Making head and neck cancer clinical data Findable-Accessible-Interoperable-Reusable to support multi-institutional collaboration and federated learning
Bibliographic record
Abstract
Objectives: Federated learning (FL) is a group of methodologies where statistical modelling can be performed without exchanging identifiable patient data between cooperating institutions. To realize its potential for AI development on clinical data, a number of bottlenecks need to be addressed. One of these is making data Findable-Accessible-Interoperable-Reusable (FAIR). The primary aim of this work is to show that tools making data FAIR allow consortia to collaborate on privacy-aware data exploration, data visualization, and training of models on each other's original data. Methods: We propose a "Schema-on-Read" FAIR-ification method that adapts for different (re)analyses without needing to change the underlying original data. The procedure involves (1) decoupling the contents of the data from its schema and database structure, (2) annotation with semantic ontologies as a metadata layer, and (3) readout using semantic queries. Open-source tools are given as Docker containers to help local investigators prepare their data on-premises. Results: We created a federated privacy-preserving visualization dashboard for case mix exploration of 5 distributed datasets with no common schema at the point of origin. We demonstrated robust and flexible prognostication model development and validation, linking together different data sources-clinical risk factors and radiomics. Conclusions: Our procedure leads to successful (re)use of data in FL-based consortia without the need to impose a common schema at every point of origin of data. Advances in knowledge: This work supports the adoption of FL within the healthcare AI community by sharing means to make data more FAIR.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.053 | 0.109 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.003 |
| Bibliometrics | 0.004 | 0.004 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.010 | 0.014 |
| Open science | 0.005 | 0.023 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".