Active
Cancer data interoperability and harmonization
Cancer Data Interoperability and Harmonization
This program brings together HARMONY Alliance work on harmonized hematologic malignancy data and IKIM work on turning real-world pathology reports and clinical text into structured representations — including automated tumor ICD coding and clinical embeddings for retrieval-augmented generation.
What problem are we trying to solve?
How can hematologic and oncologic data from different hospitals, languages and recording systems be represented so that they can be compared, retrieved and reused — in large consortia and in routine care?
Why it matters
Multi-center research and clinical AI both fail when codes, text and cohorts cannot be aligned. Harmonized Alliance data and hospital-held text are complementary parts of the same problem.
Approach
HARMONY Alliance collaboration on shared, harmonized cohorts for hematologic malignancies, including unsupervised AML classification. With IKIM colleagues, published work on self-hosted language models for tumor ICD coding of real-world pathology reports, and on fine-tuning clinical embedding models to improve retrieval-augmented generation in health care. Amin Turki is a co-author on the IKIM papers, not first or last author.
Who is working on it
HARMONY Alliance
Consortium genomics and MRD work sits with the HARMONY programs, not only with the IKIM text papers below.
Papers and abstracts
- Automated Tumor International Classification of Diseases Coding of Real-World Pathology Reports Using Self-Hosted Large Language Models JCO Clin Cancer Inform 2026
- Improving Retrieval Augmented Generation for Health Care by Fine-Tuning Clinical Embedding Models: Development and Evaluation Study J Med Internet Res 2026