This study evaluated GPT-4's accuracy in MRI sequence selection based on radiology request forms (RRFs), comparing its performance to radiology residents. This retrospective study included 100 RRFs across four subspecialties (cardiac imaging, neuroradiology, musculoskeletal, and oncology). GPT-4 and two radiology residents (R1: 2 years, R2: 5 years MRI experience) selected sequences based on each patient's medical history and clinical questions. Considering imaging society guidelines, five board-certified specialized radiologists assessed protocols based on completeness, quality, and utility in consensus, using 5-point Likert scales. Clinical applicability was rated binarily by the institution's lead radiographer. GPT-4 achieved median scores of 3 (1-5) for completeness, 4 (1-5) for quality, and 4 (1-5) for utility, comparable to R1 (3 (1-5), 4 (1-5), 4 (1-5); each p > 0.05) but inferior to R2 (4 (1-5), 5 (1-5); p < 0.01, respectively, and 5 (1-5); p < 0.001). Subspecialty protocol quality varied: GPT-4 matched R1 (4 (2-4) vs. 4 (2-5), p = 0.20) and R2 (4 (2-5); p = 0.47) in cardiac imaging; showed no differences in neuroradiology (all 5 (1-5), p > 0.05); scored lower than R1 and R2 in musculoskeletal imaging (3 (2-5) vs. 4 (3-5); p < 0.01, and 5 (3-5); p < 0.001); and matched R1 (4 (1-5) vs. 2 (1-4), p = 0.12) as well as R2 (5 (2-5); p = 0.20) in oncology. GPT-4-based protocols were clinically applicable in 95% of cases, comparable to R1 (95%) and R2 (96%). GPT-4 generated MRI protocols with notable completeness, quality, utility, and clinical applicability, excelling in standardized subspecialties like cardiac and neuroradiology imaging while yielding lower accuracy in musculoskeletal examinations. Question Long MRI acquisition times limit patient access, making accurate protocol selection crucial for efficient diagnostics, though it's time-consuming and error-prone, especially for inexperienced residents. Findings GPT-4 generated MRI protocols of remarkable yet inconsistent quality, performing on par with an experienced resident in standardized fields, but moderately in musculoskeletal examinations. Clinical relevance The large language model can assist less experienced radiologists in determining detailed MRI protocols and counteract increasing workloads. The model could function as a semi-automatic tool, generating MRI protocols for radiologists' confirmation, optimizing resource allocation, and improving diagnostics and cost-effectiveness.

Clinical magnetic resonance imaging (MRI) is a high-resolution tool widely used for detailed anatomical imaging. However, prolonged scan times often lead to motion artefacts and patient discomfort. Fast acquisition techniques can reduce scan times but often produce noisy, low-contrast images, compromising segmentation accuracy essential for diagnosis and treatment planning. To address these limitations, we developed an end-to-end framework that incorporates BIDS-based data organiser and anonymizer, a GAN-based MR image enhancement model (GAN-MRI), AssemblyNet for brain region segmentation, and an attention-residual U-Net with Guided loss for abdominal and thigh segmentation. Thirty brain scans (5,400 slices) and 32 abdominal (1,920 slices) and 55 thigh scans (2,200 slices) acquired from multiple MRI scanners (GE, Siemens, Toshiba) underwent evaluation. Image quality improved significantly, with SNR and CNR for brain scans increasing from 28.44 to 42.92 (p < 0.001) and 11.88 to 18.03 (p < 0.001), respectively. Abdominal scans exhibited SNR increases from 35.30 to 50.24 (p < 0.001) and CNR from 10,290.93 to 93,767.22 (p < 0.001). Double-blind evaluations highlighted improved visualisations of anatomical structures and bias field correction. Segmentation performance improved substantially in the thigh (muscle: + 21%, IMAT: + 9%) and abdominal regions (SSAT: + 1%, DSAT: + 2%, VAT: + 12%), while brain segmentation metrics remained largely stable, reflecting the robustness of the baseline model. Proposed framework is designed to handle data from multiple anatomies with variations from different MRI scanners and centres by enhancing MRI scan and improving segmentation accuracy, diagnostic precision and treatment planning while reducing scan times and maintaining patient comfort.

Preoperative T-staging in rectal cancer is essential for treatment planning, yet conventional MRI shows limited accuracy (~ 60-78). Our study investigates whether radiomic analysis of high-resolution T2-weighted MRI can non-invasively improve staging accuracy through a retrospective evaluation in a real-world surgical cohort. This single-center retrospective study included 200 patients (January 2024-April 2025) with pathologically confirmed rectal cancer, all undergoing preoperative high-resolution T2-weighted MRI within one week prior to curative surgery and no neoadjuvant therapy. Manual segmentation was performed using ITK‑SNAP, followed by extraction of 107 radiomic features via PyRadiomics. Feature selection employed mRMR and LASSO logistic regression, culminating in a Rad-score predictive model. Statistical performance was evaluated using ROC curves (AUC), accuracy, sensitivity, specificity, and Delong's test. Among 200 patients, 95 were pathologically staged as T2 and 105 as T3-T4 (55 T3, 50 T4). After preprocessing, 26 radiomic features were retained; key features including ngtdm_contrast and ngtdm_coarseness showed AUC values > 0.70. The LASSO-based model achieved an AUC of 0.82 (95% CI: 0.75-0.89), with overall accuracy of 81%, sensitivity of 78%, and specificity of 84%. Radiomic analysis of standard preoperative T2-weighted MRI provides a reliable, non-invasive method to predict rectal cancer T-stage. This approach has the potential to enhance staging accuracy and inform personalized surgical planning. Prospective multicenter validation is required for broader clinical implementation.

Amyotrophic Lateral Sclerosis is a devastating motor neuron disease characterized by its diagnostic difficulty. Currently, no reliable biomarkers exist in the diagnosis process. In this scenario, our purpose is the application of machine learning algorithms to imaging MRI-derived variables for the development of diagnostic models that facilitate and shorten the process. A dataset of 211 patients (114 ALS, 45 mimic, 22 genetic carriers and 30 control) with MRI-derived features of volumetry, cortical thickness and local iron (via T2* mapping, and visual assessment of susceptibility imaging). A binary classification task approach has been taken to classify patients with and without ALS. A sequential modeling methodology, understood from an iterative improvement perspective, has been followed, analyzing each group's performance separately to adequately improve modelling. Feature filtering techniques, dimensionality reduction techniques (PCA, kernel PCA), oversampling techniques (SMOTE, ADASYN) and classification techniques (logistic regression, LASSO, Ridge, ElasticNet, Support Vector Classifier, K-neighbors, random forest) were included. Three subsets of available data have been used for each proposed architecture: a subset containing automatic retrieval MRI-derived data, a subset containing the variables from the visual analysis of the susceptibility imaging and a subset containing all features. The best results have been attained with all the available data through a voting classifier composed of five different classifiers: accuracy = 0.896, AUC = 0.929, sensitivity = 0.886, specificity = 0.929. These results confirm the potential of ML techniques applied to imaging variables of volumetry, cortical thickness, and local iron for the development of diagnostic model as a clinical tool for decision-making support.

This study aimed to develop and validate a hybrid deep learning (DL) model that integrates convolutional neural network (CNN) and vision transformer (ViT) architectures to predict distant metastasis (DM) in patients with non-small cell lung cancer (NSCLC) using ¹⁸F-FDG PET/CT images. A retrospective analysis was conducted on a cohort of consecutively registered patients who were newly diagnosed and untreated for NSCLC. A total of 167 patients with available PET/CT images were included in the analysis. DL features were extracted using a combination of CNN and ViT architectures, followed by feature selection, model construction, and evaluation of model performance using the receiver operating characteristic (ROC) and the area under the curve (AUC). The ViT-based DL model exhibited strong predictive capabilities in both the training and validation cohorts, achieving AUCs of 0.824 and 0.830 for CT features, and 0.602 and 0.694 for PET features, respectively. Notably, the model that integrated both PET and CT features demonstrated a notable AUC of 0.882 in the validation cohort, outperforming models that utilized either PET or CT features alone. Furthermore, this model outperformed the CNN model (ResNet 50), which achieved an AUC of 0.752 [95% CI 0.613, 0.890], p < 0.05. Decision curve analysis further supported the efficacy of the ViT-based DL model. The ViT-based DL developed in this study demonstrates considerable potential in predicting DM in patients with NSCLC, potentially informing the creation of personalized treatment strategies. Future validation through prospective studies with larger cohorts is necessary.

Community-acquired pneumonia (CAP) is a significant concern for children worldwide and is associated with a high morbidity and mortality. To improve patient outcomes, early intervention and accurate diagnosis are essential. Artificial intelligence (AI) can mine and label imaging data and thus may contribute to precision research and personalized clinical management. The baseline characteristics of 230 children with severe CAP hospitalized from January 2023 to October 2024 were retrospectively analyzed. The patients were divided into two groups according to the presence of respiratory failure. The predictive ability of AI-derived chest CT (computed tomography) indices alone for respiratory failure was assessed via logistic regression analysis. ROC (receiver operating characteristic) curves were plotted for these regression models. After adjusting for age, white blood cell count, neutrophils, lymphocytes, creatinine, wheezing, and fever > 5 days, a greater number of involved lung lobes [odds ratio 1.347, 95% confidence interval (95% CI) 1.036-1.750, P = 0.026] and bilateral lung involvement (odds ratio 2.734, 95% CI 1.084-6.893, P = 0.033) were significantly associated with respiratory failure. The discriminatory power (as measured by the area under curve) of Model 2 and Model 3, which included electronic health record data and the accuracy of CT imaging features, was better than that of Model 0 and Model 1, which contained only the chest CT parameters. The sensitivity and specificity of Model 2 at the optimal critical value (0.441) were 84.3% and 59.8%, respectively. The sensitivity and specificity of Model 3 at the optimal critical value (0.446) were 68.6% and 76.0%, respectively. The use of AI-derived chest CT indices may achieve high diagnostic accuracy and guide precise interventions for patients with severe CAP. However, clinical, laboratory, and AI-derived chest CT indices should be included to accurately predict and treat severe CAP.

Accurate thyroid volume assessment is critical in thyroid disease diagnostics, yet conventional high-resolution 2D ultrasound has limitations. Freehand 3D ultrasound with AI-assisted segmentation presents a potential advancement, but its clinical accuracy requires validation. This prospective clinical trial included 14 patients scheduled for total thyroidectomy. Preoperative thyroid volume was measured using both 2D ultrasound (ellipsoid method) and freehand 3D ultrasound with AI segmentation. Postoperative thyroid volume, determined via the water displacement method, served as the reference standard. The median postoperative thyroid volume was 14.8 mL (IQR 8.8-20.2). The median volume difference was 1.7 mL (IQR 1.2-3.3) for 3D ultrasound and 3.6 mL (IQR 2.3-6.6) for 2D ultrasound (p = 0.02). The inter-operator reliability coefficient for 3D ultrasound was 0.986 (p < 0.001). These findings suggest that freehand 3D ultrasound with AI-assisted segmentation provides superior accuracy and reproducibility compared to 2D ultrasound and may enhance clinical thyroid volume assessment. ClinicalTrials.gov identifier: NCT05510609.

BackgroundAlzheimer's disease (AD) is an irreversible neurodegenerative disorder characterized by progressive cognitive and memory decline. Accurate prediction of high-risk individuals enables early detection and better patient care.ObjectiveThis study aims to enhance MRI-based AD classification through advanced image preprocessing, optimal feature selection, and ensemble deep learning techniques.MethodsThe study employs advanced image preprocessing techniques such as normalization, affine transformation, and denoising to improve MRI quality. Brain structure segmentation is performed using the adaptive DeepLabV3 + approach for precise AD diagnosis. A novel optimal feature selection framework, H-IBMFO, integrates the Improved Beluga Whale Optimizer and Manta Foraging Optimization. An ensemble deep learning model combining MobileNet V2, DarkNet, and ResNet is used for classification. MATLAB is utilized for implementation.ResultsThe proposed system achieves 98.7% accuracy, with 98% precision, 98% sensitivity, 99% specificity, and 98% F-measure, demonstrating superior classification performance with minimal false positives and negatives.ConclusionsThe study establishes an efficient framework for AD classification, significantly improving early detection through optimized feature selection and deep learning. The high accuracy and reliability of the system validate its effectiveness in diagnosing AD stages.

To evaluate the utility of a deep learning (DL)-based image enhancement for improving the image quality and diagnostic performance of 3D contrast-enhanced T1-weighted black blood (BB) MR imaging for brain metastases. This retrospective study included 126 patients with and 121 patients without brain metastasis who underwent 3-T MRI examinations. Commercially available DL-based MR image enhancement software was utilized for image post-processing. The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of enhancing lesions were measured. For qualitative assessment and diagnostic performance evaluation, two radiologists graded the overall image quality, noise, and artifacts of each image and the conspicuity of visible lesions. The Wilcoxon signed-rank test and regression analyses with generalized estimating equations (GEEs) were used for statistical analysis. For MR images that were not previously processed using other DL-based methods, SNR and CNR were higher in the DL-enhanced images than in the standard images (438.3 vs. 661.1, p < 0.01; 173.9 vs. 223.5, p < 0.01). Overall image quality and noise were improved in the DL images (p < 0.01, average score-5 proportion 38% vs. 65%; p < 0.01, 43% vs. 74%), whereas artifacts did not significantly differ (p ≥ 0.07). Sensitivity was increased after post-processing from 79 to 86% (p = 0.02), especially for lesions smaller than 5 mm (69 to 78%, p = 0.03), and changes in specificity (p = 0.24) and average false-positive (FP) count (p = 0.18) were not significant. DL image enhancement improves the image quality and diagnostic performance of 3D contrast-enhanced T1-weighted BB MR imaging for the detection of small brain metastases. Question Can deep learning (DL)-based image enhancement improve the image quality and diagnostic performance of 3D contrast-enhanced T1-weighted black blood (BB) MR imaging for brain metastases? Findings DL-based image enhancement improved image quality of thin slice BB MR images and sensitivity for brain metastasis, particularly for lesions smaller than 5 mm. Clinical relevance DL-based image enhancement on BB images may assist in the accurate diagnosis of brain metastasis by achieving better sensitivity while maintaining comparable specificity.

By carefully analyzing latent image properties, content-based image retrieval (CBIR) systems are able to recover pertinent images without relying on text descriptions, natural language tags, or keywords related to the image. This search procedure makes it quite easy to automatically retrieve images in huge, well-balanced datasets. However, in the medical field, such datasets are usually not available. This study proposed an advanced DL technique to enhance the accuracy of image retrieval in complex medical datasets. The proposed model can be integrated into five stages, namely pre-processing, decomposing the images, feature extraction, dimensionality reduction, and classification with an image retrieval mechanism. The hybridized Wavelet-Hadamard Transform (HWHT) was utilized to obtain both low and high frequency detail for analysis. In order to extract the main characteristics, the Gray Level Co-occurrence Matrix (GLCM) was employed. Furthermore, to minimize feature complexity, Sine chaos based artificial rabbit optimization (SCARO) was utilized. By employing the Bhattacharyya Coefficient for improved similarity matching, the Bhattacharya Context performance aware global attention-based Transformer (BCGAT) improves classification accuracy. The experimental results proved that the COVID-19 Chest X-ray image dataset attained higher accuracy, precision, recall, and F1-Score of 99.5%, 97.1%, 97.1%, and 97.1%, 97.1%, respectively. However, the chest x-ray image (pneumonia) dataset has attained higher accuracy, precision, recall, and F1-score values of 98.60%, 98.49%, 97.40%, and 98.50%, respectively. For the NIH chest X-ray dataset, the accuracy value is 99.67%.