25 studies from the fixed 4 November 2026 snapshot, resource version 0.3.2. Modality, task, reference standard and level descriptions are preserved from the manuscript table. Author verification is incomplete. Mechanism level does not imply clinical benefit.
All evidence resources · CSV · JSON
| Key | Study | Modality and task | Data and reference standard | Level | Source url |
|---|---|---|---|---|---|
| shen2026evomedagent | Evo-MedAgent | Chest radiograph question answering | ChestAgentBench answers, revealed after each case | 1 | Original source |
| feng2026medexpmem | MedExpMem | CT, MRI, radiography and ultrasound differential diagnosis | Eurorad teaching-file diagnoses; temporal split | 1 | Original source |
| wu2026medrsi | MedRSI | Imaging diagnosis and tool construction | Public benchmark labels and private retrospective cohorts | 2, implicit | Original source |
| chen2026medforge | MedForge-RSI | Medical-image forgery detection | Verified forgery labels; separate final test set | 1 | Original source |
| coevolving2025 | TissueLab | Imaging and omics analysis with expert interaction | Clinician annotations and accept/reject feedback | 1 | Original source |
| fan2026macro | MACRO | Multiple medical imaging analysis tasks | Public dataset labels | 2, implicit | Original source |
| lee2026revision3d | ReVision3D | CT liver and kidney lesion localization | Reference segmentation masks (LiTS, KiTS23) | 2, explicit | Original source |
| xu2026oaims | OAIMS | Interactive segmentation under modality and pathology shift | Reference masks; correction clicks simulated from labels | 1 | Original source |
| tzanis2026reclaim | ReclAIm | MRI, CT and radiograph classifier maintenance | Dataset labels; monitoring cohort reused for reassessment | 1 | Original source |
| hu2026templar | TEMPLAR | PET/CT and CT reporting templates | Report corpora; expert and LLM-judged ratings | 1 | Original source |
| zhuang2025doctor | Learning to Be A Doctor | Dermatology image diagnosis with workflow search | Skin Concepts and augmented Skin Conditions labels | 1 | Original source |
| sayeedi2026r4 | R4Agent | Chest radiograph detection and reporting | Public chest X-ray datasets; memory written from self-critique | 1 | Original source |
| zhu2026ultrasoundagents | Ultrasound-Agents | Breast ultrasound malignancy and BI-RADS assessment | Public datasets; attribute, box and BI-RADS labels | 2, implicit | Original source |
| chen2026theraagent | TheraAgent | PSMA PET/CT-based radioligand therapy outcome prediction | Single-centre cohort; documented PSA response and survival | 1 | Original source |
| chen2026skingptx | SkinGPT-X | Dermatology image diagnosis | Public clinical and dermoscopic datasets; expert labels | 1 | Original source |
| meng2026topoagent | TopoAgent | Topological descriptors for 2D image classification across 11 modalities | Public benchmark labels; reflection on feature statistics | 1 | Original source |
| qu2026sageagent | SAGEAgent | Glioma survival prediction with staged radiology acquisition | TCGA-LGG, TCGA-GBM and BraTS outcomes | 1 | Original source |
| wang2026mira | MIRA | Medical image question answering | Public VQA benchmark answers | 2, implicit | Original source |
| liu2026bat | BaT | Medical-imaging AI development and radiology viewer tasks | Synthetic sandboxes scored by a rubric verifier | 2, explicit | Original source |
| shao2026case | CASE | Longitudinal multi-sequence MRI diagnosis questions | UK Biobank MRI; ICD-10-derived multiple-choice labels | 2, implicit | Original source |
| feng2026gatedopenclaw | Gated OpenClaw | Fundus-image diagnosis with adapter and guidance updates | Private fundus images; report-simulated feedback; test cases reused | 1 | Original source |
| li2026emai | EMAI | Chest-radiology report summarization | MIMIC-CXR and Open-I report texts; reference impressions and clinician ratings | 1 | Original source |
| li2026decisionrules | Evolved decision rules | Multi-image medical video question answering | MedFrameQA answers; video-grouped evolution, selection and test partitions | 1 | Original source |
| fan2026medenv | MedEnv | Simulated diagnosis with optional chest radiograph review | Virtual EHR environment; author-reported held-out cases and reward ablations | 2, implicit | Original source |
| chen2027mdtace | MDTace | Multidisciplinary medical text and image QA | MedQA, PubMedQA, VQA-RAD and SLAKE; training-derived Playbook only | 1 | Original source |