Distributing Deep Learning Hyperparameter Tuning for 3D Medical Image Segmentation

TL;DR

This paper explores distributed deep learning training pipelines for 3D medical image segmentation, comparing data and experiment parallelism on multi-GPU and multi-node setups to enhance scalability and efficiency.

Contribution

It introduces a novel pipeline design for distributed training in 3D medical image segmentation, benchmarking data and experiment parallelism approaches on high-performance computing resources.

Findings

Experiment parallelism achieves better scalability than data parallelism.

Distributed training significantly reduces processing time for 3D segmentation tasks.

Open-source implementation facilitates adoption and adaptation in medical imaging research.

Abstract

Most research on novel techniques for 3D Medical Image Segmentation (MIS) is currently done using Deep Learning with GPU accelerators. The principal challenge of such technique is that a single input can easily cope computing resources, and require prohibitive amounts of time to be processed. Distribution of deep learning and scalability over computing devices is an actual need for progressing on such research field. Conventional distribution of neural networks consist in data parallelism, where data is scattered over resources (e.g., GPUs) to parallelize the training of the model. However, experiment parallelism is also an option, where different training processes are parallelized across resources. While the first option is much more common on 3D image segmentation, the second provides a pipeline design with less dependence among parallelized processes, allowing overhead reduction and more potential scalability. In this work we present a design for distributed deep learning training pipelines, focusing on multi-node and multi-GPU environments, where the two different distribution approaches are deployed and benchmarked. We take as proof of concept the 3D U-Net architecture, using the MSD Brain Tumor Segmentation dataset, a state-of-art problem in medical image segmentation with high computing and space requirements. Using the BSC MareNostrum supercomputer as benchmarking environment, we use TensorFlow and Ray as neural network training and experiment distribution platforms. We evaluate the experiment speed-up, showing the potential for scaling out on GPUs and nodes. Also comparing the different parallelism techniques, showing how experiment distribution leverages better such resources through scaling. Finally, we provide the implementation of the design open to the community, and the non-trivial steps and methodology for adapting and deploying a MIS case as the here presented.