Lightweight Brain Tumor Segmentation on Low-Resource Systems: A Step-by-Step Guide with 3D U-Net

Ayomide B. Oladele

Keywords: Brain Tumor Segmentation, AI Democratization, BraTS-Africa, CPU-only de-ployment, Resource-Constrained Settings, SPARK Academy.

TL;DR: Specifically designed for computers with limited computing resources, this tutorial demonstrates how to achieve promising results on a CPU, making deep learning accessible to a wider range of learners and users.

Abstract:

The application of deep learning methods in the healthcare has gained significant popularity and relevance among researchers and consumers in clinical, academic, and industry settings, leading to impactful discoveries that are improving human health. One such application is in automated brain tumor segmentation, which aids in the precise identification of tumor regions on magnetic resonance imaging (MRI) scans for accurate diagnosis, treatment planning, and prognosis.

However, the implementation of conventional deep learning models for this task often requires high computational resources, limiting their use by researchers in resource-constrained settings. This computational burden also limits skills training in deep learning methods in resource-constrained settings.

This tutorial presents an approach to address the need for high computing resources in deep learning methods development. It provides a step-by-step guide to developing a lightweight brain tumor segmentation model using a 3D U-Net architecture, optimized for low-resource systems. Using the Brain Tumor Segmentation (BraTS) in Sub-Sharan African Population (BraTS-Africa) 2024 dataset, the architecture was trained efficiently and evaluated on standard CPUs, without relying on GPUs. The approach taken in this tutorial seeks to balance computational efficiency with segmentation accuracy. The lightweight model achieved a Dice score of 0.67% on the validation data, and the segmentation output was visually compared with the ground truth. Despite being trained on low computing resources, the model showed promising results.

The main objective of this tutorial is to empower researchers in resource-constrained settings to learn how to develop, validate and deploy deep learning methods using existing frameworks and without reliance on expensive computational resources such as GPUs. More importantly, the tutorial will enable a wider audience to gain practical AI skills, facilitating development of local relevant tools for early detection, ultimately improving patient outcomes.

Submission Number: 6