medical-ai

This survey reviews the role of knowledge graphs in medicine across five key domains—clinical decision support, disease prediction, health recommender systems, precision medicine, and medical question answering—discussing applications, challenges, and future directions.

MedLatentDx proposes a latent multi-agent communication framework for cross-hospital rare-disease diagnosis, using latent KV blocks to share diagnostic evidence without exposing clinical text, and introduces the CrossRare-Bench benchmark.

The article envisions a future by 2050 where AI assistants are in every home, education is personalized, medical treatments are advanced, cities are smart, and human-AI collaboration is widespread.

The paper introduces LungKG, the first structured pulmonary knowledge graph, and Lung-R1, a LLM trained via KG-constrained reasoning and reinforcement learning for pulmonary diagnostic reasoning from EMRs. Lung-R1-14B achieves state-of-the-art performance on EMR diagnosis.

This paper presents LiverRisk, a machine learning framework for NAFLD risk prediction that combines gradient-boosted decision trees with conformal prediction to provide calibrated, distribution-free coverage guarantees on individual risk estimates, achieving high AUROC on internal and external cohorts.

PathPocket is a multimodal AI agentic co-pilot for evidence-grounded pathology, utilizing a comprehensive evidence corpus and hypergraph to outperform existing state-of-the-art methods on over 200,000 real-world cases.

This paper evaluates the open-weight LLM LLaMA 3.1 for automatic extraction of structured data from Dutch brain MRI reports, achieving high performance on visual rating scores and accurate detection of findings, with few-shot prompting improving extraction of numerical variables.

A large-scale study across 5 models (7B–72B), 10 biomedical QA datasets, 4 retrieval methods, and 4 corpora finds that RAG yields only small and inconsistent gains (1–2 points) over no-retrieval baselines in biomedical question answering. The study concludes that the main bottleneck is not retrieval quality but models' limited ability to effectively use retrieved evidence.

A user shares their experience using ChatGPT for complex medical caregiving and proposes the idea of aggregating multiple AI models to improve reliability by seeking consensus among different LLMs.

This paper introduces StenCE, a pretraining framework that uses cross-modal contrastive learning between ECG and X-ray angiography representations to detect severe coronary stenosis from ECGs, achieving high performance and enabling early diagnosis even in asymptomatic patients.

Researchers introduce MedSP1000, a 1,638-case interactive benchmark derived from standardized patient scenarios to evaluate LLMs as dynamic clinical agents across multi-turn encounters. Results show even the best model (GPT-5.5) completes only 60.4% of expert rubric items, suggesting current LLMs are not yet reliable enough for clinical practice.

This paper investigates whether compact, task-specific bi-encoders fine-tuned on synthetic data from large language models can outperform general-purpose embeddings for clinical code retrieval in non-English languages, achieving state-of-the-art results on Spanish benchmarks CodiESP and DISTEMIST.

EHRBench is an automated and reliable benchmark for evaluating LLMs on clinical decision-making tasks using real-world electronic health records, covering nearly 1M QA items across diagnosis, treatment, and prognosis tasks.

AutoMedBench is a workflow-aware benchmark for autonomous medical-AI research, evaluating agents across five stages on diverse medical imaging tasks. Stage-level scoring reveals validation as the weakest stage, highlighting the need for reliable verification in agentic workflows.

A medical professional shares their positive experience using ChatGPT to assist in diagnostic pathology, demonstrating the AI's ability to provide accurate and detailed analysis comparable to a dermatopathologist.

This paper presents a hybrid neural-symbolic pipeline for extracting follow-up instructions from clinical notes, using BioBERT and deterministic date arithmetic. It achieves high performance (Pair F1 ~0.99) compared to generative baselines.

This paper addresses the problem of tool failures in medical AI agents by proposing a GRPO-based reinforcement learning framework that leverages instance-level selection, disagreement-aware synergy learning, and entropy-guided sampling to correct erroneous tool consensus and improve reliability across seven medical benchmarks.

MedGuideX transforms clinical practice guidelines into executable decision logic to generate factual and counterfactual QA data for training medical LLMs, achieving a 10.28% relative improvement in average accuracy across clinical reasoning benchmarks.

This paper introduces HRVConformer, a hybrid Convolution-Transformer architecture for classifying neonatal hypoxic-ischemic encephalopathy directly from raw heart rate signals, achieving an AUC of 83.23% and outperforming baseline models like ResNet50 and Transformer.

Proposes RAG4Outcome, a retrieval-augmented generation framework integrating multimodal clinical data (PET-CT reports, surgical records, follow-up notes) to improve prognostic prediction in chronic osteomyelitis, enhancing interpretability and clinical reliability.