This study aimed to compare six large language models (LLMs) [Chat Generative Pre-trained Transformer (ChatGPT)o1-preview, ChatGPT-4o, ChatGPT-4o with canvas, Google Gemini 1.5 Pro, Claude 3.5 Sonnet, and Claude 3 Opus] in generating radiology references, assessing accuracy, fabrication, and bibliographic completeness. In this cross-sectional observational study, 120 open-ended questions were administered across eight radiology subspecialties (neuroradiology, abdominal, musculoskeletal, thoracic, pediatric, cardiac, head and neck, and interventional radiology), with 15 questions per subspecialty. Each question prompted the LLMs to provide responses containing four references with in-text citations and complete bibliographic details (authors, title, journal, publication year/month, volume, issue, page numbers, and PubMed Identifier). References were verified using Medline, Google Scholar, the Directory of Open Access Journals, and web searches. Each bibliographic element was scored for correctness, and a composite final score [(FS): 0-36] was calculated by summing the correct elements and multiplying this by a 5-point verification score for content relevance. The FS values were then categorized into a 5-point Likert scale reference accuracy score (RAS: 0 = fabricated; 4 = fully accurate). Non-parametric tests (Kruskal-Wallis, Tamhane's T2, Wilcoxon signed-rank test with Bonferroni correction) were used for statistical comparisons. Claude 3.5 Sonnet demonstrated the highest reference accuracy, with 80.8% fully accurate references (RAS 4) and a fabrication rate of 3.1%, significantly outperforming all other models (<i>P</i> < 0.001). Claude 3 Opus ranked second, achieving 59.6% fully accurate references and a fabrication rate of 18.3% (<i>P</i> < 0.001). ChatGPT-based models (ChatGPT-4o, ChatGPT-4o with canvas, and ChatGPT o1-preview) exhibited moderate accuracy, with fabrication rates ranging from 27.7% to 52.9% and <8% fully accurate references. Google Gemini 1.5 Pro had the lowest performance, achieving only 2.7% fully accurate references and the highest fabrication rate of 60.6% (<i>P</i> < 0.001). Reference accuracy also varied by subspecialty, with neuroradiology and cardiac radiology outperforming pediatric and head and neck radiology. Claude 3.5 Sonnet significantly outperformed all other models in generating verifiable radiology references, and Claude 3 Opus showed moderate performance. In contrast, ChatGPT models and Google Gemini 1.5 Pro delivered substantially lower accuracy with higher rates of fabricated references, highlighting current limitations in automated academic citation generation. The high accuracy of Claude 3.5 Sonnet can improve radiology literature reviews, research, and education with dependable references. The poor performance of other models, with high fabrication rates, risks misinformation in clinical and academic settings and highlights the need for refinement to ensure safe and effective use.

Automatic radiology report generation can enhance diagnostic efficiency and accuracy. However, clean open-source imaging scan-report pairs are limited in scale and variety. Moreover, the vast amount of radiological texts available online is often too noisy to be directly employed. To address this challenge, we introduce a novel task called Noisy Report Refinement (NRR), which generates radiology reports from noisy free-texts. To achieve this, we propose a report refinement pipeline that leverages large language models (LLMs) enhanced with guided self-critique and report selection strategies. To address the inability of existing radiology report generation metrics in measuring cleanliness, radiological usefulness, and factual correctness across various modalities of reports in NRR task, we introduce a new benchmark, NRRBench, for NRR evaluation. This benchmark includes two online-sourced datasets and four clinically explainable LLM-based metrics: two metrics evaluate the matching rate of radiology entities and modality-specific template attributes respectively, one metric assesses report cleanliness, and a combined metric evaluates overall NRR performance. Experiments demonstrate that guided self-critique and report selection strategies significantly improve the quality of refined reports. Additionally, our proposed metrics show a much higher correlation with noisy rate and error count of reports than radiology report generation metrics in evaluating NRR.

Predicting long-term functional outcomes for individuals with stroke is a significant challenge. Solving this challenge will open new opportunities for improving stroke management by informing acute interventions and guiding personalized rehabilitation strategies. The location of the stroke is a key predictor of outcomes, yet no clinically deployed tools incorporate lesion location information for outcome prognostication. This study responds to this critical need by introducing a fully automated, three-stage neuroimaging processing and machine learning pipeline that predicts personalized outcomes from clinical imaging in adult ischemic stroke patients. In the first stage, our system automatically processes raw DICOM inputs, registers the brain to a standard template, and uses deep learning models to segment the stroke lesion. In the second stage, lesion location and automatically derived network features are input into statistical models trained to predict long-term impairments from a large independent cohort of lesion patients. In the third stage, a structured PDF report is generated using a large language model that describes the strokes location, the arterial distribution, and personalized prognostic information. We demonstrate the viability of this approach in a proof-of-concept application predicting select cognitive outcomes in a stroke cohort. Brain-behavior models were pre-trained to predict chronic impairment on 28 different cognitive outcomes in a large cohort of patients with focal brain lesions (N=604). The automated pipeline used these models to predict outcomes from clinically acquired MRIs in an independent ischemic stroke cohort (N=153). Starting from raw clinical DICOM images, we show that our pipeline can generate outcome predictions for individual patients in less than 3 minutes with 96% concordance relative to methods requiring manual processing. We also show that prediction accuracy is enhanced using models that incorporate lesion location, lesion-associated network information, and demographics. Our results provide a strong proof-of-concept and lay the groundwork for developing imaging-based clinical tools for stroke outcome prognostication.

Deep learning has revolutionized medical image segmentation, yet its full potential remains constrained by the paucity of annotated datasets. While diffusion models have emerged as a promising approach for generating synthetic image-mask pairs to augment these datasets, they paradoxically suffer from the same data scarcity challenges they aim to mitigate. Traditional mask-only models frequently yield low-fidelity images due to their inability to adequately capture morphological intricacies, which can critically compromise the robustness and reliability of segmentation models. To alleviate this limitation, we introduce Siamese-Diffusion, a novel dual-component model comprising Mask-Diffusion and Image-Diffusion. During training, a Noise Consistency Loss is introduced between these components to enhance the morphological fidelity of Mask-Diffusion in the parameter space. During sampling, only Mask-Diffusion is used, ensuring diversity and scalability. Comprehensive experiments demonstrate the superiority of our method. Siamese-Diffusion boosts SANet's mDice and mIoU by 3.6% and 4.4% on the Polyps, while UNet improves by 1.52% and 1.64% on the ISIC2018. Code is available at GitHub.

The purpose of this study was to compare the ability of two multimodal models (GPT-4o and Gemini 1.5 Pro) with that of radiologists to generate differential diagnoses from textual context alone, key images alone, or a combination of both using complex neuroradiology cases. This retrospective study included neuroradiology cases from the "Diagnosis Please" series published in the Radiology journal between January 2008 and September 2024. The two multimodal models were asked to provide three differential diagnoses from textual context alone, key images alone, or the complete case. Six board-certified neuroradiologists solved the cases in the same setting, randomly assigned to two groups: context alone first and images alone first. Three radiologists solved the cases without, and then with the assistance of Gemini 1.5 Pro. An independent radiologist evaluated the quality of the image descriptions provided by GPT-4o and Gemini for each case. Differences in correct answers between multimodal models and radiologists were analyzed using McNemar test. GPT-4o and Gemini 1.5 Pro outperformed radiologists using clinical context alone (mean accuracy, 34.0 % [18/53] and 44.7 % [23.7/53] vs. 16.4 % [8.7/53]; both P < 0.01). Radiologists outperformed GPT-4o and Gemini 1.5 Pro using images alone (mean accuracy, 42.0 % [22.3/53] vs. 3.8 % [2/53], and 7.5 % [4/53]; both P < 0.01) and the complete cases (48.0 % [25.6/53] vs. 34.0 % [18/53], and 38.7 % [20.3/53]; both P < 0.001). While radiologists improved their accuracy when combining multimodal information (from 42.1 % [22.3/53] for images alone to 50.3 % [26.7/53] for complete cases; P < 0.01), GPT-4o and Gemini 1.5 Pro did not benefit from the multimodal context (from 34.0 % [18/53] for text alone to 35.2 % [18.7/53] for complete cases for GPT-4o; P = 0.48, and from 44.7 % [23.7/53] to 42.8 % [22.7/53] for Gemini 1.5 Pro; P = 0.54). Radiologists benefited significantly from the suggestion of Gemini 1.5 Pro, increasing their accuracy from 47.2 % [25/53] to 56.0 % [27/53] (P < 0.01). Both GPT-4o and Gemini 1.5 Pro correctly identified the imaging modality in 53/53 (100 %) and 51/53 (96.2 %) cases, respectively, but frequently failed to identify key imaging findings (43/53 cases [81.1 %] with incorrect identification of key imaging findings for GPT-4o and 50/53 [94.3 %] for Gemini 1.5). Radiologists show a specific ability to benefit from the integration of textual and visual information, whereas multimodal models mostly rely on the clinical context to suggest diagnoses.

 Fibrotic lung disease is a progressive illness that causes scarring and ultimately respiratory failure, with irreversible damage by the time it is diagnosed on computed tomography imaging. Recent research postulates the role of the lung vasculature on the pathogenesis of the disease. With the recent development of high-resolution hierarchical phase-contrast tomography (HiP-CT), we have the potential to understand and detect changes in the lungs long before conventional imaging. However, to gain quantitative insight into vascular changes you first need to be able to segment the vessels before further downstream analysis can be conducted. Aside from this, HiP-CT generates large-volume, high-resolution data which is time-consuming and expensive to label.  This project aims to qualitatively assess the latest machine learning methods for vessel segmentation in HiP-CT data to enable label propagation as the first step for imaging biomarker discovery, with the goal to identify early-stage interstitial lung disease amenable to treatment, before fibrosis begins.  Semisupervised learning (SSL) has become a growing method to tackle sparsely labeled datasets due to its leveraging of unlabeled data. In this study, we will compare two SSL methods; Seg PL, based on pseudo-labeling, and MisMatch, using consistency regularization against state-of-the-art supervised learning method, nnU-Net, on vessel segmentation in sparsely labeled lung HiP-CT data.  On initial experimentation, both MisMatch and SegPL showed promising performance on qualitative review. In comparison with supervised learning, both MisMatch and SegPL showed better out-of-distribution performance within the same sample (different vessel morphology and texture vessels), though supervised learning provided more consistent segmentations for well-represented labels in the limited annotations.  Further quantitative research is required to better assess the generalizability of these findings, though they show promising first steps toward leveraging this novel data to tackle fibrotic lung disease.

This study introduces APD-FFNet, a novel, explainable deep learning architecture for automated periodontitis diagnosis using panoramic radiographs. A total of 337 panoramic radiographs, annotated by a periodontist, served as the dataset. APD-FFNet combines custom convolutional and transformer-based layers within a deep feature fusion framework that captures both local and global contextual features. Performance was evaluated using accuracy, the F1 score, the area under the receiver operating characteristic curve, the Jaccard similarity coefficient, and the Matthews correlation coefficient. McNemar's test confirmed statistical significance, and SHapley Additive exPlanations provided interpretability insights. APD-FFNet achieved 94% accuracy, a 93.88% F1 score, 93.47% area under the receiver operating characteristic curve, 88.47% Jaccard similarity coefficient, and 88.46% Matthews correlation coefficient, surpassing comparable approaches. McNemar's test validated these findings (p < 0.05). Explanations generated by SHapley Additive exPlanations highlighted important regions in each radiograph, supporting clinical applicability. By merging convolutional and transformer-based layers, APD-FFNet establishes a new benchmark in automated, interpretable periodontitis diagnosis, with low hyperparameter sensitivity facilitating its integration into regular dental practice. Its adaptable design suggests broader relevance to other medical imaging domains. This is the first feature fusion method specifically devised for periodontitis diagnosis, supported by an expert-curated dataset and advanced explainable artificial intelligence. Its robust accuracy, low hyperparameter sensitivity, and transparent outputs set a new standard for automated periodontal analysis.

The diffusion model has recently emerged as a potent approach in computer vision, demonstrating remarkable performances in the field of generative artificial intelligence. Capable of producing high-quality synthetic images, diffusion models have been successfully applied across a range of applications. However, a significant challenge remains with the high computational cost associated with training and generating these models. This study focuses on the efficiency and inference time of diffusion-based generative models, highlighting their applications in both natural and medical imaging. We present the most recent advances in diffusion models by categorizing them into three key models: the Denoising Diffusion Probabilistic Model (DDPM), the Latent Diffusion Model (LDM), and the Wavelet Diffusion Model (WDM). These models play a crucial role in medical imaging, where producing fast, reliable, and high-quality medical images is essential for accurate analysis of abnormalities and disease diagnosis. We first investigate the general framework of DDPM, LDM, and WDM and discuss the computational complexity gap filled by these models in natural and medical imaging. We then discuss the current limitations of these models as well as the opportunities and future research directions in medical imaging.

The use of Artificial Intelligence (AI) methods for the analysis of CT (computed tomography) images has greatly contributed to the development of an effective computer-assisted diagnosis (CAD) system for lung cancer (LC). However, complex structures, multiple radiographic interrelations, and the dynamic locations of abnormalities within lung CT images make extracting relevant information to process and implement LC CAD systems difficult. These prominent problems are addressed in this paper by presenting a hybrid method of LC malignancy classification, which may help researchers and experts properly engineer the model's performance by observing how the model makes decisions. The proposed methodology is named IncCat-LCC: Explainer (Inception Net Cat Boost LC Classification: Explainer), which consists of feature extraction (FE) using the handcrafted radiomic Feature (HcRdF) extraction technique, InceptionNet CNN Feature (INCF) extraction, Vision Transformer Feature (ViTF) extraction, and XGBOOST (XGB)-based feature selection, and the GPU based CATBOOST (CB) classification technique. The proposed framework achieves better and highest performance scores for lung nodule multiclass malignancy classification when evaluated using metrics such as accuracy, precision, recall, f-1 score, specificity, and area under the roc curve as 96.74 %, 93.68 %, 96.74 %, 95.19 %, 98.47 % and 99.76 % consecutively for classifying highly normal class. Observing the explainable artificial intelligence (XAI) explanations will help readers understand the model performance and the statistical outcomes of the evaluation parameter. The work presented in this article may improve the existing LC CAD system and help assess the important parameters using XAI to recognize the factors contributing to enhanced performance and reliability.

The rapid evolution of large language models (LLMs) offers promising opportunities for radiology report annotation, aiding in determining the presence of specific findings. This study evaluates the effectiveness of a human-optimized prompt in labeling radiology reports across multiple institutions using LLMs. Six distinct institutions collected 500 radiology reports: 100 in each of 5 categories. A standardized Python script was distributed to participating sites, allowing the use of one common locally executed LLM with a standard human-optimized prompt. The script executed the LLM's analysis for each report and compared predictions to reference labels provided by local investigators. Models' performance using accuracy was calculated, and results were aggregated centrally. The human-optimized prompt demonstrated high consistency across sites and pathologies. Preliminary analysis indicates significant agreement between the LLM's outputs and investigator-provided reference across multiple institutions. At one site, eight LLMs were systematically compared, with Llama 3.1 70b achieving the highest performance in accurately identifying the specified findings. Comparable performance with Llama 3.1 70b was observed at two additional centers, demonstrating the model's robust adaptability to variations in report structures and institutional practices. Our findings illustrate the potential of optimized prompt engineering in leveraging LLMs for cross-institutional radiology report labeling. This approach is straightforward while maintaining high accuracy and adaptability. Future work will explore model robustness to diverse report structures and further refine prompts to improve generalizability.