Neural networks : the official journal of the International Neural Network Society · 2025
Explainable machine learning algorithm for classifying resting-state functional MRI in amyotrophic lateral sclerosis.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Shimano Kaoru, Hattori Takaaki, Yasuda Eiji, Hase Takeshi, Kina Satoko, Miyake Toru, Yokota Takanori
Аннотация
BACKGROUND: Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease that affects multiple brain systems. Altered brain function can be observed through resting-state functional magnetic resonance imaging (rs-fMRI). While machine learning offers significant advantages in capturing complex signal patterns across numerous voxels, its decision-making process often lacks transparency. This study aimed to develop an explainable machine learning pipeline to classify patients with ALS and healthy control (HC) using rs-fMRI data.
METHODS: Thirty patients with ALS and 30 HCs were enrolled. The pipeline consisted of three key components: (1) preprocessing of rs-fMRI data using independent component analysis, followed by dual regression to reduce dimensionality and generate individual network maps; (2) training of a three-dimensional convolutional neural network (3D-CNN) to classify each individual image as either ALS or HC; and (3) application of saliency map and Grad-CAM++ to visualize the reasoning behind the model's classification.
RESULTS: The 3D-CNN achieved high classification accuracy using the sensorimotor network (SMN) map (78.3%) and the visual network (VN) map (83.3%). Simultaneously, saliency map and Grad-CAM++ highlighted brain regions that contributed to the classification, and some of which were consistent with regions showing intergroup differences in the dual regression analysis.
DISCUSSION: This study developed a novel explainable machine learning model capable of extracting features and classifying rs-fMRI data. Our results showed altered functional integrity in the SMN and VN in ALS. Our pipeline holds the potential to extract features of rs-fMRI data, enabling classification of neurological diseases with explainability.
METHODS: Thirty patients with ALS and 30 HCs were enrolled. The pipeline consisted of three key components: (1) preprocessing of rs-fMRI data using independent component analysis, followed by dual regression to reduce dimensionality and generate individual network maps; (2) training of a three-dimensional convolutional neural network (3D-CNN) to classify each individual image as either ALS or HC; and (3) application of saliency map and Grad-CAM++ to visualize the reasoning behind the model's classification.
RESULTS: The 3D-CNN achieved high classification accuracy using the sensorimotor network (SMN) map (78.3%) and the visual network (VN) map (83.3%). Simultaneously, saliency map and Grad-CAM++ highlighted brain regions that contributed to the classification, and some of which were consistent with regions showing intergroup differences in the dual regression analysis.
DISCUSSION: This study developed a novel explainable machine learning model capable of extracting features and classifying rs-fMRI data. Our results showed altered functional integrity in the SMN and VN in ALS. Our pipeline holds the potential to extract features of rs-fMRI data, enabling classification of neurological diseases with explainability.