Abstract:Dose prediction plays a key role in knowledge-based planning (KBP) by automatically generating patient-specific dose distribution. Recent advances in deep learning-based dose prediction methods necessitates collaboration among data contributors for improved performance. Federated learning (FL) has emerged as a solution, enabling medical centers to jointly train deep-learning models without compromising patient data privacy. We developed the FedKBP framework to evaluate the performances of centralized, federated, and individual (i.e. separated) training of dose prediction model on the 340 plans from OpenKBP dataset. To simulate FL and individual training, we divided the data into 8 training sites. To evaluate the effect of inter-site data variation on model training, we implemented two types of case distributions: 1) Independent and identically distributed (IID), where the training and validating cases were evenly divided among the 8 sites, and 2) non-IID, where some sites have more cases than others. The results show FL consistently outperforms individual training on both model optimization speed and out-of-sample testing scores, highlighting the advantage of FL over individual training. Under IID data division, FL shows comparable performance to centralized training, underscoring FL as a promising alternative to traditional pooled-data training. Under non-IID division, larger sites outperformed smaller sites by up to 19% on testing scores, confirming the need of collaboration among data owners to achieve better prediction accuracy. Meanwhile, non-IID FL showed reduced performance as compared to IID FL, posing the need for more sophisticated FL method beyond mere model averaging to handle data variation among participating sites.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in radiotherapy, how to improve the efficiency and accuracy of cancer treatment plans through dose prediction in knowledge - based planning (KBP). Specifically, the article aims to address the following challenges: 1. **Data Limitations**: The application of deep - learning methods in dose prediction is restricted by the lack of training data. 2. **Data Privacy Protection**: In order to collaborate without violating patient privacy, a method that can jointly train models among different medical centers is required. To this end, the authors propose a framework named FedKBP, which utilizes federated learning (FL) technology to enable multiple medical centers to jointly train dose - prediction models without sharing patient data. The main purpose of the study is to evaluate the performance of centralized, federated, and independent training methods on the same dataset and explore the impact of data distribution (independent and identically distributed (IID) and non - IID) on model performance. ### Specific Problem Description - **Dose Prediction Requirements**: Knowledge - based planning (KBP) introduces the function of automatically estimating patient - specific dose distributions to simplify the evaluation and optimization process of treatment plans. - **Data Limitations and Privacy Issues**: Traditional methods require a large amount of high - quality data, but this data is scattered among different medical institutions and involves privacy issues, making it difficult to share directly. - **Application of Federated Learning**: Federated learning, as a solution, allows multiple institutions to collaboratively train models without sharing raw data, thereby protecting patient privacy and improving the generalization ability of the model. ### Research Objectives By developing and evaluating the FedKBP framework, verify the effectiveness of federated learning in dose - prediction tasks, especially its performance under different data - distribution conditions, and provide theoretical basis and technical support for future clinical applications. ### Main Contributions - Proposed and implemented a federated - learning framework FedKBP for dose prediction. - Compared the performance of centralized, federated, and independent training methods in dose - prediction tasks. - Explored the impact of data distribution (IID and non - IID) on the performance of federated - learning models. In summary, this paper aims to explore the application potential of federated learning in radiotherapy dose prediction to solve data - limitation and privacy - protection problems, thereby improving the quality and efficiency of treatment plans.

FedKBP: Federated dose prediction framework for knowledge-based planning in radiation therapy

FedDGP: Disentangling Global and Personal Models for Federated Learning

Knowledge-based Radiation Treatment Planning: A Data-driven Method Survey

pFedKT: Personalized federated learning with dual knowledge transfer

Predicting treatment response in multicenter non-small cell lung cancer patients based on federated learning

FedBN: Federated Learning on Non-IID Features via Local Batch Normalization

Personalized Federated Learning with Adaptive Feature Aggregation and Knowledge Transfer

A Hierarchical Knowledge Transfer Framework for Heterogeneous Federated Learning.

Personalized Federated Learning with Adaptive Batchnorm for Healthcare

Handling Data Heterogeneity in Federated Learning via Knowledge Distillation and Fusion

Adaptive Personlization in Federated Learning for Highly Non-i.i.d. Data

Fine-tuning Global Model Via Data-Free Knowledge Distillation for Non-IID Federated Learning

Multi-stage framework with difficulty-aware learning for progressive dose prediction

FKD-Med: Privacy-Aware, Communication-Optimized Medical Image Segmentation via Federated Learning and Model Lightweighting Through Knowledge Distillation

FedGK: Communication-Efficient Federated Learning through Group-Guided Knowledge Distillation

PRIOR: Personalized Prior for Reactivating the Information Overlooked in Federated Learning

OpenKBP: The open-access knowledge-based planning grand challenge

Communication-Efficient and Privacy-Preserving Aggregation in Federated Learning With Adaptability

The Best of Both Worlds: Accurate Global and Personalized Models through Federated Learning with Data-Free Hyper-Knowledge Distillation

A Personalized Federated Learning Method Based on Clustering and Knowledge Distillation

One-shot Federated Learning on Medical Data using Knowledge Distillation with Image Synthesis and Client Model Adaptation