JMIR Publications

ABSTRACT

Background:

Accurate delineation of the gross tumor volume (GTV) is crucial in radiotherapy for dose calculation and precise imaging-guided treatment of lung cancer patients. Conventionally, this task has been performed manually by radiation oncologists, which can be subjective and vary among clinicians. Deep learning has enabled automated GTV segmentation, with the potential to revolutionize the radiotherapy workflow by improving efficiency and consistency, ultimately enhancing patient outcomes while reducing clinician workload. However, the adoption of deep learning based GTV segmentation tools is hindered by the challenges of data privacy and the need for large, diverse datasets across multiple institutions. Federated learning (FL) offers a promising solution, allowing collaborative development of AI models without the need to share individual subject-level data.

Objective:

The objective is to introduce an innovative federated learning infrastructure called the Personal Health Train (PHT) that includes the procedural, technical, and governance components needed to implement federated learning on real-world healthcare data, including training deep learning neural networks. The study aims to apply this federated deep learning infrastructure to the use case of gross tumor volume (GTV) segmentation on chest CT images of lung cancer patients, and present the results from a proof-of-concept experiment.

Methods:

The PHT framework addresses the challenges of data privacy concerns of data sharing by keeping data close to the source, and instead sending analysis to the data. Technologically, PHT requires three interdependent components: "tracks" (protected communication channels), "trains" (containerized software applications), and "stations" (institutional data repositories), which are supported by the open source "Vantage6" software. The study applies this federated deep learning infrastructure to the use case of GTV segmentation on chest CT images of lung cancer patients, with the introduction of an additional component called the Secure Aggregations Server, where the model averaging is done in a trusted and inaccessible environment.

Results:

In this paper we demonstrated the feasibility of executing deep learning algorithms in a federated manner using PHT and presented the results from a proof-of-concept study. The infrastructure linked 12 hospitals across 8 nations, covering 4 continents, demonstrating the scalability and global reach of the proposed approach. In the entire execution and training of the deep learning algorithm, no data has been shared outside the hospital.

Conclusions:

The findings of the proof-of-concept study, as well as the implications and limitations of the infrastructure and the results, are discussed. The application of federated deep learning to unstructured medical imaging data, facilitated by the PHT framework and Vantage6 platform, represents a significant advancement in the field. The proposed infrastructure addresses the challenges of data privacy and enables collaborative model development, paving the way for the widespread adoption of deep learning-based tools in the medical domain and beyond. The introduction of the Secure Aggregation Server implied that data leakage problems in federated learning can be prevented by careful design decisions of the infrastructure. Clinical Trial: ARtificial Intelligence for Gross Tumour vOlume Segmentation (ARGOS) ClinicalTrials.gov ID NCT05775068 Sponsor Maastricht Radiation Oncology Information provided by Andre Dekker, Maastricht Radiation Oncology (Responsible Party) https://clinicaltrials.gov/study/NCT05775068