Assess the feasibility of using a large language model (LLM) to identify valuable radiology teaching cases through report discrepancy detection. Retrospective study included after-hours head CT and musculoskeletal radiograph exams from January 2017 to December 2021. Discrepancy level between trainee's preliminary interpretation and final attending report was annotated on a 5-point scale. RadBERT, an LLM pretrained on a vast corpus of radiology text, was fine-tuned for discrepancy detection. For comparison and to ensure the robustness of the approach, Mixstral 8×7B, Mistral 7B, and Llama2 were also evaluated. The model's performance in detecting discrepancies was evaluated using a randomly selected hold-out test set. A subset of discrepant cases identified by the LLM was compared to a random case set by recording clinical parameters, discrepant pathology, and evaluating possible educational value. F1 statistic was used for model comparison. Pearson's chi-squared test was employed to assess discrepancy prevalence and score between groups (significance set at p<0.05). The fine-tuned LLM model achieved an overall accuracy of 90.5% with a specificity of 95.5% and a sensitivity of 66.3% for discrepancy detection. The model sensitivity significantly improved with higher discrepancy scores, 49% (34/70) for score 2 versus 67% (47/62) for score 3, and 81% (35/43) for score 4/5 (p<0.05 compared to score 2). LLM-curated set showed a significant increase in the prevalence of all discrepancies and major discrepancies (scores 4 or 5) compared to a random case set (P<0.05 for both). Evaluation of the clinical characteristics from both the random and discrepant case sets demonstrated a broad mix of pathologies and discrepancy types. An LLM can detect trainee report discrepancies, including both higher and lower-scoring discrepancies, and may improve case set curation for resident education as well as serve as a trainee oversight tool.

Echo features of lymph nodes (LNs) influence target selection during endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA). This study evaluates deep learning's diagnostic capabilities on EBUS images for detecting mediastinal LN metastasis in lung cancer, emphasising the added value of integrating a region of interest (ROI), LN size on CT, and PET-CT findings. We analysed 2901 EBUS images from 2055 mediastinal LN stations in 1454 lung cancer patients. ResNet18-based deep learning models were developed to classify images of true positive malignant and true negative benign LNs diagnosed by EBUS-TBNA using different inputs: original images, ROI images, and CT size and PET-CT data. Model performance was assessed using the area under the receiver operating characteristic curve (AUROC) and other diagnostic metrics. The model using only original EBUS images showed the lowest AUROC (0.870) and accuracy (80.7%) in classifying LN images. Adding ROI information slightly increased the AUROC (0.896) without a significant difference (p = 0.110). Further adding CT size resulted in a minimal change in AUROC (0.897), while adding PET-CT (original + ROI + PET-CT) showed a significant improvement (0.912, p = 0.008 vs. original; p = 0.002 vs. original + ROI + CT size). The model combining original and ROI EBUS images with CT size and PET-CT findings achieved the highest AUROC (0.914, p = 0.005 vs. original; p = 0.018 vs. original + ROI + PET-CT) and accuracy (82.3%). Integrating an ROI, LN size on CT, and PET-CT findings into the deep learning analysis of EBUS images significantly enhances the diagnostic capability of models for detecting mediastinal LN metastasis in lung cancer, with the integration of PET-CT data having a substantial impact.

The SAM (segment anything model) is a foundation model for general purpose image segmentation, however, when it comes to a specific medical application, such as segmentation of both ventricles from the 2D cardiac MRI, the results are not satisfactory. The scarcity of labeled medical image data further increases the difficulty to apply the SAM to medical image processing. To address these challenges, we propose the SAMBV by fine-tuning the SAM for semi-supervised segmentation of bi-ventricle from the 2D cardiac MRI. The SAM is tuned in three aspects, (i) the position and feature adapters are introduced so that the SAM can adapt to bi-ventricle segmentation. (ii) a dual-branch encoder is incorporated to collect missing local feature information in SAM so as to improve bi-ventricle segmentation. (iii) the interpolation consistency regularization (ICR) semi-supervised manner is utilized, allowing the SAMBV to achieve competitive performance with only 40% of the labeled data in the ACDC dataset. Experimental results demonstrate that the proposed SAMBV achieves an average Dice score improvement of 17.6% over the original SAM, raising its performance from 74.49% to 92.09%. Furthermore, the SAMBV outperforms other supervised SAM fine-tuning methods, showing its effectiveness in semi-supervised medical image segmentation tasks. Notably, the proposed method is specifically designed for 2D MRI data.

Artificial intelligence (AI) is rapidly transforming cardiovascular imaging, offering innovative solutions to enhance diagnostic precision, prognostic accuracy, and therapeutic decision-making. This review explores the role of AI in cardiovascular imaging, highlighting its applications, advantages, limitations, and future challenges. The discussion is structured by imaging modalities, including echocardiography, cardiac and coronary computed tomography, cardiac magnetic resonance, and nuclear cardiology. For each modality, we examine AI's contributions across the patient care continuum: from patient selection and image acquisition to quantitative and qualitative analysis, interpretation support, prognostic stratification, therapeutic guidance, and integration with other clinical data. AI applications demonstrate significant potential to streamline workflows, improve diagnostic accuracy, and provide advanced insights for complex clinical scenarios. However, several limitations must be addressed. Many AI algorithms are developed using data from single, high-expertise centers, raising concerns about their generalizability to routine clinical practice. In some cases, these algorithms may even produce misleading results. Additionally, the "black box" nature of certain AI systems poses challenges for cardiologists, making discrepancies difficult to interpret or rectify. Importantly, AI should be seen as a complementary tool rather than a replacement for cardiologists, designed to expedite routine tasks and allow clinicians to focus on complex cases. Future challenges include fostering clinician involvement in algorithm development and extending AI implementation to peripheral healthcare centers. This approach aims to enhance accessibility, understanding, and applicability of AI in everyday clinical practice, ultimately democratizing its benefits and ensuring equitable integration into healthcare systems.

The management of skeletal-related events (SREs), particularly the prevention of pathological fractures, is crucial for cancer patients. Current clinical assessment of fracture risk is mostly based on medical images, but incorporating sequential images in the assessment remains challenging. This study addressed this issue by leveraging a comprehensive dataset consisting of 260 longitudinal micro-computed tomography (μCT) scans acquired in normal and breast cancer bearing mice. A machine learning (ML) model based on a spatial-temporal neural network was built to forecast bone structures from previous μCT scans, which were found to have an overall similarity coefficient (Dice) of 0.814 with ground truths. Despite the predicted lesion volumes (18.5 ​% ​± ​15.3 ​%) being underestimated by ∼21 ​% than the ground truths' (22.1 ​% ​± ​14.8 ​%), the time course of the lesion growth was better represented in the predicted images than the preceding scans (10.8 ​% ​± ​6.5 ​%). Under virtual biomechanical testing using finite element analysis (FEA), the predicted bone structures recapitulated the loading carrying behaviors of the ground truth structures with a positive correlation (y ​= ​0.863x) and a high coefficient of determination (R² ​= ​0.955). Interestingly, the compliances of the predicted and ground truth structures demonstrated nearly identical linear relationships with the lesion volumes. In summary, we have demonstrated that bone deterioration could be proficiently predicted using machine learning in our preclinical dataset, suggesting the importance of large longitudinal clinical imaging datasets in fracture risk assessment for cancer bone metastasis.

To evaluate the performance of large language models (LLMs) in generating clinically follow-up recommendations for pulmonary nodules by leveraging radiological report findings and management guidelines. This retrospective study included CT follow-up reports of pulmonary nodules documented by senior radiologists from September 1st, 2023, to April 30th, 2024. Sixty reports were collected for prompting engineering additionally, based on few-shot learning and the Chain of Thought methodology. Radiological findings of pulmonary nodules, along with finally prompt, were input into GPT-4o-mini or ERNIE-4.0-Turbo-8K to generate follow-up recommendations. The AI-generated recommendations were evaluated against radiologist-defined guideline-based standards through binary classification, assessing nodule risk classifications, follow-up intervals, and harmfulness. Performance metrics included sensitivity, specificity, positive/negative predictive values, and F1 score. On 1009 reports from 996 patients (median age, 50.0 years, IQR, 39.0-60.0 years; 511 male patients), ERNIE-4.0-Turbo-8K and GPT-4o-mini demonstrated comparable performance in both accuracy of follow-up recommendations (94.6 % vs 92.8 %, P = 0.07) and harmfulness rates (2.9 % vs 3.5 %, P = 0.48). In nodules classification, ERNIE-4.0-Turbo-8K and GPT-4o-mini performed similarly with accuracy rates of 99.8 % vs 99.9 % sensitivity of 96.9 % vs 100.0 %, specificity of 99.9 % vs 99.9 %, positive predictive value of 96.9 % vs 96.9 %, negative predictive value of 100.0 % vs 99.9 %, f1-score of 96.9 % vs 98.4 %, respectively. LLMs show promise in providing guideline-based follow-up recommendations for pulmonary nodules, but require rigorous validation and supervision to mitigate potential clinical risks. This study offers insights into their potential role in automated radiological decision support.

Accurate and timely diagnosis of brain tumors is critical for patient management and treatment planning. Magnetic resonance imaging (MRI) is a widely used modality for brain tumor detection and characterization, often aided by the administration of gadolinium-based contrast agents (GBCAs) to improve tumor visualization. Recently, deep learning models have shown remarkable success in predicting contrast-enhancement in medical images, thereby reducing the need of GBCAs and potentially minimizing patient discomfort and risks. In this paper, we present a study aimed at investigating the generalization capabilities of a neural network trained to predict full contrast in brain tumor images from noncontrast MRI scans. While initial results exhibited promising performance on a specific tumor type at a certain stage using a specific dataset, our attempts to extend this success to other tumor types and diverse patient populations yielded unexpected challenges and limitations. Through a rigorous analysis of the factor contributing to these negative results, we aim to shed light on the complexities associated with generalizing contrast enhancement prediction in medical brain tumor imaging, offering valuable insights for future research and clinical applications.

Accurate internal target volume (ITV) estimation is essential for effective and safe radiation therapy in liver cancer. This study evaluates the clinical value of an ultraquality 4-dimensional magnetic resonance imaging (UQ 4D-MRI) technique for ITV estimation. The UQ 4D-MRI technique maps motion information from a low spatial resolution dynamic volumetric MRI onto a high-resolution 3-dimensional MRI used for radiation treatment planning. It was validated using a motion phantom and data from 13 patients with liver cancer. ITV generated from UQ 4D-MRI (ITV_4D) was compared with those obtained through isotropic expansions (ITV_{2 mm} and ITV_{5 mm}) and those measured using conventional 4D-computed tomography (computed tomography-based ITV, ITV_CT) for each patient. Phantom studies showed a displacement measurement difference of <5% between UQ 4D-MRI and single-slice 2-dimensional cine MRI. In patient studies, the maximum superior-inferior displacements of the tumor on UQ 4D-MRI showed no significant difference compared with single-slice 2-dimensional cine imaging (<i>P</i> = .985). Computed tomography-based ITV showed no significant difference (<i>P</i> = .72) with ITV_4D, whereas ITV_{2 mm} and ITV_{5 mm} significantly overestimated the volume by 29.0% (<i>P</i> = .002) and 120.7% (<i>P</i> < .001) compared with ITV_4D, respectively. UQ 4D-MRI enables accurate motion assessment for liver tumors, facilitating precise ITV delineation for radiation treatment planning. Despite uncertainties from artificial intelligence-based delineation and variations in patients' respiratory patterns, UQ 4D-MRI excels at capturing tumor motion trajectories, potentially improving treatment planning accuracy and reducing margins in liver cancer radiation therapy.

VSBONE^® BSI (VSBONE), an automated bone scan index (BSI) measurement system was updated from version 2.1 (ver.2) to 3.0 (ver.3). VSBONE ver.3 incorporates deep learning of the skeletal structures of 957 new women, and it can be applied in patients with breast cancer. However, the performance of the updated VSBONE remains unclear. This study aimed to validate the diagnostic accuracy of the VSBONE system in patients with breast cancer. In total, 220 Japanese patients with breast cancer who underwent bone scintigraphy with single-photon emission computed tomography/computed tomography (SPECT/CT) were retrospectively analyzed. The patients were diagnosed with active bone metastases (<i>n</i> = 20) and non-bone metastases (<i>n</i> = 200) according to the physician's radiographic image interpretation. The patients were assessed using the VSBONE ver.2 and VSBONE ver.3, and the BSI findings were compared with the interpretation results by the physicians. The occurrence of segmentation errors, the association of BSI between VSBONE ver.2 and VSBONE ver.3, and the diagnostic accuracy of the systems were evaluated. VSBONE ver.2 and VSBONE ver.3 had segmentation errors in four and two patients. Significant positive linear correlations were confirmed in both versions of the BSI (<i>r</i> = 0.92). The diagnostic accuracy was 54.1% in VSBOBE ver.2, and 80.5% in VSBONE ver.3 <i>(P</i> < 0.001), respectively. The diagnostic accuracy of VSBONE was improved through deep learning of the female skeletal structures. The updated VSBONE ver.3 can be a reliable automated system for measuring BSI in patients with breast cancer.

The article discusses advanced three-dimensional evaluation of hallux valgus deformity using weightbearing computed tomography. Conventional two-dimensional radiographs fall short in assessing the complexity of hallux valgus deformities, whereas weightbearing computed tomography provides detailed insights into bone alignment and joint stability in a weightbearing state. Recent studies have highlighted the significance of first ray hypermobility and intrinsic metatarsal rotation in hallux valgus, influencing surgical planning and outcomes. The integration of semiautomatic and artificial intelligence-assisted tools with weightbearing computed tomography is enhancing the precision of deformity assessment, leading to more personalized and effective hallux valgus management.