SFCNeXt: a simple fully convolutional network for effective brain age estimation with small sample size

Yu Fu,Yanyan Huang,Shunjie Dong,Yalin Wang,Tianbai Yu,Meng Niu,Cheng Zhuo

2023-05-30

Abstract:Deep neural networks (DNN) have been designed to predict the chronological age of a healthy brain from T1-weighted magnetic resonance images (T1 MRIs), and the predicted brain age could serve as a valuable biomarker for the early detection of development-related or aging-related disorders. Recent DNN models for brain age estimations usually rely too much on large sample sizes and complex network structures for multi-stage feature refinement. However, in clinical application scenarios, researchers usually cannot obtain thousands or tens of thousands of MRIs in each data center for thorough training of these complex models. This paper proposes a simple fully convolutional network (SFCNeXt) for brain age estimation in small-sized cohorts with biased age distributions. The SFCNeXt consists of Single Pathway Encoded ConvNeXt (SPEC) and Hybrid Ranking Loss (HRL), aiming to estimate brain ages in a lightweight way with a sufficient exploration of MRI, age, and ranking features of each batch of subjects. Experimental results demonstrate the superiority and efficiency of our approach.

Image and Video Processing,Computer Vision and Pattern Recognition,Machine Learning

What problem does this paper attempt to address?

The paper aims to address the following main issues:

1. **Brain age estimation with small sample sizes**: Existing deep learning models for brain age estimation typically require a large amount of training data (e.g., thousands or tens of thousands of MRI images), which is a challenge for clinical applications as it is often difficult to obtain such large-scale datasets.

2. **Complex network structures**: Many existing brain age estimation models have complex network structures, which not only increase the demand for computational resources but also limit the deployment and application of these models in real clinical scenarios.

To address the above issues, the authors propose a simple and efficient fully convolutional neural network architecture—SFCNeXt (Simple Fully Convolutional Network next), for brain age estimation tasks based on small-scale datasets. The model achieves its goals through the following two key technical components:

- **Single Pathway Encoded ConvNeXt (SPEC)**: This includes a main pathway that uses a four-stage ConvNeXt network to progressively downsample and aggregate 3D brain MRI features, and a branch pathway to integrate gender features. Finally, a Conformer encoder module is used for adaptive encoding of high-order features, and a multi-layer perceptron (MLP) is used for feature fusion to complete the brain age estimation.

- **Hybrid Ranking Loss (HRL)**: This combines mean squared error (MSE) loss and fast differentiable sorting loss to explore the soft ranking relationship between brain MRI features and true age within each batch, thereby improving the model's performance.

Experimental results show that SFCNeXt outperforms other lightweight baseline models on multiple evaluation metrics (such as mean absolute error (MAE), Pearson correlation coefficient (PCC), and Spearman rank correlation coefficient (SRCC)), while consuming comparable GPU resources, demonstrating the model's effectiveness and practicality.