This study evaluates the performance of ChatGPT-4, a Large Language Model (LLM), in automatically extracting U scores from free-text thyroid ultrasound reports collected from University Hospitals Birmingham (UHB), UK, between 2014 and 2024. The LLM was provided with guidelines on the U classification system and extracted U scores independently from 14,248 de-identified reports, without access to human-assigned scores. The LLM-extracted scores were compared to initial clinician-assigned and refined U scores provided by expert reviewers. The LLM achieved 97.7% agreement with refined human U scores, successfully identifying the highest U score in 98.1% of reports with multiple nodules. Most discrepancies (2.5%) were linked to ambiguous descriptions, multi-nodule reports, and cases with human-documented uncertainty. While the results demonstrate the potential for LLMs to improve reporting consistency and reduce manual workload, ethical and governance challenges such as transparency, privacy, and bias must be addressed before routine clinical deployment. Embedding LLMs into reporting workflows, such as Online Analytical Processing (OLAP) tools, could further enhance reporting quality and consistency.

Medical imaging datasets often contain heterogeneous biases ranging from erroneous labels to inconsistent labeling styles. Such biases can negatively impact deep segmentation networks performance. Yet, the identification and characterization of such biases is a particularly tedious and challenging task. In this paper, we introduce HyperSORT, a framework using a hyper-network predicting UNets' parameters from latent vectors representing both the image and annotation variability. The hyper-network parameters and the latent vector collection corresponding to each data sample from the training set are jointly learned. Hence, instead of optimizing a single neural network to fit a dataset, HyperSORT learns a complex distribution of UNet parameters where low density areas can capture noise-specific patterns while larger modes robustly segment organs in differentiated but meaningful manners. We validate our method on two 3D abdominal CT public datasets: first a synthetically perturbed version of the AMOS dataset, and TotalSegmentator, a large scale dataset containing real unknown biases and errors. Our experiments show that HyperSORT creates a structured mapping of the dataset allowing the identification of relevant systematic biases and erroneous samples. Latent space clusters yield UNet parameters performing the segmentation task in accordance with the underlying learned systematic bias. The code and our analysis of the TotalSegmentator dataset are made available: https://github.com/ImFusionGmbH/HyperSORT

Deep-learning models for prostate cancer detection typically require large datasets, limiting clinical applicability across institutions due to domain shift issues. This study aimed to develop a few-shot learning deep-learning model for prostate cancer detection on multiparametric MRI that requires minimal training data and to compare its diagnostic performance with experienced radiologists. In this retrospective study, we used 99 cases (80 positive, 19 negative) of biopsy-confirmed prostate cancer (2017-2022), with 20 cases for training, 5 for validation, and 74 for testing. A 2D transformer model was trained on T2-weighted, diffusion-weighted, and apparent diffusion coefficient map images. Model predictions were compared with two radiologists using Matthews correlation coefficient (MCC) and F1 score, with 95% confidence intervals (CIs) calculated via bootstrap method. The model achieved an MCC of 0.297 (95% CI: 0.095-0.474) and F1 score of 0.707 (95% CI: 0.598-0.847). Radiologist 1 had an MCC of 0.276 (95% CI: 0.054-0.484) and F1 score of 0.741; Radiologist 2 had an MCC of 0.504 (95% CI: 0.289-0.703) and F1 score of 0.871, showing that the model performance was comparable to Radiologist 1. External validation on the Prostate158 dataset revealed that ImageNet pretraining substantially improved model performance, increasing study-level ROC-AUC from 0.464 to 0.636 and study-level PR-AUC from 0.637 to 0.773 across all architectures. Our findings demonstrate that few-shot deep-learning models can achieve clinically relevant performance when using pretrained transformer architectures, offering a promising approach to address domain shift challenges across institutions.

Hepatic echinococcosis (HE) is a widespread parasitic disease in underdeveloped pastoral areas with limited medical resources. While CNN-based and Transformer-based models have been widely applied to medical image segmentation, CNNs lack global context modeling due to local receptive fields, and Transformers, though capable of capturing long-range dependencies, are computationally expensive. Recently, state space models (SSMs), such as Mamba, have gained attention for their ability to model long sequences with linear complexity. In this paper, we propose EAGLE, a U-shaped network composed of a Progressive Visual State Space (PVSS) encoder and a Hybrid Visual State Space (HVSS) decoder that work collaboratively to achieve efficient and accurate segmentation of hepatic echinococcosis (HE) lesions. The proposed Convolutional Vision State Space Block (CVSSB) module is designed to fuse local and global features, while the Haar Wavelet Transformation Block (HWTB) module compresses spatial information into the channel dimension to enable lossless downsampling. Due to the lack of publicly available HE datasets, we collected CT slices from 260 patients at a local hospital. Experimental results show that EAGLE achieves state-of-the-art performance with a Dice Similarity Coefficient (DSC) of 89.76%, surpassing MSVM-UNet by 1.61%.

To address the high computational complexity of the Transformer in the segmentation of ultrasound thyroid nodules and the loss of image details or omission of key spatial information caused by traditional image sampling techniques when dealing with high-resolution, complex texture or uneven density two-dimensional ultrasound images, this paper proposes a thyroid nodule segmentation method that integrates the receiving weighted key-value (RWKV) architecture and spherical geometry feature (SGF) sampling technology. This method effectively captures the details of adjacent regions through two-dimensional offset prediction and pixel-level sampling position adjustment, achieving precise segmentation. Additionally, this study introduces a patch attention module (PAM) to optimize the decoder feature map using a regional cross-attention mechanism, enabling it to focus more precisely on the high-resolution features of the encoder. Experiments on the thyroid nodule segmentation dataset (TN3K) and the digital database for thyroid images (DDTI) show that the proposed method achieves dice similarity coefficients (DSC) of 87.24% and 80.79% respectively, outperforming existing models while maintaining a lower computational complexity. This approach may provide an efficient solution for the precise segmentation of thyroid nodules.

The objective of this study is to investigate the impact of deep learning reconstruction and accelerated acquisition on reproducibility and variability of radiomic features in abdominal MRI. Seventeen volunteers were prospectively included to undergo abdominal MRI on a 3-T scanner for axial T2-weighted, axial T2-weighted fat-suppressed, and coronal T2-weighted sequences. Each sequence was scanned for four times using clinical reference acquisition with standard reconstruction, clinical reference acquisition with deep learning reconstruction, accelerated acquisition with standard reconstruction, and accelerated acquisition with deep learning reconstruction, respectively. The regions of interest were drawn for ten anatomical sites with rigid registrations. Ninety-three radiomic features were extracted via PyRadiomics after z-score normalization. The reproducibility was evaluated using clinical reference acquisition with standard reconstruction as reference by intraclass correlation coefficient (ICC) and concordance correlation coefficient (CCC). The variability among four scans was assessed by coefficient of variation (CV) and quartile coefficient of dispersion (QCD). Our study found that the median (first and third quartile) of overall ICC and CCC values were 0.451 (0.305, 0.583) and 0.450 (0.304, 0.582). The overall percentage of radiomic features with ICC > 0.90 and CCC > 0.90 was 8.1% and 8.1%, and was considered acceptable. The median (first and third quartile) of overall CV and QCD values was 9.4% (4.9%, 17.2%) and 4.9% (2.5%, 9.7%). The overall percentage of radiomic features with CV < 10% and QCD < 10% was 51.9% and 75.0%, and was considered acceptable. Without respect to clinical significance, deep learning reconstruction and accelerated acquisition led to a poor reproducibility of radiomic features, but more than a half of the radiomic features varied within an acceptable range.

Sarcopenia, defined as the progressive loss of skeletal muscle mass and function, has been associated with poor prognosis in patients with pancreatic cancer, particularly those with borderline resectable pancreatic cancer (BRPC). Although body composition can be extracted from routine CT imaging, sarcopenia assessment remains underused in clinical practice. Recent advances in artificial intelligence (AI) offer the potential to automate and standardize this process, but their clinical translation remains limited. This narrative review aims to critically evaluate (1) the clinical impact of CT-defined sarcopenia in BRPC, and (2) the performance and maturity of AI-based methods for automated muscle and fat segmentation on CT images. A dual-axis literature search was conducted to identify clinical studies assessing the prognostic role of sarcopenia in BRPC, and technical studies developing AI-based segmentation models for body composition analysis. Structured data extraction was applied to 13 clinical and 71 technical studies. A PRISMA-inspired flow diagram was included to ensure methodological transparency. Sarcopenia was consistently associated with worse survival and treatment tolerance in BRPC, yet clinical definitions and cut-offs varied widely. AI models-mostly 2D U-Nets trained on L3-level CT slices-achieved high segmentation accuracy (mean DSC >0.93), but external validation and standardization were often lacking. CT-based AI assessment of sarcopenia holds promise for improving patient stratification in BRPC. However, its clinical adoption will require standardization, integration into decision-support frameworks, and prospective validation across diverse populations.

Differentiating intrahepatic cholangiocarcinoma (ICC) from hepatocellular carcinoma (HCC) is essential for selecting the most effective treatment strategies. However, traditional imaging modalities and serum biomarkers often lack sufficient specificity. Radiomics, a sophisticated image analysis approach that derives quantitative data from medical imaging, has emerged as a promising non-invasive tool. To systematically review and meta-analyze the radiomics diagnostic accuracy in differentiating ICC from HCC. PubMed, EMBASE, and Web of Science databases were systematically searched through January 24, 2025. Studies evaluating radiomics models for distinguishing ICC from HCC were included. Assessing the quality of included studies was done by using Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) and METhodological RadiomICs Score tools. Pooled sensitivity, specificity, and area under the curve (AUC) were calculated using a bivariate random-effects model. Subgroup and publication bias analyses were also performed. 12 studies with 2541 patients were included, with 14 validation cohorts entered into meta-analysis. The pooled sensitivity and specificity of radiomics models were 0.82 (95% CI: 0.76-0.86) and 0.90 (95% CI: 0.85-0.93), respectively, with an AUC of 0.88 (95% CI: 0.85-0.91). Subgroup analyses revealed variations based on segmentation method, software used, and sample size, though not all differences were statistically significant. Publication bias was not detected. Radiomics demonstrates high diagnostic accuracy in distinguishing ICC from HCC and offers a non-invasive adjunct to conventional diagnostics. Further prospective, multicenter studies with standardized workflows are needed to enhance clinical applicability and reproducibility.

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.

This study aimed to develop a deep learning (DL) model based on three-dimensional multi-parametric magnetic resonance imaging (mpMRI) for preoperative assessment of lymph node metastasis (LNM) in rectal cancer (RC) and to investigate the contribution of different MRI sequences. A total of 613 eligible patients with RC from four medical centres who underwent preoperative mpMRI were retrospectively enrolled and randomly assigned to training (n = 372), validation (n = 106), internal test (n = 88) and external test (n = 47) cohorts. A multi-parametric multi-scale EfficientNet (MMENet) was designed to effectively extract LNM-related features from mpMR for preoperative LNM assessment. Its performance was compared with other DL models and radiologists using metrics of area under the receiver operating curve (AUC), accuracy (ACC), sensitivity, specificity and average precision with 95% confidence interval (CI). To investigate the utility of various MRI sequences, the performances of the mono-parametric model and the MMENet with different sequences combinations as input were compared. The MMENet using a combination of T2WI, DWI and DCE sequence achieved an AUC of 0.808 (95% CI 0.720-0.897) with an ACC of 71.6% (95% CI 62.3-81.0) in the internal test cohort and an AUC of 0.782 (95% CI 0.636-0.925) with an ACC of 76.6% (95% CI 64.6-88.6) in the external test cohort, outperforming the mono-parametric model, the MMENet with other sequences combinations and the radiologists. The MMENet, leveraging a combination of T2WI, DWI and DCE sequences, can accurately assess LNM in RC preoperatively and holds great promise for automated evaluation of LNM in clinical practice.