To evaluate the efficacy of thin-slice T₂-weighted imaging (T₂WI) and super-resolution reconstruction (SRR) for preoperative assessment of vascular invasion in pancreatic ductal adenocarcinoma (PDAC). Ninety-five PDACs with preoperative MRI were retrospectively enrolled as a training set, with non-reconstructed T₂WI (NRT₂) in different slice thicknesses (NRT₂-3, 3 mm; NRT₂-5, ≥ 5 mm). A prospective test set was collected with NRT₂-5 (n = 125) only. A deep-learning network was employed to generate reconstructed super-resolution T₂WI (SRT₂) in different slice thicknesses (SRT₂-3, 3 mm; SRT₂-5, ≥ 5 mm). Image quality was assessed, including the signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and signal-intensity ratio (SIR_t/p, tumor/pancreas; SIR_t/b, tumor/background). Diagnostic efficacy for vascular invasion was evaluated using the area under the curve (AUC) and compared across different slice thicknesses before and after reconstruction. SRT₂-5 demonstrated higher SNR and SIR_t/p compared to NRT₂-5 (74.18 vs 72.46; 1.42 vs 1.30; p < 0.05). SRT₂-3 showed increased SIR_t/p and SIR_t/b over NRT₂-3 (1.35 vs 1.31; 2.73 vs 2.58; p < 0.05). SRT₂-5 showed higher CNR, SIR_t/p and SIR_t/b than NRT₂-3 (p < 0.05). NRT₂-3 outperformed NRT₂-5 in evaluating venous invasion (AUC: 0.732 vs 0.597, p = 0.021). SRR improved venous assessment (AUC: NRT₂-3, 0.927 vs 0.732; NRT₂-5, 0.823 vs 0.597; p < 0.05), and SRT₂-5 exhibits comparable efficacy to NRT₂-3 in venous assessment (AUC: 0.823 vs 0.732, p = 0.162). Thin-slice T₂WI and SRR effectively improve the image quality and diagnostic efficacy for assessing venous invasion in PDAC. Thick-slice T₂WI with SRR is a potential alternative to thin-slice T₂WI. Both thin-slice T₂-WI and SRR effectively improve image quality and diagnostic performance, providing valuable options for optimizing preoperative vascular assessment in PDAC. Non-invasive and accurate assessment of vascular invasion supports treatment planning and avoids futile surgery. Vascular invasion evaluation is critical for the surgical eligibility of PDAC. SRR improved image quality and vascular assessment in T₂WI. Utilizing thin-slice T₂WI and SRR aids in clinical decision making for PDAC.

Multimodal large language models (MLLMs) have enormous potential to perform few-shot in-context learning in the context of medical image analysis. However, safe deployment of these models into real-world clinical practice requires an in-depth analysis of the accuracies of their predictions, and their associated calibration errors, particularly across different demographic subgroups. In this work, we present the first investigation into the calibration biases and demographic unfairness of MLLMs' predictions and confidence scores in few-shot in-context learning for medical image classification. We introduce CALIN, an inference-time calibration method designed to mitigate the associated biases. Specifically, CALIN estimates the amount of calibration needed, represented by calibration matrices, using a bi-level procedure: progressing from the population level to the subgroup level prior to inference. It then applies this estimation to calibrate the predicted confidence scores during inference. Experimental results on three medical imaging datasets: PAPILA for fundus image classification, HAM10000 for skin cancer classification, and MIMIC-CXR for chest X-ray classification demonstrate CALIN's effectiveness at ensuring fair confidence calibration in its prediction, while improving its overall prediction accuracies and exhibiting minimum fairness-utility trade-off.

Federated learning (FL) has emerged as a promising paradigm for collaboratively training deep learning models across institutions without exchanging sensitive medical data. However, its effectiveness is often hindered by limited data availability and non-independent, identically distributed data across participating clients, which can degrade model performance and generalization. To address these challenges, we propose a generative AI based data augmentation framework that integrates synthetic image sharing into the federated training process for breast cancer diagnosis via ultrasound images. Specifically, we train two simple class-specific Deep Convolutional Generative Adversarial Networks: one for benign and one for malignant lesions. We then simulate a realistic FL setting using three publicly available breast ultrasound image datasets: BUSI, BUS-BRA, and UDIAT. FedAvg and FedProx are adopted as baseline FL algorithms. Experimental results show that incorporating a suitable number of synthetic images improved the average AUC from 0.9206 to 0.9237 for FedAvg and from 0.9429 to 0.9538 for FedProx. We also note that excessive use of synthetic data reduced performance, underscoring the importance of maintaining a balanced ratio of real and synthetic samples. Our findings highlight the potential of generative AI based data augmentation to enhance FL results in the breast ultrasound image classification task.

Multimodal large language models (MLLMs) have enormous potential to perform few-shot in-context learning in the context of medical image analysis. However, safe deployment of these models into real-world clinical practice requires an in-depth analysis of the accuracies of their predictions, and their associated calibration errors, particularly across different demographic subgroups. In this work, we present the first investigation into the calibration biases and demographic unfairness of MLLMs' predictions and confidence scores in few-shot in-context learning for medical image classification. We introduce CALIN, an inference-time calibration method designed to mitigate the associated biases. Specifically, CALIN estimates the amount of calibration needed, represented by calibration matrices, using a bi-level procedure: progressing from the population level to the subgroup level prior to inference. It then applies this estimation to calibrate the predicted confidence scores during inference. Experimental results on three medical imaging datasets: PAPILA for fundus image classification, HAM10000 for skin cancer classification, and MIMIC-CXR for chest X-ray classification demonstrate CALIN's effectiveness at ensuring fair confidence calibration in its prediction, while improving its overall prediction accuracies and exhibiting minimum fairness-utility trade-off.

Accurate carotid plaque grading (CPG) is vital to assess the risk of cardiovascular and cerebrovascular diseases. Due to the small size and high intra-class variability of plaque, CPG is commonly evaluated using a combination of transverse and longitudinal ultrasound views in clinical practice. However, most existing deep learning-based multi-view classification methods focus on feature fusion across different views, neglecting the importance of representation learning and the difference in class features. To address these issues, we propose a novel Corpus-View-Category Refinement Framework (CVC-RF) that processes information from Corpus-, View-, and Category-levels, enhancing model performance. Our contribution is four-fold. First, to the best of our knowledge, we are the foremost deep learning-based method for CPG according to the latest Carotid Plaque-RADS guidelines. Second, we propose a novel center-memory contrastive loss, which enhances the network's global modeling capability by comparing with representative cluster centers and diverse negative samples at the Corpus level. Third, we design a cascaded down-sampling attention module to fuse multi-scale information and achieve implicit feature interaction at the View level. Finally, a parameter-free mixture-of-experts weighting strategy is introduced to leverage class clustering knowledge to weight different experts, enabling feature decoupling at the Category level. Experimental results indicate that CVC-RF effectively models global features via multi-level refinement, achieving state-of-the-art performance in the challenging CPG task.

The recent release of RadGenome-Chest CT has significantly advanced CT-based report generation. However, existing methods primarily focus on global features, making it challenging to capture region-specific details, which may cause certain abnormalities to go unnoticed. To address this, we propose MedRegion-CT, a region-focused Multi-Modal Large Language Model (MLLM) framework, featuring three key innovations. First, we introduce Region Representative ($R^2$) Token Pooling, which utilizes a 2D-wise pretrained vision model to efficiently extract 3D CT features. This approach generates global tokens representing overall slice features and region tokens highlighting target areas, enabling the MLLM to process comprehensive information effectively. Second, a universal segmentation model generates pseudo-masks, which are then processed by a mask encoder to extract region-centric features. This allows the MLLM to focus on clinically relevant regions, using six predefined region masks. Third, we leverage segmentation results to extract patient-specific attributions, including organ size, diameter, and locations. These are converted into text prompts, enriching the MLLM's understanding of patient-specific contexts. To ensure rigorous evaluation, we conducted benchmark experiments on report generation using the RadGenome-Chest CT. MedRegion-CT achieved state-of-the-art performance, outperforming existing methods in natural language generation quality and clinical relevance while maintaining interpretability. The code for our framework is publicly available.

The connection between cognition, eye, and brain remains inconclusive in Alzheimer's disease (AD) spectrum disorders. To explore the relationship between cognitive function, retinal biometrics, and brain alterations in the AD spectrum. Prospective. Healthy control (HC) (n = 16), subjective cognitive decline (SCD) (n = 35), mild cognitive impairment (MCI) (n = 18), and AD group (n = 7). 3-T, 3D T1-weighted Brain Volume (BRAVO) and resting-state functional MRI (fMRI). In all subgroups, cortical thickness was measured from BRAVO and segmented using the Desikan-Killiany-Tourville (DKT) atlas. The fractional amplitude of low-frequency fluctuations (FALFF) and regional homogeneity (ReHo) were measured in fMRI using voxel-based analysis. The eye was imaged by optical coherence tomography angiography (OCTA), with the deep learning model FARGO segmenting the foveal avascular zone (FAZ) and retinal vessels. FAZ area and perimeter, retinal blood vessels curvature (RBVC), thicknesses of the retinal nerve fiber layer (RNFL) and ganglion cell layer-inner plexiform layer (GCL-IPL) were calculated. Cognition-eye-brain associations were compared across the HC group and each AD spectrum stage using multivariable linear regression. Multivariable linear regression analysis. Statistical significance was set at p < 0.05 with FWE correction for fMRI and p < 1/62 (Bonferroni-corrected) for structural analyses. Reductions of FALFF in temporal regions, especially the left superior temporal gyrus (STG) in MCI patients, were linked to decreased RNFL thickness and increased FAZ area significantly. In AD patients, reduced ReHo values in occipital regions, especially the right middle occipital gyrus (MOG), were significantly associated with an enlarged FAZ area. The SCD group showed widespread cortical thickening significantly associated with all aforementioned retinal biometrics, with notable thickening in the right fusiform gyrus (FG) and right parahippocampal gyrus (PHG) correlating with reduced GCL-IPL thickness. Brain function and structure may be associated with cognition and retinal biometrics across the AD spectrum. Specifically, cognition-eye-brain connections may be present in SCD. 2. 3.

We introduce MRF-DiPh, a novel physics informed denoising diffusion approach for multiparametric tissue mapping from highly accelerated, transient-state quantitative MRI acquisitions like Magnetic Resonance Fingerprinting (MRF). Our method is derived from a proximal splitting formulation, incorporating a pretrained denoising diffusion model as an effective image prior to regularize the MRF inverse problem. Further, during reconstruction it simultaneously enforces two key physical constraints: (1) k-space measurement consistency and (2) adherence to the Bloch response model. Numerical experiments on in-vivo brain scans data show that MRF-DiPh outperforms deep learning and compressed sensing MRF baselines, providing more accurate parameter maps while better preserving measurement fidelity and physical model consistency-critical for solving reliably inverse problems in medical imaging.

Collecting pixel-level labels for medical datasets can be a laborious and expensive process, and enhancing segmentation performance with a scarcity of labeled data is a crucial challenge. This work introduces AugPaint, a data augmentation framework that utilizes inpainting to generate image-label pairs from limited labeled data. AugPaint leverages latent diffusion models, known for their ability to generate high-quality in-domain images with low overhead, and adapts the sampling process for the inpainting task without need for retraining. Specifically, given a pair of image and label mask, we crop the area labeled with the foreground and condition on it during reversed denoising process for every noise level. Masked background area would gradually be filled in, and all generated images are paired with the label mask. This approach ensures the accuracy of match between synthetic images and label masks, setting it apart from existing dataset generation methods. The generated images serve as valuable supervision for training downstream segmentation models, effectively addressing the challenge of limited annotations. We conducted extensive evaluations of our data augmentation method on four public medical image segmentation datasets, including CT, MRI, and skin imaging. Results across all datasets demonstrate that AugPaint outperforms state-of-the-art label-efficient methodologies, significantly improving segmentation performance.

Accurate preoperative differentiation of benign and malignant thyroid nodules is critical for optimal patient management. However, conventional imaging modalities present inherent diagnostic limitations. To develop a non-contrast computed tomography-based machine learning model integrating radiomics and clinical features for preoperative thyroid nodule classification. This multicenter retrospective study enrolled 272 patients with thyroid nodules (376 thyroid lobes) from center A (May 2021-April 2024), using histopathological findings as the reference standard. The dataset was stratified into a training cohort (264 lobes) and an internal validation cohort (112 lobes). Additional prospective temporal (97 lobes, May-August 2024, center A) and external multicenter (81 lobes, center B) test cohorts were incorporated to enhance generalizability. Thyroid lobes were segmented along the isthmus midline, with segmentation reliability confirmed by an intraclass correlation coefficient (≥ 0.80). Radiomics feature extraction was performed using Pearson correlation analysis followed by least absolute shrinkage and selection operator regression with 10-fold cross-validation. Seven machine learning algorithms were systematically evaluated, with model performance quantified through the area under the receiver operating characteristic curve (AUC), Brier score, decision curve analysis, and DeLong test for comparison with radiologists interpretations. Model interpretability was elucidated using SHapley Additive exPlanations (SHAP). The extreme gradient boosting model demonstrated robust diagnostic performance across all datasets, achieving AUCs of 0.899 [95% confidence interval (CI): 0.845-0.932] in the training cohort, 0.803 (95%CI: 0.715-0.890) in internal validation, 0.855 (95%CI: 0.775-0.935) in temporal testing, and 0.802 (95%CI: 0.664-0.939) in external testing. These results were significantly superior to radiologists assessments (AUCs: 0.596, 0.529, 0.558, and 0.538, respectively; <i>P</i> < 0.001 by DeLong test). SHAP analysis identified radiomic score, age, tumor size stratification, calcification status, and cystic components as key predictive features. The model exhibited excellent calibration (Brier scores: 0.125-0.144) and provided significant clinical net benefit at decision thresholds exceeding 20%, as evidenced by decision curve analysis. The non-contrast computed tomography-based radiomics-clinical fusion model enables robust preoperative thyroid nodule classification, with SHAP-driven interpretability enhancing its clinical applicability for personalized decision-making.