Adenocarcinoma (AD) and squamous cell carcinoma (SCC) are frequently observed forms of non-small cell lung cancer (NSCLC), playing a significant role in global cancer mortality. This research categorizes NSCLC subtypes by analyzing image details using computer-assisted semi-automatic segmentation and radiomic features in model development. This study includes 80 patients with 50 AD and 30 SCC which were analyzed using 3D Slicer software and extracted 107 quantitative radiomic features per patient. After eliminating correlated attributes, LASSO binary logistic regression model and 10-fold cross-validation were used for feature selection. The Shapiro-Wilk test assessed radiomic score normality, and the Mann-Whitney U test compared score distributions. Random Forest (RF) and Support Vector Machine (SVM) classification models were implemented for subtype classification. Receiver-Operator Characteristic (ROC) curves evaluated the radiomics score, showing a moderate predictive ability with training set area under curve (AUC) of 0.679 (95 % CI, 0.541-0.871) and validation set AUC of 0.560 (95 % CI, 0.342-0.778). Rad-Score distributions were normal for AD and not normal for SCC. RF and SVM classification models, which are based on selected features, resulted RF accuracy (95 % CI) of 0.73 and SVM accuracy (95 % CI) of 0.87, with respective AUC values of 0.54 and 0.87. These findings enhance the understanding that the two subtypes of NSCLC can be differentiated. The study demonstrated radiomic analysis improves diagnostic accuracy and offers a non-invasive alternative. However, the AUCs and ROC curves for the machine learning models must be critically evaluated to ensure clinical acceptability. If robust, these models could reduce the need for biopsies and enhance personalized treatment planning. Further research is needed to validate these findings and integrate radiomics into NSCLC clinical practice.

With access to large-scale, unlabeled medical datasets, researchers are confronted with two questions: Should they attempt to pretrain a custom foundation model on this medical data, or use transfer-learning from an existing generalist model? And, if a custom model is pretrained, are novel methods required? In this paper we explore these questions by conducting a case-study, in which we train a foundation model on a large regional fetal ultrasound dataset of 2M images. By selecting the well-established DINOv2 method for pretraining, we achieve state-of-the-art results on three fetal ultrasound datasets, covering data from different countries, classification, segmentation, and few-shot tasks. We compare against a series of models pretrained on natural images, ultrasound images, and supervised baselines. Our results demonstrate two key insights: (i) Pretraining on custom data is worth it, even if smaller models are trained on less data, as scaling in natural image pretraining does not translate to ultrasound performance. (ii) Well-tuned methods from computer vision are making it feasible to train custom foundation models for a given medical domain, requiring no hyperparameter tuning and little methodological adaptation. Given these findings, we argue that a bias towards methodological innovation should be avoided when developing domain specific foundation models under common computational resource constraints.

This study evaluates the feasibility of a novel deep learning-accelerated half-fourier single-shot turbo spin-echo sequence (HASTE-DL) compared to the conventional HASTE sequence (HASTE_S) in postoperative single-sequence MRI for the detection of fluid collections following abdominal surgery. As small fluid collections are difficult to visualize using other techniques, HASTE-DL may offer particular advantages in this clinical context. A retrospective analysis was conducted on 76 patients (mean age 65±11.69 years) who underwent abdominal MRI for suspected septic foci following abdominal surgery. Imaging was performed using 3-T MRI scanners, and both sequences were analyzed in terms of image quality, contrast, sharpness, and artifact presence. Quantitative assessments focused on fluid collection detectability, while qualitative assessments evaluated visualization of critical structures. Inter-reader agreement was measured using Cohen's kappa coefficient, and statistical significance was determined with the Mann-Whitney U test. HASTE-DL achieved a 46% reduction in scan time compared to HASTE_S, while significantly improving overall image quality (p<0.001), contrast (p<0.001), and sharpness (p<0.001). The inter-reader agreement for HASTE-DL was excellent (κ=0.960), with perfect agreement on overall image quality and fluid collection detection (κ=1.0). Fluid detectability and characterization scores were higher for HASTE-DL, and visualization of critical structures was significantly enhanced (p<0.001). No relevant artifacts were observed in either sequence. HASTE-DL offers superior image quality, improved visualization of critical structures, such as drainages, vessels, bile and pancreatic ducts, and reduced acquisition time, making it an effective alternative to the standard HASTE sequence, and a promising complementary tool in the postoperative imaging workflow.

This study compared the capabilities of two-dimensional (2D) and three-dimensional (3D) deep learning (DL), radiomics, and fusion models to predict postpartum hemorrhage (PPH), using sagittal T2-weighted MRI images. This retrospective study successively included 581 pregnant women suspected of placenta accreta spectrum (PAS) disorders who underwent placental MRI assessment between May 2018 and June 2024 in two hospitals. Clinical information was collected, and MRI images were analyzed by two experienced radiologists. The study cohort was divided into training (hospital 1, n=470) and validation (hospital 2, n=160) sets. Radiomics features were extracted after image segmentation to develop the radiomics model, 2D and 3D DL models were developed, and two fusion strategies (early and late fusion) were used to construct the fusion models. ROC curves, AUC, sensitivity, specificity, calibration curves, and decision curve analysis were used to evaluate the models' performance. The late-fusion model (DLRad_LF) yielded the highest performance, with AUCs of 0.955 (95% CI: 0.935-0.974) and 0.898 (95% CI: 0.848-0.949) in the training and validation sets, respectively. In the validation set, the AUC of the 3D DL model was significantly larger than those of the radiomics (AUC=0.676, P<0.001) and 2D DL (AUC=0.752, P<0.001) models. Subgroup analysis found that placenta previa and PAS did not impact the models' performance significantly. The DLRad_LF model could predict PPH reasonably accurately based on sagittal T2-weighted MRI images.

Generating reports for computed tomography (CT) images is a challenging task, while similar to existing studies for medical image report generation, yet has its unique characteristics, such as spatial encoding of multiple images, alignment between image volume and texts, etc. Existing solutions typically use general 2D or 3D image processing techniques to extract features from a CT volume, where they firstly compress the volume and then divide the compressed CT slices into patches for visual encoding. These approaches do not explicitly account for the transformations among CT slices, nor do they effectively integrate multi-level image features, particularly those containing specific organ lesions, to instruct CT report generation (CTRG). In considering the strong correlation among consecutive slices in CT scans, in this paper, we propose a large language model (LLM) based CTRG method with recurrent visual feature extraction and stereo attentions for hierarchical feature modeling. Specifically, we use a vision Transformer to recurrently process each slice in a CT volume, and employ a set of attentions over the encoded slices from different perspectives to selectively obtain important visual information and align them with textual features, so as to better instruct an LLM for CTRG. Experiment results and further analysis on the benchmark M3D-Cap dataset show that our method outperforms strong baseline models and achieves state-of-the-art results, demonstrating its validity and effectiveness.

Computed Tomography (CT) plays a crucial role in clinical diagnosis, but the growing demand for CT examinations has raised concerns about diagnostic errors. While Multimodal Large Language Models (MLLMs) demonstrate promising comprehension of medical knowledge, their tendency to produce inaccurate information highlights the need for rigorous validation. However, existing medical visual question answering (VQA) benchmarks primarily focus on simple visual recognition tasks, lacking clinical relevance and failing to assess expert-level knowledge. We introduce MedErr-CT, a novel benchmark for evaluating medical MLLMs' ability to identify and correct errors in CT reports through a VQA framework. The benchmark includes six error categories - four vision-centric errors (Omission, Insertion, Direction, Size) and two lexical error types (Unit, Typo) - and is organized into three task levels: classification, detection, and correction. Using this benchmark, we quantitatively assess the performance of state-of-the-art 3D medical MLLMs, revealing substantial variation in their capabilities across different error types. Our benchmark contributes to the development of more reliable and clinically applicable MLLMs, ultimately helping reduce diagnostic errors and improve accuracy in clinical practice. The code and datasets are available at https://github.com/babbu3682/MedErr-CT.

Convolutional denoising autoencoders (DAEs) are powerful tools for image restoration. However, they inherit a key limitation of convolutional neural networks (CNNs): they tend to recover low-frequency features, such as smooth regions, more effectively than high-frequency details. This leads to the loss of fine details, which is particularly problematic in dental radiographs where preserving subtle anatomical structures is crucial. While self-attention mechanisms can help mitigate this issue by emphasizing important features, conventional attention methods often prioritize features corresponding to cleaner regions and may overlook those obscured by noise. To address this limitation, we propose a noise-aware self-attention method, which allows the model to effectively focus on and recover key features even within noisy regions. Building on this approach, we introduce the noise-aware attention-enhanced denoising autoencoder (NAADA) network for enhancing noisy panoramic dental radiographs. Compared with the recent state of the art (and much heavier) methods like Uformer, MResDNN etc., our method improves the reconstruction of fine details, ensuring better image quality and diagnostic accuracy.

Although new vision foundation models such as Segment Anything Model 2 (SAM2) have significantly enhanced zero-shot image segmentation capabilities, reliance on human-provided prompts poses significant challenges in adapting SAM2 to medical image segmentation tasks. Moreover, SAM2's performance in medical image segmentation was limited by the domain shift issue, since it was originally trained on natural images and videos. To address these challenges, we proposed SAM2 with support-set guided prompting (SAM2-SGP), a framework that eliminated the need for manual prompts. The proposed model leveraged the memory mechanism of SAM2 to generate pseudo-masks using image-mask pairs from a support set via a Pseudo-mask Generation (PMG) module. We further introduced a novel Pseudo-mask Attention (PMA) module, which used these pseudo-masks to automatically generate bounding boxes and enhance localized feature extraction by guiding attention to relevant areas. Furthermore, a low-rank adaptation (LoRA) strategy was adopted to mitigate the domain shift issue. The proposed framework was evaluated on both 2D and 3D datasets across multiple medical imaging modalities, including fundus photography, X-ray, computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET), and ultrasound. The results demonstrated a significant performance improvement over state-of-the-art models, such as nnUNet and SwinUNet, as well as foundation models, such as SAM2 and MedSAM2, underscoring the effectiveness of the proposed approach. Our code is publicly available at https://github.com/astlian9/SAM_Support.

Panoramic X-rays (PX) are extensively utilized in dental and maxillofacial diagnostics, offering comprehensive imaging of teeth and surrounding structures. This study investigates the automatic determination of Kennedy's classification in partially edentulous jaws. A retrospective study involving 209 PX images from 206 patients was conducted. The established Mask R-CNN, a deep learning-based instance segmentation model, was trained for the automatic detection, position labeling (according to the international dental federation's scheme), and segmentation of teeth in PX. Subsequent post-processing steps filter duplicate outputs by position label and by geometric overlap. Finally, a rule-based determination of Kennedy's class of partially edentulous jaws was performed. In a fivefold cross-validation, Kennedy's classification was correctly determined in 83.0% of cases, with the most common errors arising from the mislabeling of morphologically similar teeth. The underlying algorithm demonstrated high sensitivity (97.1%) and precision (98.1%) in tooth detection, with an F1 score of 97.6%. FDI position label accuracy was 94.7%. Ablation studies indicated that post-processing steps, such as duplicate filtering, significantly improved algorithm performance. Our findings show that automatic dentition analysis in PX images can be extended to include clinically relevant jaw classification, reducing the workload associated with manual labeling and classification.

To evaluate PANCANAI, a previously developed AI model for pancreatic cancer (PC) detection, on a longitudinal cohort of patients. In particular, aiming for PC detection on scans acquired before histopathologic diagnosis was assessed. The model has been previously trained to predict PC suspicion on 2134 portal venous CTs. In this study, the algorithm was evaluated on a retrospective cohort of Danish patients with biopsy-confirmed PC and with CT scans acquired between 2006 and 2016. The sensitivity was measured, and bootstrapping was performed to provide median and 95% CI. The study included 1083 PC patients (mean age: 69 y ± 11, 575 men). CT scans were divided into 2 groups: (1) concurrent diagnosis (CD): 1022 CT scans acquired within 2 months around histopathologic diagnosis, and (2) prediagnosis (PD): 198 CT scans acquired before histopathologic diagnosis (median 7 months before diagnosis). The sensitivity was 91.8% (938 of 1022; 95% CI: 89.9-93.5) and 68.7% (137 of 198; 95% CI: 62.1-75.3) on the CD and PD groups, respectively. Sensitivity on CT scans acquired 1 year or more before diagnosis was 53.9% (36 of 67; 95% CI: 41.8-65.7). Sensitivity on CT scans acquired at stage I was 82.9% (29 of 35; 95% CI: 68.6-94.3). PANCANAI showed high sensitivity for automatic PC detection on a large retrospective cohort of biopsy-confirmed patients. PC suspicion was detected in more than half of the CT scans that were acquired at least a year before histopathologic diagnosis.