The reliance on computationally intensive U-Net and Transformer architectures significantly limits their accessibility in low-resource environments, creating a technological divide that hinders global healthcare equity, especially in medical diagnostics and treatment planning. This divide is most pronounced in low- and middle-income countries, primary care facilities, and conflict zones. We introduced MED-NCA, Neural Cellular Automata (NCA) based segmentation models characterized by their low parameter count, robust performance, and inherent quality control mechanisms. These features drastically lower the barriers to high-quality medical image analysis in resource-constrained settings, allowing the models to run efficiently on hardware as minimal as a Raspberry Pi or a smartphone. Building upon the foundation laid by MED-NCA, this paper extends its validation across eight distinct anatomies, including the hippocampus and prostate (MRI, 3D), liver and spleen (CT, 3D), heart and lung (X-ray, 2D), breast tumor (Ultrasound, 2D), and skin lesion (Image, 2D). Our comprehensive evaluation demonstrates the broad applicability and effectiveness of MED-NCA in various medical imaging contexts, matching the performance of two magnitudes larger UNet models. Additionally, we introduce NCA-VIS, a visualization tool that gives insight into the inference process of MED-NCA and allows users to test its robustness by applying various artifacts. This combination of efficiency, broad applicability, and enhanced interpretability makes MED-NCA a transformative solution for medical image analysis, fostering greater global healthcare equity by making advanced diagnostics accessible in even the most resource-limited environments.

Optimize deep learning-based vertebrae segmentation in longitudinal CT scans of multiple myeloma patients using structural uncertainty analysis. Retrospective CT scans from 474 multiple myeloma patients were divided into train (179 patients, 349 scans, 2005-2011) and test cohort (295 patients, 671 scans, 2012-2020). An enhanced segmentation pipeline was developed on the train cohort. It integrated vertebrae segmentation using an open-source deep learning method (Payer's) with a post-hoc structural uncertainty analysis. This analysis identified inconsistencies, automatically correcting them or flagging uncertain regions for human review. Segmentation quality was assessed through vertebral shape analysis using topology. Metrics included 'identification rate', 'longitudinal vertebral match rate', 'success rate' and 'series success rate' and evaluated across age/sex subgroups. Statistical analysis included McNemar and Wilcoxon signed-rank tests, with p < 0.05 indicating significant improvement. Payer's method achieved an identification rate of 95.8% and success rate of 86.7%. The proposed pipeline automatically improved these metrics to 98.8% and 96.0%, respectively (p < 0.001). Additionally, 3.6% of scans were marked for human inspection, increasing the success rate from 96.0% to 98.8% (p < 0.001). The vertebral match rate increased from 97.0% to 99.7% (p < 0.001), and the series success rate from 80.0% to 95.4% (p < 0.001). Subgroup analysis showed more consistent performance across age and sex groups. The proposed pipeline significantly outperforms Payer's method, enhancing segmentation accuracy and reducing longitudinal matching errors while minimizing evaluation workload. Its uncertainty analysis ensures robust performance, making it a valuable tool for longitudinal studies in multiple myeloma.

In this paper, we introduce the Visual Query-based task of Video Clip Localization (VQ-VCL) for medical video understanding. Specifically, we aim to retrieve a video clip containing frames similar to a given exemplar frame from a given input video. To solve the task, we propose a novel visual query-based video clip localization model called TIER-LOC. TIER-LOC is designed to improve video clip retrieval, especially in fine-grained videos by extracting features from different levels, i.e., coarse to fine-grained, referred to as TIERS. The aim is to utilize multi-Tier features for detecting subtle differences, and adapting to scale or resolution variations, leading to improved video-clip retrieval. TIER-LOC has three main components: (1) a Multi-Tier Spatio-Temporal Transformer to fuse spatio-temporal features extracted from multiple Tiers of video frames with features from multiple Tiers of the visual query enabling better video understanding. (2) a Multi-Tier, Dual Anchor Contrastive Loss to deal with real-world annotation noise which can be notable at event boundaries and in videos featuring highly similar objects. (3) a Temporal Uncertainty-Aware Localization Loss designed to reduce the model sensitivity to imprecise event boundary. This is achieved by relaxing hard boundary constraints thus allowing the model to learn underlying class patterns and not be influenced by individual noisy samples. To demonstrate the efficacy of TIER-LOC, we evaluate it on two ultrasound video datasets and an open-source egocentric video dataset. First, we develop a sonographer workflow assistive task model to detect standard-frame clips in fetal ultrasound heart sweeps. Second, we assess our model's performance in retrieving standard-frame clips for detecting fetal anomalies in routine ultrasound scans, using the large-scale PULSE dataset. Lastly, we test our model's performance on an open-source computer vision video dataset by creating a VQ-VCL fine-grained video dataset based on the Ego4D dataset. Our model outperforms the best-performing state-of-the-art model by 7%, 4%, and 4% on the three video datasets, respectively.

This study aimed to develop a computed tomography (CT)-based multi-organ segmentation model for delineating organs-at-risk (OARs) in pediatric upper abdominal tumors and evaluate its robustness across multiple datasets. In-house postoperative CTs from pediatric patients with renal tumors and neuroblastoma (n = 189) and a public dataset (n = 189) with CTs covering thoracoabdominal regions were used. Seventeen OARs were delineated: nine by clinicians (Type 1) and eight using TotalSegmentator (Type 2). Auto-segmentation models were trained using in-house (Model-PMC-UMCU) and a combined dataset of public data (Model-Combined). Performance was assessed with Dice Similarity Coefficient (DSC), 95 % Hausdorff Distance (HD95), and mean surface distance (MSD). Two clinicians rated clinical acceptability on a 5-point Likert scale across 15 patient contours. Model robustness was evaluated against sex, age, intravenous contrast, and tumor type. Model-PMC-UMCU achieved mean DSC values above 0.95 for five of nine OARs, while the spleen and heart ranged between 0.90 and 0.95. The stomach-bowel and pancreas exhibited DSC values below 0.90. Model-Combined demonstrated improved robustness across both datasets. Clinical evaluation revealed good usability, with both clinicians rating six of nine Type 1 OARs above four and six of eight Type 2 OARs above three. Significant performance differences were only found across age groups in both datasets, specifically in the left lung and pancreas. The 0-2 age group showed the lowest performance. A multi-organ segmentation model was developed, showcasing enhanced robustness when trained on combined datasets. This model is suitable for various OARs and can be applied to multiple datasets in clinical settings.

Idiopathic Pulmonary Fibrosis (IPF) is a progressive lung disease that continuously scars and thickens lung tissue, leading to respiratory difficulties. Timely assessment of IPF progression is essential for developing treatment plans and improving patient survival rates. However, current clinical standards require multiple (usually two) CT scans at certain intervals to assess disease progression. This presents a dilemma: the disease progression is identified only after the disease has already progressed. To address this issue, a feasible solution is to generate the follow-up CT image from the patient's initial CT image to achieve early prediction of IPF. To this end, we propose a lung structure and function information-guided residual diffusion model. The key components of our model include (1) using a 2.5D generation strategy to reduce computational cost of generating 3D images with the diffusion model; (2) designing structural attention to mitigate negative impact of spatial misalignment between the two CT images on generation performance; (3) employing residual diffusion to accelerate model training and inference while focusing more on differences between the two CT images (i.e., the lesion areas); and (4) developing a CLIP-based text extraction module to extract lung function test information and further using such extracted information to guide the generation. Extensive experiments demonstrate that our method can effectively predict IPF progression and achieve superior generation performance compared to state-of-the-art methods.

To evaluate and compare the diagnostic performance of CT texture analysis (CTTA), perfusion CT (PCT), and dual-energy CT (DECT) in distinguishing between clear-cell renal cell carcinoma (ccRCC) and non-ccRCC. This retrospective study included 66 patients with RCC (52 ccRCC and 14 non-ccRCC) who underwent DECT and PCT imaging before surgery (2017-2022). This DECT parameters (iodine concentration, iodine ratio [IR]) and PCT parameters (blood flow, blood volume, mean transit time, time to peak) were measured using circular regions of interest (ROIs). CT texture analysis features were extracted from manually annotated corticomedullary-phase images. A machine learning (ML) model was developed to differentiate RCC subtypes, with performance evaluated using k-fold cross-validation. Multivariate logistic regression analysis was performed to assess the predictive value of each imaging modality. All 3 imaging modalities demonstrated high diagnostic accuracy, with F1 scores of 0.9107, 0.9358, and 0.9348 for PCT, DECT, and CTTA, respectively. None of the 3 models were significantly different (P > 0.05). While each modality could effectively differentiate between ccRCC and non-ccRCC, higher IR on DECT and increased entropy on CTTA were independent predictors of ccRCC, with F1 scores of 0.9345 and 0.9272, respectively (P < 0.001). Dual-energy CT achieved the highest individual performance, with IR being the best predictor (F1 = 0.902). Iodine ratio was significantly higher in ccRCC (65.12 ± 23.73) compared to non-ccRCC (35.17 ± 17.99, P < 0.001), yielding an Area under curve (AUC) of 0.91, sensitivity of 87.5%, and specificity of 89.3%. Entropy on CTTA was the strongest texture feature, with higher values in ccRCC (7.94 ± 0.336) than non-ccRCC (6.43 ± 0.297, P < 0.001), achieving an AUC of 0.94, sensitivity of 83.0%, and specificity of 92.3%. The combined ML model integrating DECT, PCT, and CTTA parameters yielded the highest diagnostic accuracy, with an F1 score of 0.954. PCT, DECT, and CTTA effectively differentiate RCC subtypes. However, IR (DECT) and entropy (CTTA) emerged as key independent markers, suggesting their clinical utility in RCC characterization. Accurate, non-invasive biomarkers are essential to differentiate RCC subtypes, aiding in prognosis and guiding targeted therapies, particularly in ccRCC, where treatment options differ significantly.

The recent popularity of foundation models and the pre-train-and-adapt paradigm, where a large-scale model is transferred to downstream tasks, is gaining attention for volumetric medical image segmentation. However, current transfer learning strategies devoted to full fine-tuning for transfer learning may require significant resources and yield sub-optimal results when the labeled data of the target task is scarce. This makes its applicability in real clinical settings challenging since these institutions are usually constrained on data and computational resources to develop proprietary solutions. To address this challenge, we formalize Few-Shot Efficient Fine-Tuning (FSEFT), a novel and realistic scenario for adapting medical image segmentation foundation models. This setting considers the key role of both data- and parameter-efficiency during adaptation. Building on a foundation model pre-trained on open-access CT organ segmentation sources, we propose leveraging Parameter-Efficient Fine-Tuning and black-box Adapters to address such challenges. Furthermore, novel efficient adaptation methodologies are introduced in this work, which include Spatial black-box Adapters that are more appropriate for dense prediction tasks and constrained transductive inference, leveraging task-specific prior knowledge. Our comprehensive transfer learning experiments confirm the suitability of foundation models in medical image segmentation and unveil the limitations of popular fine-tuning strategies in few-shot scenarios. The project code is available: https://github.com/jusiro/fewshot-finetuning.

Medical image segmentation is a pivotal task within the realms of medical image analysis and computer vision. While current methods have shown promise in accurately segmenting major regions of interest, the precise segmentation of boundary areas remains challenging. In this study, we propose a novel network architecture named CTO, which combines Convolutional Neural Networks (CNNs), Vision Transformer (ViT) models, and explicit edge detection operators to tackle this challenge. CTO surpasses existing methods in terms of segmentation accuracy and strikes a better balance between accuracy and efficiency, without the need for additional data inputs or label injections. Specifically, CTO adheres to the canonical encoder-decoder network paradigm, with a dual-stream encoder network comprising a mainstream CNN stream for capturing local features and an auxiliary StitchViT stream for integrating long-range dependencies. Furthermore, to enhance the model's ability to learn boundary areas, we introduce a boundary-guided decoder network that employs binary boundary masks generated by dedicated edge detection operators to provide explicit guidance during the decoding process. We validate the performance of CTO through extensive experiments conducted on seven challenging medical image segmentation datasets, namely ISIC 2016, PH2, ISIC 2018, CoNIC, LiTS17, BraTS, and BTCV. Our experimental results unequivocally demonstrate that CTO achieves state-of-the-art accuracy on these datasets while maintaining competitive model complexity. The codes have been released at: CTO.

Lung cancer remains the leading cause of cancer-related mortality worldwide, with its early detection and effective treatment posing significant clinical challenges. Radiomics, the extraction of quantitative features from medical imaging, has emerged as a promising approach for enhancing diagnostic accuracy, predicting treatment responses, and personalising patient care. This review explores the role of radiomics in lung cancer diagnosis and management, with methods ranging from handcrafted radiomics to deep learning techniques that can capture biological intricacies. The key applications are highlighted across various stages of lung cancer care, including nodule detection, histology prediction, and disease staging, where artificial intelligence (AI) models demonstrate superior specificity and sensitivity. The article also examines future directions, emphasising the integration of large language models, explainable AI (XAI), and super-resolution imaging techniques as transformative developments. By merging diverse data sources and incorporating interpretability into AI models, radiomics stands poised to redefine clinical workflows, offering more robust and reliable tools for lung cancer diagnosis, treatment planning, and outcome prediction. These advancements underscore radiomics' potential in supporting precision oncology and improving patient outcomes through data-driven insights.

This study aims to investigate the impact of adaptive statistical iterative reconstruction-Veo (ASIR-V) and deep learning image reconstruction (DLIR) algorithms on the quantification of pericoronary adipose tissue (PCAT) and epicardial adipose tissue (EAT). Furthermore, we propose to explore the feasibility of correcting the effects through fat threshold adjustment. A retrospective analysis was conducted on the imaging data of 134 patients who underwent coronary CT angiography (CCTA) between December 2023 and January 2024. These data were reconstructed into seven datasets using filtered back projection (FBP), ASIR-V at three different intensities (ASIR-V 30%, ASIR-V 50%, ASIR-V 70%), and DLIR at three different intensities (DLIR-L, DLIR-M, DLIR-H). Repeated-measures ANOVA was used to compare differences in fat, PCAT and EAT attenuation values among the reconstruction algorithms, and Bland-Altman plots were used to analyze the agreement between ASIR-V or DLIR and FBP algorithms in PCAT attenuation values. Compared to FBP, ASIR-V 30 %, ASIR-V 50 %, ASIR-V 70 %, DLIR-L, DLIR-M, and DLIR-H significantly increased fat attenuation values (-103.91 ± 12.99 HU, -102.53 ± 12.68 HU, -101.14 ± 12.78 HU, -101.81 ± 12.41 HU, -100.87 ± 12.25 HU, -99.08 ± 12.00 HU vs. -105.95 ± 13.01 HU, all p < 0.001). When the fat threshold was set at -190 to -30 HU, ASIR-V and DLIR algorithms significantly increased PCAT and EAT attenuation values compared to FBP algorithm (all p < 0.05), with these values increasing as the reconstruction intensity level increased. After correction with a fat threshold of -200 to -35 HU for ASIR-V 30 %, -200 to -40 HU for ASIR-V 50 % and DLIR-L, and -200 to -45 HU for ASIR-V 70 %, DLIR-M, and DLIR-H, the mean differences in PCAT attenuation values between ASIR-V or DLIR and FBP algorithms decreased (-0.03 to 1.68 HU vs. 2.35 to 8.69 HU), and no significant difference was found in PCAT attenuation values between FBP and ASIR-V 30 %, ASIR-V 50 %, ASIR-V 70 %, DLIR-L, and DLIR-M (all p > 0.05). Compared to the FBP algorithm, ASIR-V and DLIR algorithms increase PCAT and EAT attenuation values. Adjusting the fat threshold can mitigate the impact of ASIR-V and DLIR algorithms on PCAT attenuation values.