Privacy-Preserving Dashboard for F.A.I.R Head and Neck Cancer data supporting multi-centered collaborations
Bibliographic record
Abstract
Research in modern healthcare requires vast volumes of data from various healthcare centers across the globe. It is not always feasible to centralize clinical data without compromising privacy. A tool addressing these issues and facilitating reuse of clinical data is the need of the hour. The Federated Learning approach, governed in a set of agreements such as the Personal Health Train (PHT) manages to tackle these concerns by distributing models to the data centers instead of the traditional approach of centralizing datasets. One of the prerequisites of PHT is using semantically interoperable datasets for the models to be able to find them. FAIR (Findable, Accessible, Interoperable, Reusable) principles help in building interoperable and reusable data by adding knowledge representation and providing descriptive metadata. However, the process of making data FAIR is not always easy and straight-forward. Our main objective is to disentangle this process by using domain and technical expertise and get data prepared for federated learning. This paper introduces applications that are easily deployable as Docker containers, which will automate parts of the aforementioned process and significantly simplify the task of creating FAIR clinical data. Our method bypasses the need for clinical researchers to have a high degree of technical skills. We demonstrate the FAIR-ification process by applying it to five Head and Neck cancer datasets (four public and one private). The PHT paradigm is explored by building a distributed visualization dashboard from the aggregated summaries of the FAIR-ified datasets. Using the PHT infrastructure for exchanging only statistical summaries or model coefficients allows researchers to explore data from multiple centers without breaching privacy.
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 | 0.012 | 0.031 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.004 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.009 | 0.009 |
| Open science | 0.003 | 0.010 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.010 | 0.004 |
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".