Abstract
Multidisciplinary tumor board (MTB) protocols condense years of oncological history into dense narrative text, a rich but unstructured knowledge source. Converting them into machine-readable form is a prerequisite for oncology Digital Twins, yet remains labor-intensive. We present a self-hosted framework that maps ovarian carcinoma patient journeys onto a structured data model using an open-weight large language model (gpt-oss-120b) with schema-constrained generation. The pipeline runs entirely within the institutional network, so no patient data leaves the hospital. We extract patient journeys for 200 patients and evaluate fidelity on a 50-patient expert gold standard, using content-based field alignment robust to treatment-line ordering. Agreement is highest for well-defined structural fields (identifiers, staging, histology, document metadata), reaching 98–100%, and remains strong across categorical fields; sparse result-valued fields such as the biomarker value are the mainremainingchallenge.Clinicianreviewraisedoverallfield-levelagreement to 92.5%. The gold standard is further validated against pharmacy records, which independently corroborate most of its dispensable drugs (recall 0.92, precision 0.84). We thereby establish a privacy-compliant ⋆ Corresponding author. 2 L. Houri et al. pathway toward structured substrates for medical Digital Twins, as an enabling data ingestion component rather than a Digital Twin itself.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/DT4H_033.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=3pv0Edkfjv
BibTex
@InProceedings{HouLee_Toward_MICCAISAT2026,
author = { Houri, Leen AND Tahiri, Natyra AND Voelker, Gaspar AND Arlanch, Elina AND Koechert, Marie AND Frank, Jonatan AND Kuligin, Leonid AND Lübberstedt, Jannik AND Bressem, Keno AND Boeker, Martin AND Tschochohei, Maximilian AND Lammert, Jacqueline},
title = { { Toward Oncology Digital Twins: Leveraging LLMs for Automated Longitudinal Data Ingestion in Ovarian Cancer Care } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17275},
month = {pending},
page = {pending}
}