A robust DWT–CNN‐based CAD system for early diagnosis of autism using task‐based fMRI

Abstract

Task-based fMRI (TfMRI) is a diagnostic imaging modality for observing the effects of a disease or other condition on the functional activity of the brain. Autism spectrum disorder (ASD) is a pervasive developmental disorder associated with impairments in social and linguistic abilities. Machine learning algorithms have been widely utilized for brain imaging aiming for objective ASD diagnostics. Recently, deep learning methods have been gaining more attention for fMRI classification. The goal of this paper is to develop a convolutional neural network (CNN)-based framework to help in global diagnosis of ASD using TfMRI data that are collected from a response to speech experiment. Methods: To achieve this goal, the proposed framework adopts a novel imaging marker integrating both spatial and temporal information that are related to the functional activity of the brain. The developed pipeline consists of three main components. In the first step, the collected TfMRI data are preprocessed and parcellated using the Harvard-Oxford probabilistic atlas included with the fMRIB Software Library (FSL). Second, a group analysis using FSL is performed between ASD and typically developing (TD) children to identify significantly activated brain areas in response to the speech task. In order to reduce brain spatial dimensionality, a K-means clustering technique is performed on such significant brain areas. Informative blood oxygen level-dependent (BOLD) signals are extracted from each cluster. A compression step for each extracted BOLD signal using discrete wavelet transform (DWT) has been proposed. The adopted wavelets are similar to the expected hemodynamic response which enables DWT to compress the BOLD signal while highlighting its activation information. Finally, a deep learning 2D CNN network is used to classify the patients as ASD or TD based on extracted features from the previous step. Results: Preliminary results on 100 TfMRI dataset (50 ASD, 50 TD) obtain 80% correct global classification using tenfold cross validation (with sensitivity = 84%, specificity = 76%). Conclusion:The experimental results show the high accuracy of the proposed framework and hold promise for the presented framework as a helpful adjunct to currently used ASD diagnostic tools.

A Review on Autism Spectrum Disorder Diagnosis Using Task-Based Functional Mri

“…Inference based on brain images of children and adults can be generalized to young children and toddlers to a limited extent. Therefore, studies in [67][68][69] established their analysis and classification on toddlers with ages in the range of 1 to 3 years old to pursue more reliable and stable diagnosis. Since fMRI is a high dimensional 4D data with too many data points, feature extraction and reduction techniques are very crucial before developing a classifier, especially traditional ML classifiers.…”

“…Fortunately, the spatial dimension representing the whole brain contains many redundant data shared in each brain area. Therefore, many studies perform ROI selection [60][64] [69]. Either when ROI selection is performed or not, features are selected from each brain area by average, histogram, k-means clustering or down sampling with bootstrapping.…”

“…They begin their pipeline with selecting ROI based on literature or GLM group analysis as a feature reduction step [64][69]. Some studies included datasets from public shareable data repository with a license such as NDAR [67][68][69]. Other studies collected data from their labs.…”

“…After applying several classifiers, best accuracy of 78% was achieved by RF. The most recent research in ASD diagnosis with TfMRI was proposed by [69]. They developed a comprehensive feature extraction framework for both temporal and spatial feature extraction.…”

ASD-SAENet: A Sparse Autoencoder, and Deep-Neural Network Model for Detecting Autism Spectrum Disorder (ASD) Using fMRI Data

“…In fMRI studies, functional connectivity is based on the correlation of the activation time series in pairs of brain areas and are studied for both ASD and healthy brains. However, it is not possible to detect subtle biomarker patterns using conventional computational and statistical methods Haweel et al, 2020;Nogay and Adeli, 2020). Machine learning algorithms have been successful in identifying biomarkers from functional Magnetic Resonance Imaging (fMRI) datasets for biomarker discovery, and classification of various brain disorders (Deshpande et al, 2015;Sarraf and Tofighi, 2016;Dvornek et al, 2017;Saeed, 2018, 2019;El-Gazzar et al, 2019a;Yao and Lu, 2019).…”