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.

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.

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.

Osteosarcoma (OS) is the most common primary bone malignancy, and neoadjuvant chemotherapy (NAC) improves survival rates. However, OS heterogeneity results in variable treatment responses, highlighting the need for reliable, non-invasive tools to predict NAC response. Radiomics-based machine learning (ML) offers potential for identifying imaging biomarkers to predict treatment outcomes. This systematic review and meta-analysis evaluated the accuracy and reliability of radiomics models for predicting NAC response in OS. A systematic search was conducted in PubMed, Embase, Scopus, and Web of Science up to November 2024. Studies using radiomics-based ML for NAC response prediction in OS were included. Pooled sensitivity, specificity, and AUC for training and validation cohorts were calculated using bivariate random-effects modeling, with clinical-combined models analyzed separately. Quality assessment was performed using the QUADAS-2 tool, radiomics quality score (RQS), and METRICS scores. Sixteen studies were included, with 63 % using MRI and 37 % using CT. Twelve studies, comprising 1639 participants, were included in the meta-analysis. Pooled metrics for training cohorts showed an AUC of 0.93, sensitivity of 0.89, and specificity of 0.85. Validation cohorts achieved an AUC of 0.87, sensitivity of 0.81, and specificity of 0.82. Clinical-combined models outperformed radiomics-only models. The mean RQS score was 9.44 ± 3.41, and the mean METRICS score was 60.8 % ± 17.4 %. Radiomics-based ML shows promise for predicting NAC response in OS, especially when combined with clinical indicators. However, limitations in external validation and methodological consistency must be addressed.

Critical findings in radiology reports are life threatening conditions that need to be communicated promptly to physicians for timely management of patients. Although challenging, advancements in natural language processing (NLP), particularly large language models (LLMs), now enable the automated identification of key findings from verbose reports. Given the scarcity of labeled critical findings data, we implemented a two-phase, weakly supervised fine-tuning approach on 15,000 unlabeled Mayo Clinic reports. This fine-tuned model then automatically extracted critical terms on internal (Mayo Clinic, n = 80) and external (MIMIC-III, n = 123) test datasets, validated against expert annotations. Model performance was further assessed on 5000 MIMIC-IV reports using LLM-aided metrics, G-eval and Prometheus. Both manual and LLM-based evaluations showed improved task alignment with weak supervision. The pipeline and model, publicly available under an academic license, can aid in critical finding extraction for research and clinical use ( https://github.com/dasavisha/CriticalFindings_Extract ).

BackgroundIn glioblastoma (GBM), promoter methylation of the O6-methylguanine-DNA methyltransferase (MGMT) is associated with beneficial chemotherapy but has not been accurately evaluated based on radiological and pathological sections. To develop and validate an MRI and pathology image-based deep learning radiopathomics model for predicting MGMT promoter methylation in patients with GBM. MethodsA retrospective collection of pathologically confirmed isocitrate dehydrogenase (IDH) wild-type GBM patients (n=207) from three centers was performed, all of whom underwent MRI scanning within 2 weeks prior to surgery. The pre-trained ResNet50 was used as the feature extractor. Features of 1024 dimensions were extracted from MRI and pathological images, respectively, and the features were screened for modeling. Then feature fusion was performed by calculating the normalized multimode MRI fusion features and pathological features, and prediction models of MGMT based on deep learning radiomics, pathomics, and radiopathomics (DLRM, DLPM, DLRPM) were constructed and applied to internal and external validation cohorts. ResultsIn the training, internal and external validation cohorts, the DLRPM further improved the predictive performance, with a significantly better predictive performance than the DLRM and DLPM, with AUCs of 0.920 (95% CI 0.870-0.968), 0.854 (95% CI 0.702-1), and 0.840 (95% CI 0.625-1). ConclusionWe developed and validated cross-scale radiology and pathology models for predicting MGMT methylation status, with DLRPM predicting the best performance, and this cross-scale approach paves the way for further research and clinical applications in the future.

In abdominal interventional procedures, achieving precise registration of 2D ultrasound (US) frames with 3D computed tomography (CT) scans presents a significant challenge. Traditional tracking methods often rely on high-precision sensors, which can be prohibitively expensive. Furthermore, the clinical need for real-time registration with a broad capture range frequently exceeds the performance of standard image-based optimization techniques. Current automatic registration methods that utilize deep learning are either heavily reliant on manual annotations for training or struggle to effectively bridge the gap between different imaging domains. To address these challenges, we propose a novel diffusion-stimulated CT-US registration model. This model harnesses the physical diffusion properties of US to generate synthetic US images from preoperative CT data. Additionally, we introduce a synthetic-to-real domain adaptation strategy using a diffusion model to mitigate the discrepancies between real and synthetic US images. A dual-stream self-supervised regression neural network, trained on these synthetic images, is then used to estimate the pose within the CT space. The effectiveness of our proposed approach is verified through validation using US and CT scans from a dual-modality human abdominal phantom. The results of our experiments confirm that our method can accurately initialize the US image pose within an acceptable range of error and subsequently refine it to achieve precise alignment. This enables real-time, tracker-independent, and robust rigid registration of CT and US images.

Background and ObjectivesBottom-of-sulcus dysplasia (BOSD) is a diagnostically challenging subtype of focal cortical dysplasia, 60% being missed on patients first MRI. Automated MRI-based detection methods have been developed for focal cortical dysplasia, but not BOSD specifically. Use of FDG-PET alongside MRI is not established in automated methods. We report the development and performance of an automated BOSD detector using combined MRI+PET data. MethodsThe training set comprised 54 mostly operated patients with BOSD. The test sets comprised 17 subsequently diagnosed patients with BOSD from the same center, and 12 published patients from a different center. 81% patients across training and test sets had reportedly normal first MRIs and most BOSDs were <1.5cm3. In the training set, 12 features from T1-MRI, FLAIR-MRI and FDG-PET were evaluated using a novel "pseudo-control" normalization approach to determine which features best distinguished dysplastic from normal-appearing cortex. Using the Multi-centre Epilepsy Lesion Detection groups machine-learning detection method with the addition of FDG-PET, neural network classifiers were then trained and tested on MRI+PET features, MRI-only and PET-only. The proportion of patients whose BOSD was overlapped by the top output cluster, and the top five output clusters, were assessed. ResultsCortical and subcortical hypometabolism on FDG-PET were superior in discriminating dysplastic from normal-appearing cortex compared to MRI features. When the BOSD detector was trained on MRI+PET features, 87% BOSDs were overlapped by one of the top five clusters (69% top cluster) in the training set, 76% in the prospective test set (71% top cluster) and 75% in the published test set (42% top cluster). Cluster overlap was similar when the detector was trained and tested on PET-only features but lower when trained and tested on MRI-only features. ConclusionDetection of BOSD is possible using established MRI-based automated detection methods, supplemented with FDG-PET features and trained on a BOSD-specific cohort. In clinical practice, an MRI+PET BOSD detector could improve assessment and outcomes in seemingly MRI-negative patients being considered for epilepsy surgery.

This study proposes a novel approach to predict the efficacy of bevacizumab (BEV) in treating peritumoral edema in metastatic brain tumor patients by integrating advanced machine learning (ML) techniques with comprehensive imaging and clinical data. A retrospective analysis was performed on 300 patients who received BEV treatment from September 2013 to January 2024. The dataset incorporated 13 predictive features: 8 clinical variables and 5 radiological variables. The dataset was divided into a training set (70%) and a test set (30%) using stratified sampling. Data preprocessing was carried out through methods such as handling missing values with the MICE method, detecting and adjusting outliers, and feature scaling. Four algorithms, namely Random Forest (RF), Logistic Regression, Gradient Boosting Tree, and Naive Bayes, were selected to construct binary classification models. A tenfold cross-validation strategy was implemented during training, and techniques like regularization, hyperparameter optimization, and oversampling were used to mitigate overfitting. The RF model demonstrated superior performance, achieving an accuracy of 0.89, a precision of 0.94, F1-score of 0.92, with both AUC-ROC and AUC-PR values reaching 0.91. Feature importance analysis consistently identified edema volume as the most significant predictor, followed by edema index, patient age, and tumor volume. Traditional multivariate logistic regression corroborated these findings, confirming that edema volume and edema index were independent predictors (p < 0.01). Our results highlight the potential of ML-driven predictive models in optimizing BEV treatment selection, reducing unnecessary treatment risks, and improving clinical decision-making in neuro-oncology.

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.