As Artificial intelligence (AI) is increasingly integrated into surgical practice, particularly in robotic surgery, the clinical intraoperative implementation remains limited. Continued progress will require not only technical advances but also a clear understanding of which functions surgeons find valuable in practice. This study aimed to assess surgeons’ perceptions, knowledge, attitudes, and current use of AI-driven intraoperative assistance. We conducted a structured, web-based survey of 53 surgeons across 5 continents, assessing demographics, attitudes, knowledge, current use, and perceived usefulness of five AI-based intraoperative guidance components, using video footage from robotic upper gastrointestinal surgeries. Participants were stratified by surgical experience level. Ordinal and categorical data were analyzed using non-parametric tests, and paired comparisons, with statistical significance set at p < 0.05. Perceived knowledge of AI tools for surgery was rated as average or lower by 83.0

The development of deep learning methods for magnetic resonance spectroscopy (MRS) is often hindered by limited availability of large, high-quality training datasets. While physics-based simulations are commonly used to mitigate this limitation, accurately modeling all in-vivo signal components remains challenging. In this work, we propose a data-driven framework for synthesizing in-vivo MRS data using a variational autoencoder (VAE) trained exclusively on measured single-voxel spectroscopy data. The model learns a low-dimensional latent representation of complex-valued spectra and enables generation of new samples through latent-space sampling and interpolation. The generative performance of the proposed approach is evaluated using a comprehensive set of complementary analyses, including reconstruction quality, feature-level similarity using low-dimensional embeddings, application-based signal quality metrics, and metabolite quantification agreement. The results demonstrate that the VAE accurately reconstructs dominant spectral patterns and generates synthetic spectra that occupy the same feature space as in-vivo data. In an example application targeting GABA-edited spectroscopy, augmenting limited subsets of transients with synthetic spectra improves signal quality metrics such as signal-to-noise ratio, linewidth, and shape scores. However, the results also reveal limitations of the generative approach, including under-representation of stochastic noise and reduced accuracy in absolute metabolite quantification, particularly for applications sensitive to concentration estimates. These findings highlight both potential and limitations of data-driven MRS synthesis. Beyond the proposed model, this study introduces a structured evaluation framework for generative MRS methods, emphasizing the importance of application-aware validation when synthetic data are used for downstream analysis.

Accurate recognition of anatomical structures is essential for safe and effective minimally invasive surgery (MIS), yet it remains underexplored in surgical computer vision due to limited annotated data and methods tailored primarily to natural scenes. In this work, we present a combined dataset and model framework to advance anatomy-aware perception in MIS. First, we introduce ATLAS-120k, a large-scale clip-level semantic segmentation dataset comprising over 120,000 annotated frames from 100 surgical videos spanning 14 procedures and multiple modalities, including laparoscopic and robot-assisted surgery. The dataset captures substantial procedural variability and was created using a scalable annotation pipeline that integrates expert manual labeling, automated propagation, iterative refinement, and surgeon verification to ensure high-quality annotations. Second, we propose ATLAS (Anatomy Recognition with Context Learning using Foundation Representations), a video semantic segmentation model specifically designed for surgical anatomy recognition. Unlike conventional approaches that emphasize object tracking, ATLAS leverages foundation-model embeddings together with lightweight temporal reasoning to incorporate contextual cues such as procedure type, surgical phase, and short-term visual memory. This design enables temporally consistent and accurate predictions while maintaining real-time feasibility. Together, the dataset and model establish a practical foundation for robust surgical scene understanding and support the development of clinically applicable guidance systems for minimally invasive surgery. The models, dataset annotations and annotation platform are publicly available at: https://github.com/TimJaspers0801/ATLAS.

Visual Question Answering (VQA) in robotic surgery, referred to as surgical VQA, requires high-level understanding of complex surgical scenes and the integration of visual perception with language reasoning, with the potential to support surgical training and intraoperative decision-making. Recent Vision-Language Models (VLMs) have shown promising performance through parameter-efficient fine-tuning; however, most existing approaches rely on coarse visual grounding, typically limited to bounding boxes, which fails to capture the fine-grained spatial structure of surgical objects. In this work, we propose a unified framework that jointly performs pixel-level segmentation and visual question answering within a single framework. Our approach integrates a VLM with a Segment Anything Model (SAM)-based decoder and represents scene elements as object tokens generated by the VLM. These object tokens guide answer prediction and are further projected to the SAM-based decoder to produce segmentation masks. By optimizing the object token embeddings through both segmentation and question answering objectives, the model learns spatially grounded representations that enhance visual reasoning while providing explicit pixel-level grounding. We evaluate the proposed method on the private RAMIE (Robot-Assisted Minimally Invasive Esophagectomy) dataset and the public EndoVis18 dataset, where it consistently outperforms baseline methods for surgical VQA. These results demonstrate that incorporating context-aware object tokens into vision-language models improves fine-grained surgical scene understanding.

Medical image segmentation is clinically important, yet data privacy and the cost of expert annotation limit the availability of labeled data. Federated semi-supervised learning (FSSL) offers a solution but faces two challenges: pseudo-label reliability depends on the strength of local models, and client devices often require compact or heterogeneous architectures due to limited computational resources. These constraints reduce the quality and stability of pseudo-labels, while large models, though more accurate, cannot be trained or used for routine inference on client devices. We propose SAM-Fed, a federated semi-supervised framework that leverages a high-capacity segmentation foundation model to guide lightweight clients during training. SAM-Fed combines dual knowledge distillation with an adaptive agreement mechanism to refine pixel-level supervision. Experiments on skin lesion and polyp segmentation across homogeneous and heterogeneous settings show that SAM-Fed consistently outperforms state-of-the-art FSSL methods.

PURPOSE:The aim of this study is to improve the traditional T1-weighted (T1w) over T2-weighted (T2w) ratio as a proxy for myelin by investigating the optimal T2w TE and combination of exponent-weighted T1w and T2w images ( T1w x 1 / T2w x 2 ). METHODS:T1w and T2w Gradient And Spin Echo (GRASE) data were acquired from 14 volunteers, 6 of whom had a repeated GRASE scan. T1w x 1 / T2w x 2 ratios were computed for combinations of x 1 and x 2 ranging 0-5 in steps of 0.1 and T2w GRASE TE ranging 10-160 ms. Ratios were correlated with myelin-water fraction (MWF) maps, as a reference MRI myelin biomarker. Analyses were performed on white matter (WM) and deep gray matter (dGM). Reliability was evaluated for six subjects. RESULTS:The optimized ratio with TE = 10 ms , x 1 = 2 . 3 and x 2 = 1 . 1 significantly increased correlation with MWF compared to the traditional T1w / T2w ( 80 ms ) ratio (paired t -test on Fisher- z values, optimized ρ = 0 . 55 , traditional ρ = 0 . 38 , p < 0 . 001 ) while preserving excellent test-retest reliability (ICC(3,1) = 0.95 for both). CONCLUSION:The ratio associated with myelin can be improved using optimized parameters without compromising reliability. Specifically, low T2w TE can significantly increase the correlation with MWF. T1w over proton density-weighted (T1w/PDw) metric can be a viable alternative to T1w / T2w , more strongly associated with MWF.

Deep learning-based surgical vision systems achieve strong performance in semantic segmentation and phase recognition, but their black-box nature limits traceability in safety-critical clinical settings. Prototype-based networks offer an interpretable alternative by grounding predictions in learned visual exemplars, yet their suitability for surgical video understanding remains insufficiently characterized. We adapted a prototype-based architecture to two surgical datasets, laparoscopic cholecystectomy and robot-assisted minimally invasive esophagectomy (RAMIE), and benchmarked it against conventional baselines. We evaluated a segmentation-only setting, in which prototype size and capacity were ablated, and a multitask setting, in which three strategies for coupling prototype learning to semantic segmentation and surgical phase recognition were compared. Prototype-based models underperformed the conventional baselines across both tasks and datasets. In the segmentation-only setting, the selected prototype configurations achieved Dice scores of 72.23% on Cholecystectomy and 72.07% on RAMIE, compared with 74.35% and 74.02% for the corresponding conventional baselines, and showed weaker boundary agreement. In the multitask setting, the best prototype strategy recovered competitive segmentation performance but remained 6–12 F1 points below the conventional baseline for phase recognition. Qualitatively, prototype activation maps exposed intra-structure decompositions and contextual cues that are not directly available from black-box baselines. Prototype networks provide spatially traceable evidence for surgical scene understanding, but currently trade interpretability for reduced boundary precision and phase-recognition performance. These findings motivate future work on scene-level and temporally aware prototypes for explainable surgical AI.

BACKGROUND:Progression independent of relapse- and magnetic resonance imaging (MRI)-activity (PIRMA) in relapsing-remitting multiple sclerosis (RRMS) may result from subtle myelin pathology in normal-appearing white matter (NAWM) and perilesions. OBJECTIVE:To evaluate whether short-term myelin changes, estimated with the MRI-based T1w/FLAIR ratio, are associatied with long-term clinical progression in patients with RRMS receiving natalizumab. METHODS:For 29 patients with RRMS, short-term, six-month myelin changes were calculated for three areas, i.e. white matter hyperintensities (WMH), a perilesional area of 2 mm, and NAWM. Expanded disability status scale (EDSS) scores at baseline and after at least three years were used to define long-term clinical progression. RESULTS:Logistic regression models demonstrated significant associations between long-term progression using short-term changes in myelin proxy in NAWM (β = -77.962; p = .039), but not in perilesional tissue (β = -3.999; p = .717) and WMH (β = -4.324; p = .752). ROC-curve of the logistic regression model of NAWM showed a good discriminative value (area under the curve (AUC) = 0.892, 95% CI [.755-1.000]) for clinical progression. CONCLUSION:By assessing short-term myelin changes outside lesions, using readily available clinical MRI scans, long-term clinical progression in patients with RRMS may be predicted.

Curative treatment of resectable esophageal cancer comprises neoadjuvant chemoradiotherapy and esophagectomy. Robot-assisted minimally invasive esophagectomy (RAMIE) is the preferred technique; however, learning RAMIE is challenging due to the complex chest anatomy, patient positioning and zoomed-in camera view. Computer-aided anatomy recognition holds promise for improving surgical navigation. This study aims to develop real-time anatomy recognition and surgical phase recognition algorithms for the thoracic part of RAMIE using deep learning and to understand the challenges of current state-of-the-art algorithms. A retrospective single-center cohort study was conducted on prospectively collected RAMIE videos at University Medical Center Utrecht, The Netherlands. Two datasets were created: an anatomy segmentation dataset with 1504 frames from 53 videos, annotated for eight anatomical classes and four surgical instruments; and a surgical phase dataset, with 38 videos labeled with thirteen distinct phases. Several deep learning models were trained and tested using both datasets. The SegNeXt model achieved the most accurate segmentations with an overall overlap score of 0.72 for all classes, of which the surgical instruments were best detectable. The best surgical phase recognition model achieved an overall accuracy of 82.8

Robotic-assisted minimally invasive esophagectomy (RAMIE) is a recognized treatment for esophageal cancer, offering better patient outcomes compared to open surgery and traditional minimally invasive surgery. RAMIE is highly complex, spanning multiple anatomical areas and involving repetitive phases and non-sequential phase transitions. Our goal is to leverage deep learning for surgical phase recognition in RAMIE to provide intraoperative support to surgeons. To achieve this, we have developed a new surgical phase recognition dataset comprising 27 videos. Using this dataset, we conducted a comparative analysis of state-of-the-art surgical phase recognition models. To more effectively capture the temporal dynamics of this complex procedure, we developed a novel deep learning model featuring an encoder-decoder structure with causal hierarchical attention, which demonstrates superior performance compared to existing models.

Perivascular spaces (PVS), also known as Virchow-Robin spaces, are critical biomarkers for diagnosing cerebral small vessel disease (CSVD). Quantifying PVS visible in magnetic resonance imaging (MRI) is essential for understanding their relationship with various neurological disorders. Traditional methods for assessing PVS rely on visual scoring of MRI images, which is time-consuming, subjective, and unsuitable for large-scale studies. Additionally, due to their small size, scattered distribution, and complex morphology, PVS can easily be confused with neighboring structures, posing significant challenges for their accurate extraction. In this paper, we propose a novel graph interaction-enhanced model based on vision-language modeling (VLM) technology for accurate PVS extraction from MRI. Our approach leverages textual information to guide image feature extraction and employs a graph structure to enhance cross-modal interactions, facilitating the reasoning of relationships between different modalities. Furthermore, we introduce a cross-modal attention mechanism for global feature alignment and an attention-based dynamic fusion module to effectively integrate multimodal information, improving the accuracy of PVS segmentation. Validated on an independent T1-weighted dataset, our model demonstrates superior performance in capturing both global and local information, addressing the limitations of traditional image-only models and providing a robust solution for PVS segmentation in complex clinical scenarios.

Deep learning-based anatomy segmentation holds promise for improving real-time guidance in complex surgeries such as robot-assisted minimally invasive esophagectomy (RAMIE). However, the clinical relevance of commonly used metrics for evaluating segmentation quality remains unclear, as previous assessments have lacked direct input from surgeons. This study aims to assess how well quantitative segmentation metrics reflect surgeons’ assessments of anatomical overlay accuracy and clinical usefulness during RAMIE. We conducted a survey involving 26 upper gastrointestinal surgeons, including both trainee and attending surgeons, who assessed video clips of RAMIE procedures featuring deep learning-generated anatomical overlays. We correlated the surgeons’ qualitative evaluations of annotation accuracy and clinical usefulness with a comprehensive set of quantitative metrics, including overlap, distance, temporal, and error-specific measures. The analysis encompassed over 8000 manually annotated frames from 12 video clips, with overlays generated by two state-of-the-art deep learning models. Overlap and temporal consistency metrics show the strongest correlation with surgeon assessments. Distance-based and error-specific metrics correlate moderately. Novices show weaker correlations and tend to rate overlays more leniently. Qualitative feedback reveals issues like hallucinations and instability, often missed by current metrics. Standard quantitative metrics partially reflect surgeon perceptions but should be complemented by surgeon-informed evaluations and task-specific metrics to better capture clinically relevant errors. Aligning metric design with surgical expertise is essential for the safe and effective integration of AI-guided anatomical segmentation in the operating room.

To develop a non-iterative method for applying elliptical field-of-view (FOV) to radial imaging and evaluate it for Stack-Of-Stars (SOS) with variable radial density in the k_z direction. New analytic expressions were derived to compute the radial profile angles for an elliptical FOV with and without golden angle sampling. With a major-to-minor-axis FOV ratio of 1:0.5, anisotropic FOV and variable density SOS were evaluated, using point spread function analysis, phantom imaging, and in vivo pelvic imaging. Compared with conventional SOS, elliptical density in k_z reduced scan time by 20

Background: Qualitative motion analysis revealed that the cervical spine moves according to a consistent pattern. This analysis calculates the relative rotation between vertebral segments to determine the sequence in which they contribute to extension, demonstrating a mean sensitivity of 90% and specificity of 85%. However, the extensive time that is required limits its applicability. This study investigated the feasibility of implementing a deep-learning model to analyze qualitative cervical motion. Methods: A U-Net architecture was implemented as 2D and 2D+t models. Dice similarity coefficient (DSC) and Intersection over Union (IoU) were used to assess the performance of the models. Intraclass Correlation Coefficient (ICC) was used to compare the relative rotation of individual vertebrae and segments to the ground truth. Results: IoU ranged from 0.37 to 0.74 and DSC ranged from 0.53 to 0.80. The ICC scores for relative rotation ranged from 0.62 to 0.96 for individual vertebrae and from 0.28 to 0.72 for vertebral segments. For segments, 2D+t models presented higher ICC scores compared to 2D models. Conclusions: This study showed the feasibility of implementing deep-learning models to analyze qualitative cervical motion in dynamic X-ray recordings. Future research should focus on improving model segmentation by enhancing recording contrast and applying post-processing methods. Improved segmentation accuracy will enable routine use of the analysis of motion patterns in clinical research. The absence or presence of a motion pattern, or identification of new patterns has the potential to aid in clinical decision-making.

Accurate surgical phase recognition is crucial for computer-assisted interventions and surgical video analysis. Annotating long surgical videos is labor-intensive, driving research toward leveraging unlabeled data for strong performance with minimal annotations. Although self-supervised learning has gained popularity by enabling large-scale pretraining followed by fine-tuning on small labeled subsets, semi-supervised approaches remain largely underexplored in the surgical domain. In this work, we propose a video transformer-based model with a robust pseudo-labeling framework. Our method incorporates temporal consistency regularization for unlabeled data and contrastive learning with class prototypes, which leverages both labeled data and pseudo-labels to refine the feature space. Through extensive experiments on the private RAMIE (Robot-Assisted Minimally Invasive Esophagectomy) dataset and the public Cholec80 dataset, we demonstrate the effectiveness of our approach. By incorporating unlabeled data, we achieve state-of-the-art performance on RAMIE with a 4.9 https://github.com/IntraSurge/SemiVT-Surge .

This paper presents a comprehensive systematic review of generative models (GANs, VAEs, DMs, and LLMs) used to synthesize various medical data types, including imaging (dermoscopic, mammographic, ultrasound, CT, MRI, and X-ray), text, time-series, and tabular data (EHR). Unlike previous narrowly focused reviews, our study encompasses a broad array of medical data modalities and explores various generative models. Our aim is to offer insights into their current and future applications in medical research, particularly in the context of synthesis applications, generation techniques, and evaluation methods, as well as providing a GitHub repository as a dynamic resource for ongoing collaboration and innovation. Our search strategy queries databases such as Scopus, PubMed, and ArXiv, focusing on recent works from January 2021 to November 2023, excluding reviews and perspectives. This period emphasizes recent advancements beyond GANs, which have been extensively covered in previous reviews. The survey also emphasizes the aspect of conditional generation, which is not focused on in similar work. Key contributions include a broad, multi-modality scope that identifies cross-modality insights and opportunities unavailable in single-modality surveys. While core generative techniques are transferable, we find that synthesis methods often lack sufficient integration of patient-specific context, clinical knowledge, and modality-specific requirements tailored to the unique characteristics of medical data. Conditional models leveraging textual conditioning and multimodal synthesis remain underexplored but offer promising directions for innovation. Our findings are structured around three themes: (1) Synthesis applications, highlighting clinically valid synthesis applications and significant gaps in using synthetic data beyond augmentation, such as for validation and evaluation; (2) Generation techniques, identifying gaps in personalization and cross-modality innovation; and (3) Evaluation methods, revealing the absence of standardized benchmarks, the need for large-scale validation, and the importance of privacy-aware, clinically relevant evaluation frameworks. These findings emphasize the need for benchmarking and comparative studies to promote openness and collaboration.

Esophageal cancer is among the most common types of cancer worldwide. It is traditionally treated using open esophagectomy, but in recent years, robot-assisted minimally invasive esophagectomy (RAMIE) has emerged as a promising alternative. However, robot-assisted surgery can be challenging for novice surgeons, as they often suffer from a loss of spatial orientation. Computer-aided anatomy recognition holds promise for improving surgical navigation, but research in this area remains limited. In this study, we developed a comprehensive dataset for semantic segmentation in RAMIE, featuring the largest collection of vital anatomical structures and surgical instruments to date. Handling this diverse set of classes presents challenges, including class imbalance and the recognition of complex structures such as nerves. This study aims to understand the challenges and limitations of current state-of-the-art algorithms on this novel dataset and problem. Therefore, we benchmarked eight real-time deep learning models using two pretraining datasets. We assessed both traditional and attention-based networks, hypothesizing that attention-based networks better capture global patterns and address challenges such as occlusion caused by blood or other tissues. The benchmark includes our RAMIE dataset and the publicly available CholecSeg8k dataset, enabling a thorough assessment of surgical segmentation tasks. Our findings indicate that pretraining on ADE20k, a dataset for semantic segmentation, is more effective than pretraining on ImageNet. Furthermore, attention-based models outperform traditional convolutional neural networks, with SegNeXt and Mask2Former achieving higher Dice scores, and Mask2Former additionally excelling in average symmetric surface distance.

Foundation models have revolutionized computer vision by achieving vastly superior performance across diverse tasks through large-scale pretraining on extensive datasets. However, their application in surgical computer vision has been limited. This study addresses this gap by introducing SurgeNetXL, a novel surgical foundation model that sets a new benchmark in surgical computer vision. Trained on the largest reported surgical dataset to date, comprising over 4.7 million video frames, SurgeNetXL achieves consistent top-tier performance across six datasets spanning four surgical procedures and three tasks, including semantic segmentation, surgical phase recognition, and critical view of safety (CVS) classification. Compared with the best-performing surgical foundation model, SurgeNetXL shows mean improvements of 4.0 %, 8.9 %, and 11.4 % for semantic segmentation, phase recognition, and CVS classification, respectively. Additionally, SurgeNetXL outperforms ImageNet1k by 16.1 %, 8.0 %, and 4.3 % for the respective tasks. In addition to advancing model performance, this study provides key insights into scaling pretraining datasets, extending training durations, and optimizing model architectures specifically for surgical computer vision. These findings pave the way for improved generalization and robustness in data-scarce scenarios, offering a comprehensive framework for future research in this domain. All models and a subset of the SurgeNetXL dataset, including over 2 million video frames, are publicly available at: https://github.com/TimJaspers0801/SurgeNet.

PURPOSE:Use a conference challenge format to compare machine learning-based gamma-aminobutyric acid (GABA)-edited magnetic resonance spectroscopy (MRS) reconstruction models using one-quarter of the transients typically acquired during a complete scan. METHODS:There were three tracks: Track 1: simulated data, Track 2: identical acquisition parameters with in vivo data, and Track 3: different acquisition parameters with in vivo data. The mean squared error, signal-to-noise ratio, linewidth, and a proposed shape score metric were used to quantify model performance. Challenge organizers provided open access to a baseline model, simulated noise-free data, guides for adding synthetic noise, and in vivo data. RESULTS:Three submissions were compared. A covariance matrix convolutional neural network model was most successful for Track 1. A vision transformer model operating on a spectrogram data representation was most successful for Tracks 2 and 3. Deep learning (DL) reconstructions with 80 transients achieved equivalent or better SNR, linewidth and fit error compared to conventional 320 transient reconstructions. However, some DL models optimized linewidth and SNR without actually improving overall spectral quality, indicating a need for more robust metrics. CONCLUSION:DL-based reconstruction pipelines have the promise to reduce the number of transients required for GABA-edited MRS.