Perspectives: Comparison of Deep Learning Segmentation Models on Biophysical and Biomedical Data

TL;DR

This paper compares four deep learning segmentation models—CNNs, U-Nets, vision transformers, and vision state space models—on biophysical data, providing practical guidelines for selecting the best architecture based on dataset size and application needs.

Contribution

It offers a comprehensive comparison of popular deep learning models for segmentation in biophysics, highlighting their strengths and optimal conditions for use.

Findings

U-Nets perform best with small datasets.

Vision transformers excel with larger datasets.

Guidelines for model selection based on data and task.

Abstract

Deep learning based approaches are now widely used across biophysics to help automate a variety of tasks including image segmentation, feature selection, and deconvolution. However, the presence of multiple competing deep learning architectures, each with its own unique advantages and disadvantages, makes it challenging to select an architecture best suited for a specific application. As such, we present a comprehensive comparison of common models. Here, we focus on the task of segmentation assuming the typically small training dataset sizes available from biophysics experiments and compare the following four commonly used architectures: convolutional neural networks, U-Nets, vision transformers, and vision state space models. In doing so, we establish criteria for determining optimal conditions under which each model excels, thereby offering practical guidelines for researchers and practitioners in the field.