Longitudinal records are fragmented and cross-institutional
Clinical reports accumulate across time with missing data, uncertainty, and cross-centre inconsistency. Case context must be assembled before safe reasoning can begin.
OMGs
Ovarian tumour Multidisciplinary intelligent aGent System
Ovarian tumour management increasingly relies on multidisciplinary tumour board deliberation, yet timely expert consensus remains scarce in many centres. OMGs uses domain-specific agents to integrate multidisciplinary evidence and produce structured, traceable MDT-style recommendations.
See how OMGs supports ovarian tumour MDT decision-making.
OMGs is designed for fragmented records, role-specific MDT reasoning, and outputs that can be traced back to reports, guidelines, and literature.
Clinical reports accumulate across time with missing data, uncertainty, and cross-centre inconsistency. Case context must be assembled before safe reasoning can begin.
Primary management, histology-led pathways, platinum-sensitive relapse, platinum-resistant relapse, and reassessment each require different evidence priorities.
Outputs need explicit grounding in guidelines, literature, and patient-specific reports, together with clear points for reassessment.
Clinical reports accumulate across time with missing data, uncertainty, and cross-centre inconsistency. Case context must be assembled before safe reasoning can begin.
Primary management, histology-led pathways, platinum-sensitive relapse, platinum-resistant relapse, and reassessment each require different evidence priorities.
Outputs need explicit grounding in guidelines, literature, and patient-specific reports, together with clear points for reassessment.
OMGs brings case structuring, evidence retrieval, report selection, and five-specialty MDT deliberation into one visible workflow.
Rather than returning a single opaque answer, OMGs assembles case context, retrieves evidence, tracks provenance, and stages multi-round discussion among the chair, oncology, radiology, pathology, and nuclear medicine agents before producing a structured MDT recommendation.
Structured EHR context, laboratory trends, imaging, pathology, genomics, and trial options are consolidated into a reasoning-ready patient view.
The chair, oncologist, radiologist, pathologist, and nuclear medicine physician deliberate with evidence search, report selection, and provenance tracking before returning a plan.
The final result includes case core, timeline, specialty evidence, references, Final Assessment, Core Treatment Strategy, and Change Triggers.
Each scene represents a distinct decision setting, with different evidence priorities and intervention choices.
Evaluation spans retrospective benchmarking, re-MDT comparison, prospective comparison with routine MDT conclusions, and human-AI collaboration.
Single-centre retrospective benchmarking against CHAIR-R, CHAIR-E, and CHAIR-D.
Highest mean SPEAR profileMulticentre retrospective re-MDT evaluation with comparability to expert re-MDT.
4.45 vs 4.53 overallProspective multicentre comparison against routine MDT conclusions in real workflow.
59 prospective patientsHuman-AI collaboration showing the strongest gains in Evidence and Robustness.
Clinician recommendations improved