Map2VidSeg uses classifier-derived localization maps to generate prompts for video segmentation without segmentation training data. We revisit the pipeline under a fixed 224 & times; 224 visual-prompt regime and test whether newer foundation-model components improve performance. Starting from the CBMS 2025 system of DINOv2 attention, YOLOE, and SAM-2 on SUN-SEG polyp video, we replace the components along three axes: DINOv3 for map generation, YOLOE-26, for prompt-based segmentation, and SAM-3 for temporal refinement. A completed 2 & times; 2 prompt-source & times; YOLOE-stage factorial with SAM-3 fixed shows that the full DINOv3 + YOLOE-26 + SAM-3 upgrade reaches 0.6040 IoU and 0.6684 Dice on SUN-SEG Unseen, whereas the DINOv2 + YOLOE-11 + SAM-3 control reaches 0.6750 IoU and 0.7426 Dice. The prompt-source swap is near neutral with YOLOE-11, whereas YOLOE-26 is weaker at this prompt resolution and the full upgrade underperforms the additive expectation. In a matched comparison using identical cached YOLOE masks, SAM-3 tracker double-pass produced a modest paired gain over SAM-2 double-pass (+0.0275 IoU; 95% BCa CI [+0.0122,+0.0458]; p = 0.0012). A second-task diagnostic study on CholecSeg8k binary instruments shows the same pattern: integration changes improve the YOLOE stage, whereas SAM-3 reduced performance on multi-tool clips under shared-identity tracking, and class-aware tracking eliminated this loss. The results support pipeline-level evaluation and identify class factoring as a third design axis alongside prompt quality and prompt-model compatibility.
Objective: Tabular data learning is crucial in medical data analysis, but it has consistently overlooked a vital component within tables: column names. We aim to leverage column names and pertinent clinical knowledge when training medical tabular data, bridging the gap between the decision-making process of medical experts and conventional models. To protect privacy, all pre-processing is done locally. Materials and methods: We introduce a novel algorithm, named Interpretable Context-Aware Medical Concept Embedding strategy (ICAME), designed to emulate human experts’ data understanding. ICAME is among the first methods that explicitly utilize clinical knowledge in data analysis, applying LLMs for context-aware embedding generation. To ensure the patient data privacy, ICAME stores the embedding of predefined pseudo-values in the local database. This keeps private data local and minimizes the cost of online LLM API usage. Results: ICAME consistently improves performance across classification, regression, and prediction tasks when applied to the MIMIC-III public benchmark (average gains of 1.6%p and 2.9%p on AUROC and AUPRC, respectively). The performance improvement of ICAME is notable in a data-scarce scenario (average gains of 10.5%p and 6.3%p on AUROC and AUPRC, respectively). ICAME also provides interpretability of embeddings, confirming the incorporation of cell values and clinical knowledge. Discussion and Conclusion: ICAME effectively integrates explicit clinical knowledge into tabular data learning for the first time, boosting performance. Additionally, it showcases high scalability, covering numerical, categorical, and binary data types. Through extensive experiments, we demonstrate the potential of ICAME as a promising tool for future medical tabular data analysis.
Remote rehabilitation, or telerehabilitation, has emerged as a scalable modality for delivering rehabilitative care outside clinical settings. Human Digital Twin (HDT) offers a promising paradigm for enabling personalized and adaptive telerehabilitation by continuously representing a patient’s functional condition through digital technologies. Despite the growing interest in this field, current research still lacks a comprehensive understanding of how enabling technologies contribute to HDT-based telerehabilitation. This systematic review provides a technological perspective on telerehabilitation with the aim of identifying recurring architectural patterns, analyzing current technological limitations, and deriving design guidelines for future HDT-based rehabilitation systems. Following the PRISMA methodology, fifty-four studies published between 2021 and 2026 were retrieved from Scopus, ScienceDirect, and PubMed and analyzed according to six dimensions: clinical domain, type of Extended Reality (XR) technology, integration of Internet of Medical Things (IoMT) devices, Artificial Intelligence (AI) task, HDT implementation level, and study maturity level. Across the reviewed studies, a common technological workflow emerged in which sensing technologies, AI models, XR interfaces, and network infrastructures jointly support remote monitoring, rehabilitation assessment, and clinical supervision. To translate these findings into actionable recommendations, a Quality Function Deployment (QFD) analysis was conducted to map clinical and operational requirements to technological features. Based on the QFD analysis, a set of guidelines is proposed to translate the identified technological priorities into practical design principles for future research and the development of integrated, scalable, and clinically deployable telerehabilitation frameworks.
Identifying and understanding multi-type drug-drug interactions and their associated toxicities is crucial for preventing adverse outcomes, especially in vulnerable populations, such as cancer patients, who may already have compromised health statuses. This paper provides a benchmark study that compares various commonly-used Machine Learning-based approaches, such as graph embeddings and path analysis, aiming at predicting different types of drug-drug interaction links in a disease-specific knowledge graph, constructed from biomedical literature. These approaches are evaluated based on an external interaction dataset collected from Drugbank. We first introduce a simplified taxonomy of multi-type interactions and then compare the models' performance over the different drug-drug interaction types. Overall, the graph embeddings approach outperforms the competition, while the path-based analysis allows for an interpretation of predictions, using the most important features of each path.
Gait analysis provides an objective characterization of locomotor function and is widely used to support diagnosis and rehabilitation monitoring across neurological and orthopedic disorders. Deep learning has been increasingly applied to this domain, yet most approaches rely on supervised classifiers trained on disease-labeled data, limiting generalization to heterogeneous pathological presentations. The methodological objective of this work is to develop a label-free framework for joint-level anomaly detection and kinematic correction based on a Transformer masked autoencoder trained exclusively on normative gait sequences from 150 adults, acquired with a markerless multi-camera motion-capture system.At inference, a two-pass procedure is applied to potentially pathological input sequences: first, it estimates joint inconsistency scores by occluding individual joints and measuring deviations from the learned normative prior. Then, it withholds the flagged joints from the encoder input and reconstructs the full skeleton from the remaining spatiotemporal context, yielding corrected kinematic trajectories at the flagged positions.The validation objective is to assess whether the framework preserves unseen normative gait and reduces angular deviation in simulated abnormal gait patterns.In this proof-of-concept evaluation, data from 10 held-out normative participants, who performed seven simulated abnormal gait patterns, showed a significant reduction in angular deviation across all analyzed joints with large effect sizes, and preservation of normative kinematics.The proposed approach enables interpretable, subject-specific localization of joints that are inconsistent with learned normative gait patterns and generation of an individualized normative reconstruction without requiring disease labels. Video is available at https://youtu.be/GenGait.
