MACE. Methods: Cohort of N=58 men diagnosed with PC who underwent a 3 Tesla pre-treatment multi-parametric MRI followed by HT and followed up for MACE events were identified. MACE is defined as a composite of cardiovascular death, atrial fibrillation, myocardial infarction, ischemic stroke, heart failure, coronary artery disease, and peripheral vascular disease. Regions of interest (ROI) of skeletal muscle, including the obturator externus, pectineus, and the adductor group of muscles, were delineated on axial T1W under the guidance of an experienced radiologist. A set of 120 radiomic features, including gray level statistics, co-occurrence, wavelet, and texture energy-based features, were extracted within the ROI which quantifies underlying subvisual texture-based heterogeneity. Radiomics associated with MACE events were identified using the Wilcoxon rank-sum test and univariate Cox-regression (p&lt;0.05). Results: Of the N=58 patients included in this study, 18 experienced MACE. The most common MACE event was AF (8.6%). Median-time-to-MACE event was 7.3 years (6.2-8.2). 3 radiomic features (Table 1) quantifying the heterogeneity of skeletal muscle on prostate T1W were observed to be associated with MACE. Specifically, we observe that first order features have a lower response and wavelet features have a lower response in MACE patients which suggests a higher degree of local deformations and heterogeneity in skeletal tissue compared to non-MACE patients. Conclusions: Results from our study suggest that radiomics quantified texture characteristics of skeletal muscle tissue may indicate cardiovascular fitness, reflecting the likelihood of experiencing MACE following initiation of HT in PC men. Pre-treatment identification of PC patients at high risk of MACE may allow for alternative treatment options improving cardiovascular outcomes. Future studies will involve exploring larger, multi-site cohorts and machine learning modeling to determine cardiovascular risk.

The Impact of Artificial Intelligence on Radiologists' Reading Time in Bone Age Radiograph Assessment: A Preliminary Retrospective Observational Study.

To evaluate the real-world impact of artificial intelligence (AI) on radiologists' reading time during bone age (BA) radiograph assessments.Patients (<19 year-old) who underwent left-hand BA radiographs between December 2021 and October 2023 were retrospectively included. A commercial AI software was installed from October 2022. Radiologists' reading times, automatically recorded in the PACS log, were compared between the AI-unaided and AI-aided periods using linear regression tests and factors affecting reading time were identified.A total of 3643 radiographs (M:F=1295:2348, mean age 9.12 ± 2.31 years) were included and read by three radiologists, with 2937 radiographs (80.6%) in the AI-aided period. Overall reading times were significantly shorter in the AI-aided period compared to the AI-unaided period (mean 17.2 ± 12.9 seconds vs. mean 22.3 ± 14.7 seconds, p < 0.001). Staff reading times significantly decreased in the AI-aided period (mean 15.9 ± 11.4 seconds vs. mean 19.9 ± 13.4 seconds, p < 0.001), while resident reading times increased (mean 38.3 ± 16.4 seconds vs. 33.6 ± 15.3 seconds, p = 0.013). The use of AI and years of experience in radiology were significant factors affecting reading time (all, p≤0.001). The degree of decrease in reading time as experience increased was larger when utilizing AI (-1.151 for AI-unaided, -1.866 for AI-aided, difference =-0.715, p<0.001). In terms of AI exposure time, the staff's reading time decreased by 0.62 seconds per month (standard error 0.07, p<0.001) during the AI-aided period. The reading time of radiologists for BA assessment was influenced by AI. The time-saving effect of utilizing AI became more pronounced as the radiologists' experience and AI exposure time increased.

CNN-based glioma detection in MRI: A deep learning approach.

More than a million people are affected by brain tumors each year; high-grade gliomas (HGGs) and low-grade gliomas (LGGs) present serious diagnostic and treatment hurdles, resulting in shortened life expectancies. Glioma segmentation is still a significant difficulty in clinical settings, despite improvements in Magnetic Resonance Imaging (MRI) and diagnostic tools. Convolutional neural networks (CNNs) have seen recent advancements that offer promise for increasing segmentation accuracy, addressing the pressing need for improved diagnostic and therapeutic approaches. The study intended to develop an automated glioma segmentation algorithm using CNN to accurately identify tumor components in MRI images. The goal was to match the accuracy of experienced radiologists with commercial instruments, hence improving diagnostic precision and quantification. 285 MRI scans of high-grade gliomas (HGGs) and low-grade gliomas (LGGs) were analyzed in the study. T1-weighted sequences were utilised for segmentation both pre-and post-contrast agent administration, along with T2-weighted sequences (with and without Fluid Attenuation by Inversion Recovery [FAIRE]). The segmentation performance was assessed with a U-Net network, renowned for its efficacy in medical image segmentation. DICE coefficients were computed for the tumour core with contrast enhancement, the entire tumour, and the tumour nucleus without contrast enhancement. The U-Net network produced DICE values of 0.7331 for the tumour core with contrast enhancement, 0.8624 for the total tumour, and 0.7267 for the tumour nucleus without contrast enhancement. The results align with previous studies, demonstrating segmentation accuracy on par with professional radiologists and commercially accessible segmentation tools. The study developed a CNN-based automated segmentation system for gliomas, achieving high accuracy in recognising glioma components in MRI images. The results confirm the ability of CNNs to enhance the accuracy of brain tumour diagnoses, suggesting a promising avenue for future research in medical imaging and diagnostics. This advancement is expected to improve diagnostic processes for clinicians and patients by providing more precise and quantitative results.

Retrospective study on the strength of magnetic resonance signs for predicting breast implant rupture: assessing the impact of radiologist expertise at a breast cancer referral center.

This study evaluates the diagnostic criteria of MRI for breast implant rupture and examines the influence of radiologist experience on the accuracy of rupture detection. A retrospective study was conducted in a single center, involving patients who underwent implant replacement surgery between March 2019 and October 2022. MRI evaluations by four radiologists of varying experience levels were compared with surgical outcomes. The study included 118 explanted prostheses, with 25 identified as ruptured. Expert radiologists demonstrated near-perfect agreement (κ = 0.94) in identifying rupture signs, with high PPV and NPV, whereas non-experts showed only moderate agreement (κ = 0.44). Indicators such as the linguine sign, subcapsular line, and keyhole sign were identified with high accuracy by experts, contrasting with non-experts' lower detection rates. Expertise in radiology significantly impacts the accurate diagnosis of breast implant ruptures. This study supports updating radiological guidelines and underscores the importance of specialized training and experience in improving diagnostic outcomes in breast implant assessments.

Integrated Deep Learning Model for the Detection, Segmentation, and Morphologic Analysis of Intracranial Aneurysms Using CT Angiography.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop a deep learning model for the morphologic measurement of unruptured intracranial aneurysms (UIAs) based on CT angiography (CTA) data and validate its performance using a multicenter dataset. Materials and Methods In this retrospective study, patients with CTA examinations, including those with and without UIAs, in a tertiary referral hospital from February 2018 to February 2021 were included as the training dataset. Patients with UIAs who underwent CTA at multiple centers between April 2021 to December 2022 were included as the multicenter external testing set. An integrated deep-learning (IDL) model was developed for UIA detection, segmentation and morphologic measurement using an nnU-net algorithm. Model performance was evaluated using the Dice similarity coefficient (DSC) and intraclass correlation coefficient (ICC), with measurements by senior radiologists serving as the reference standard. The ability of the IDL model to improve performance of junior radiologists in measuring morphologic UIA features was assessed. Results The study included 1182 patients with UIAs and 578 controls without UIAs as the training dataset (55 years [IQR, 47-62], 1,012 [57.5%] females) and 535 patients with UIAs as the multicenter external testing set (57 years [IQR, 50-63], 353 [66.0%] females). The IDL model achieved 97% accuracy in detecting UIAs and achieved a DSC of 0.90 (95%CI, 0.88-0.92) for UIA segmentation. Model-based morphologic measurements showed good agreement with reference standard measurements (all ICCs > 0.85). Within the multicenter external testing set, the IDL model also showed agreement with reference standard measurements (all ICCs > 0.80). Junior radiologists assisted by the IDL model showed significantly improved performance in measuring UIA size (ICC improved from 0.88 [0.80-0.92] to 0.96 [0.92-0.97], P < .001). Conclusion The developed integrated deep learning model using CTA data showed good performance in UIA detection, segmentation and morphologic measurement and may be used to assist less experienced radiologists in morphologic analysis of UIAs. ©RSNA, 2024.

Neuroendocrine tumours and pregnancy: Real-world data from an European Neuroendocrine Tumour Centre of Excellence.

Neuroendocrine neoplasms (NENs) arise from the diffuse endocrine system and have been considered to be rare. However, the incidence and prevalence of these tumours have increased in recent years, and they are being seen in younger patients including women in the reproductive age group. Due to the paucity of data, diagnostic and therapeutic strategies in managing such tumours during pregnancy can be challenging to both treating physicians and patients. This article describes the experience and outcomes of managing pregnant women with NEN at a European Neuroendocrine Tumour Society (ENETS) Centre of Excellence. In this retrospective analysis, we evaluated a total of 22 pregnancies in 18 pregnant women with concurrent diagnoses of NENs who were managed at Royal Free Hospital ENETS Centre of Excellence throughout their pregnancy. These were identified from our tumour registry of 3500 NEN patients between 2015 and 2023. Cross-sectional imaging (computed tomography (CT)/magnetic resonance imaging (MRI)), pre- and post-pregnancy, for each patient was reviewed by an experienced radiologist. Tumour growth rate (TGR) was calculated using the formula: TGR = 100 × [exp (TG) - 1]; TG. [3 × log (D2/D1)]/time (months), where D1 is the tumour size at date 1; D2 is the tumour size at date 2; and time (months) = (Date 2 - Date 1 + 1)/30.44. Tumour growth rate pre-conception (TGRpc) and tumour growth rate post-partum (TGRpp) were calculated for each patient. In a sub-group of patients, positivity for oestrogen and progesterone receptors were analysed on the tumour tissue to evaluate whether the presence of these receptors affected tumour progression during the pregnancy. We also reviewed the pregnancy outcome in patients treated with somatostatin analogues during pregnancy. We analysed the data of a total 22 pregnancy encounters in 18 women: 15 pregnancies (68%) preceded the diagnosis of the NEN, whereas the diagnosis of NEN was made during pregnancy or in the post-partum period in 5 (23%) and 2 (9%) pregnancies respectively. Eight patients (44%) had a diagnosis of a pancreatic NEN, whereas 5 (28%) were diagnosed with mid-gut NENs, and a further 5 at other sites. The majority of the patients (n = 12, 67%) had evidence of metastatic disease at the time of diagnosis. Most pregnancies had a successful outcome (n = 19, 86%), whereas 3 patients (14%) had miscarriages in the 1st trimester. Five patients in total of 6 pregnancies were treated with somatostatin analogues as monotherapy during the pregnancy, and all of them had stable disease after pregnancy. All of them delivered healthy babies without any side effects or complications due to therapy. The average TGRpc was -0.8% (n = 5) and the average TGRpp was +0.96% (n = 6); 2 patients who did not have suitable targets for calculation of TGRpc developed new lesions suggesting disease progression. Moreover, 2 of the 4 patients who have had both pre-conception and post-pregnancy scans showed an increase in TGRpp compared to TGRpc. The management of NENs during pregnancy should be multidisciplinary with an individualised approach to each patient. Somatostatin analogues appear to be safe during pregnancy, though further robust studies are needed. Pregnancy per se may accelerate tumour progression, and patients should be counselled regarding this possibility.

The impact of radiologists’ characteristics on the detection of COVID-19 in chest CT scans

PurposeTo retrospectively assess the relationships between radiologists’ characteristics and their performance in interpreting chest CTs for COVID-19 detection. MethodsUsing DetectedX software, performance data were collected online from 74 radiologists from 29 countries, who each interpreted 30 anonymized digital chest CTs (15 positive and 15 negative scans for COVID-19). Radiologist-specific information was also collected: radiology experience, radiological specialty, number of lung CTs read weekly, familiarity with and education about COVID-19 presentation, and suspected COVID-19 cases seen weekly. The influence of this radiologist-specific information on the radiologists’ sensitivity, specificity, and ROC AUC was determined using regression analysis. ResultsRadiologists without respiratory imaging specialization had greater ROC AUCs (0.83 vs. 0.70, p = .006) and sensitivities (74.0% vs. 47.7%, p = .002) than their specialized peers. Radiologists without COVID-19 case encounters had greater sensitivity (80.6% vs. 63.1%, p = .017) but lower specificity (71.1% vs. 83.4%, p = .014) than those who encountered at least one case weekly. ConclusionSpecialization and prior experience with COVID-19 may impact the interpretation of suspected COVID-19 chest CTs. Clinical significanceThese results show the importance of diverse expertise and continuous training in managing novel diseases like COVID-19. Incorporating varied perspectives and experiences could enhance diagnostic accuracy and ensure a more comprehensive approach to disease management. In the case of novel diseases, specialization may not provide the same advantage as with more familiar disorders. The diagnostic value of specific experience with a novel disorder may help to compensate for a lack of specialization in the field, particularly in emergent situations. While specialization is important, generalization also holds value and utility in appropriate contexts.

Cross-Domain Denoising for Low-Dose Multi-Frame Spiral Computed Tomography.

Computed tomography (CT) has been used worldwide as a non-invasive test to assist in diagnosis. However, the ionizing nature of X-ray exposure raises concerns about potential health risks such as cancer. The desire for lower radiation doses has driven researchers to improve reconstruction quality. Although previous studies on low-dose computed tomography (LDCT) denoising have demonstrated the effectiveness of learning-based methods, most were developed on the simulated data. However, the real-world scenario differs significantly from the simulation domain, especially when using the multi-slice spiral scanner geometry. This paper proposes a two-stage method for the commercially available multi-slice spiral CT scanners that better exploits the complete reconstruction pipeline for LDCT denoising across different domains. Our approach makes good use of the high redundancy of multi-slice projections and the volumetric reconstructions while leveraging the over-smoothing issue in conventional cascaded frameworks caused by aggressive denoising. The dedicated design also provides a more explicit interpretation of the data flow. Extensive experiments on various datasets showed that the proposed method could remove up to 70% of noise without compromised spatial resolution, while subjective evaluations by two experienced radiologists further supported its superior performance against state-of-the-art methods in clinical practice. Code is available at https://github.com/YCL92/TMD-LDCT.

Mitigating Diagnostic Errors in Lung Cancer Classification: A Multi-Eyes Principle to Uncertainty Quantification.

In radiology, particularly in lung cancer diagnosis, diagnostic errors and cognitive biases pose substantial challenges. These issues, including perceptual errors, interpretive mistakes, and cognitive biases such as anchoring and premature closure, are often unnoticed by experienced radiologists. To address these challenges, we propose the Multi-Eyes principle approach, which utilises multiple deep learning models to reduce bias and potentially improve diagnostic accuracy. Inspired by the Four-Eyes principle in business and cybersecurity, this methodology employs various 3D and 2D (for validation) deep learning architectures and three uncertainty quantification techniques: Monte Carlo Dropout, Deep Ensemble, and Ensemble Monte Carlo Dropout. Each model functions as an independent reviewer, similar to blind reviews. With entropy selected as the uncertainty measurement, it is averaged, followed by ensemble averaging of predictions. The effectiveness of this approach was demonstrated using the LIDC-IDRI dataset for lung cancer classification. Statistical analysis of the uncertainty's distribution reveals that with more models, uncertainty in incorrect predictions becomes more peaked and left skewed, indicating consensus on uncertainty levels. This results in accuracy and F1 score improvements, even with the best performing model, addressing overconfidence in single-model systems. These findings highlight the potential of the Multi-Eyes principle to significantly improve diagnostic performance in computer-aided diagnostic systems. Future research may explore different uncertainty quantification methods and feedback mechanisms for further advancement.

Unusual Signal of Lymphadenopathy in Children with Nodular Sclerosing Hodgkin Lymphoma.

The current guidelines for initial cross-sectional imaging in pediatric lymphomas involve computed tomography (CT) of the chest, abdomen, and pelvis. However, whole-body magnetic resonance imaging (MRI) can be favored over CT for diagnosing and staging the disease, given its lack of ionizing radiation and its higher tissue contrast. Imaging characteristics of lymphoid tissue on MRI include a high T2/short tau inversion recovery (STIR) signal. A low or intermediate signal of lymphadenopathy on T2 and STIR images is an unexpected finding, noted anecdotally in nodular sclerosing Hodgkin lymphoma. This signal may be characteristic of a histological subtype of the disease and, if confirmed, could potentially be used to avoid biopsy. In this study, we aimed to review signal characteristics of lymphadenopathy in patients with biopsy-confirmed nodular sclerosing Hodgkin lymphoma. We undertook a retrospective review of relevant MR studies of patients with nodular sclerosing Hodgkin lymphoma. Studies were reviewed by an experienced pediatric radiologist regarding lymph node signal, especially on T2/STIR. Eleven children with nodular sclerosing Hodgkin lymphoma were included. Median age at the time of MRI was 14.3 (IQR: 13.9-16.1) years, and nine were boys. Five patients showed some lymphadenopathy with a low T2/STIR signal, and six showed an intermediate T2/STIR signal. Central gadolinium non-enhancement was observed in four patients. All eleven patients (100%) with a diagnosis of nodular sclerosing Hodgkin lymphoma showed some lymphadenopathy with a low or intermediate T2/STIR signal, and five children (45.5%) showed a frank low signal of some lymphadenopathy, a feature which may prove to be a biomarker for this histology.

Automatic detection of temporomandibular joint osteoarthritis radiographic features using deep learning artificial intelligence. A Diagnostic accuracy study

ObjectiveThe purpose of this study was to investigate the diagnostic performance of a neural network Artificial Intelligence model for the radiographic confirmation of Temporomandibular Joint Osteoarthritis in reference to an experienced radiologist. Materials and MethodsThe diagnostic performance of an AI model in identifying radiographic features in patients with TMJ-OA was evaluated in a diagnostic accuracy cohort study. Adult patients elected for radiographic examination by the Diagnostic Criteria for Temporomandibular Disorders decision tree were included. Cone-beam computed Tomography images were evaluated by object detection YOLO deep learning model. The diagnostic performance was verified against examiner radiographic evaluation. ResultsThe differences between the AI model and examiner were non-significant statistically, except in the subcortical cyst (P=0.049*). AI model showed substantial to near-perfect levels of agreement when compared to those of the examiner data. Regarding each radiographic phenotype, the AI model reported favorable sensitivity, specificity, accuracy, and highly statistically significant Receiver Operating Characteristic (ROC) analysis (p<0.001). Area Under Curve ranged from 0.872, for surface erosion, to 0.911 for subcortical cyst. ConclusionAI object detection model could open the horizon for a valid, automated, and convenient modality for TMJ-OA radiographic confirmation and radiomic features identification with a significant diagnostic power.

Deep learning-based segmentation of kidneys and renal cysts on T2-weighted MRI from patients with autosomal dominant polycystic kidney disease

BackgroundOur aim was to train and test a deep learning-based algorithm for automatically segmenting kidneys and renal cysts in patients with autosomal dominant polycystic kidney disease (ADPKD).MethodsWe retrospectively selected all ADPKD patients who underwent renal MRI with coronal T2-weighted imaging at our institution from 2008 to 2022. The 20 most recent examinations constituted the test dataset, to mimic pseudoprospective enrolment. The remaining ones constituted the training dataset to which eight normal renal MRIs were added. Kidneys and cysts ground truth segmentations were performed on coronal T2-weighted images by a junior radiologist supervised by an experienced radiologist. Kidneys and cysts of the 20 test MRIs were segmented by the algorithm and three independent human raters. Segmentations were compared using overlap metrics. The total kidney volume (TKV), total cystic volume (TCV), and cystic index (TCV divided by TKV) were compared using Bland–Altman analysis.ResultsWe included 164 ADPKD patients. Dice similarity coefficients ranged from 85.9% to 87.4% between the algorithms and the raters’ segmentations and from 84.2% to 86.2% across raters’ segmentations. For TCV assessment, the biases ± standard deviations (SD) were 3–19 ± 137–151 mL between the algorithm and the raters, and 22–45 ± 49–57 mL across raters. The algorithm underestimated TKV and TCV in two outliers with TCV > 2800 mL. For cystic index assessment, the biases ± SD were 2.5–6.9% ± 6.7–8.3% between the algorithm and the raters, and 2.1–9.4 ± 7.4–11.6% across raters.ConclusionThe algorithm’s performance fell within the range of inter-rater variability, but large TKV and TCV were underestimated.Relevance statementAccurate automated segmentation of the renal cysts will enable the large-scale evaluation of the prognostic value of TCV and cystic index in ADPKD patients. If these biomarkers are prognostic, then automated segmentation will facilitate their use in daily routine.Key PointsCystic volume is an emerging biomarker in ADPKD.The algorithm’s performance in segmenting kidneys and cysts fell within interrater variability.The segmentation of very large cysts, under-represented in the training dataset, needs improvement.Graphical

Grad-CAM Enabled Breast Cancer Classification with a 3D Inception-ResNet V2: Empowering Radiologists with Explainable Insights.

Breast cancer (BCa) poses a severe threat to women's health worldwide as it is the most frequently diagnosed type of cancer and the primary cause of death for female patients. The biopsy procedure remains the gold standard for accurate and effective diagnosis of BCa. However, its adverse effects, such as invasiveness, bleeding, infection, and reporting time, keep this procedure as a last resort for diagnosis. A mammogram is considered the routine noninvasive imaging-based procedure for diagnosing BCa, mitigating the need for biopsies; however, it might be prone to subjectivity depending on the radiologist's experience. Therefore, we propose a novel, mammogram image-based BCa explainable AI (BCaXAI) model with a deep learning-based framework for precise, noninvasive, objective, and timely manner diagnosis of BCa. The proposed BCaXAI leverages the Inception-ResNet V2 architecture, where the integration of explainable AI components, such as Grad-CAM, provides radiologists with valuable visual insights into the model's decision-making process, fostering trust and confidence in the AI-based system. Based on using the DDSM and CBIS-DDSM mammogram datasets, BCaXAI achieved exceptional performance, surpassing traditional models such as ResNet50 and VGG16. The model demonstrated superior accuracy (98.53%), recall (98.53%), precision (98.40%), F1-score (98.43%), and AUROC (0.9933), highlighting its effectiveness in distinguishing between benign and malignant cases. These promising results could alleviate the diagnostic subjectivity that might arise as a result of the experience-variability between different radiologists, as well as minimize the need for repetitive biopsy procedures.