Machine Learning and Deep Learning for Autism Spectrum Disorder (ASD) detection / Nour elhouda Mahdaoui

Public ISBD Titre : Machine Learning and Deep Learning for Autism Spectrum Disorder (ASD) detection Type de document : document électronique Auteurs : Nour elhouda Mahdaoui ; kaouther Belkefoul, Auteur ; Semcheddine,Moussa, Directeur de thèse Editeur : Setif:UFA Année de publication : 2025 Importance : 1 vol (75 f .) Format : 29 cm Langues : Anglais (eng) Catégories : Thèses & Mémoires:Informatique Mots-clés : Autism Spectrum Disorder Early detection Deep learning Facial images Movement Index. décimale : 004 Informatique Résumé : Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by difficulties with social interaction, communication, and behavior. Early identification ensures timely intervention and improves quality of life. The present study proposes a two-modality binary classification approach for ASD detection based on facial image data and movement data. The image dataset consists of facial photos of children labeled as ASD or non-ASD. The movement data, captured using a Kinect v2 sensor, comprises 1,259 features per subject derived from 3D joint positions and gait metrics, with a total of 800 samples. A DenseNet121 convolutional neural network (CNN) was employed for images, achieving 89% accuracy. Grad-CAM was used to provide visual explanations by highlighting important regions in the images. For the movement modality, a Multi-Layer Perceptron (MLP) trained on features learned via an autoencoder achieved 99.38% accuracy, with Shapley Additive exPlanations (SHAP) applied to identify key features influencing model decisions. Finally, a late fusion mechanism combining both models was evaluated, resulting in 88.12% accuracy. Results highlight the effectiveness of unimodal solutions, particularly the MLP with autoencoder, and indicate that multimodal fusion requires further optimization for enhanced overall performance Note de contenu : Sommaire List of Figures iv List of Tables v List of abbreviations vi General Introduction 1 Chapter 1 : Autism spectrum disorder (ASD) 3 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1 Definition and overview . . . . . . . . . . . . . . . . . . . . . . . . 4 1.1 Historical perspective on autism . . . . . . . . . . . . . . . . . 5 2 Prevalence and epidemiology of ASD . . . . . . . . . . . . . . . . . 6 3 Causes and risks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3.1 Genetic factors . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3.2 Environmental risk factors . . . . . . . . . . . . . . . . . . . . 7 3.3 Parental age and premature birth . . . . . . . . . . . . . . . . . 7 4 Signs and symptoms of autism spectrum disorder (ASD) . . . . . . . 8 5 Diagnosis of autism spectrum disorder . . . . . . . . . . . . . . . . . 9 6 Importance of early diagnosis . . . . . . . . . . . . . . . . . . . . . 11 7 Comorbidities of autism spectrum disorder (ASD) . . . . . . . . . . . 11 7.1 Common comorbidities . . . . . . . . . . . . . . . . . . . . . 11 7.2 Psychiatric conditions . . . . . . . . . . . . . . . . . . . . . . 12 8 Interventions and treatments for ASD . . . . . . . . . . . . . . . . . 12 9 The role of technology in autism spectrum disorder . . . . . . . . . . 13 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Chapter 2 : Background and Literature Review 15 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1 Machine learning (ML) . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.1 Definition and basic concepts of ML . . . . . . . . . . . . . . 16 1.2 Types of machine learning . . . . . . . . . . . . . . . . . . . . 17 2 Deep learning (DL) . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.1 Definition and basic concepts of DL . . . . . . . . . . . . . . . 19 2.2 Evolution of deep learning models . . . . . . . . . . . . . . . 20 2.3 Types of artificial neural networks . . . . . . . . . . . . . . . . 21 2.4 Evaluation metrics for deep learning and machine learning models 25 2.5 Deep learning vs. traditional machine learning . . . . . . . . . 26 3 ML and DL for ASD detection . . . . . . . . . . . . . . . . . . . . . 26 3.1 Advantages of ML/DL over traditional methods . . . . . . . . 27 3.2 Analyzing complex patterns in large datasets . . . . . . . . . . 27 4 Machine learning techniques for ASD detection . . . . . . . . . . . . 27 4.1 Neuroimaging-based approaches . . . . . . . . . . . . . . . . 28 4.2 Behavioral data and feature-based models . . . . . . . . . . . . 30 4.3 Hybrid and ensemble learning approaches . . . . . . . . . . . 31 5 Deep learning techniques for ASD detection . . . . . . . . . . . . . . 32 5.1 Convolutional neural networks for ASD classification . . . . . 33 5.2 Transfer learning for ASD detection using facial features . . . . 34 5.3 Hybrid and multi-modal learning approaches . . . . . . . . . . 36 6 Data sources for machine learning and deep learning in autism spectrum disorder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 7 Challenges and limitations in ML/DL for ASD detection . . . . . . . 37 7.1 Unavailable MRI neuroimaging datasets of ASD patients . . . 38 7.2 Challenges in artificial intelligence algorithms in diagnosing autism spectrum disorder . . . . . . . . . . . . . . . . . . . . 38 7.3 Challenges in hardware . . . . . . . . . . . . . . . . . . . . . 38 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Chapter3 : Contribution,experiments and results 40 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 1 Development environment . . . . . . . . . . . . . . . . . . . . . . . 41 1.1 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 1.2 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 2 Proposed methodology . . . . . . . . . . . . . . . . . . . . . . . . . 43 2.1 Phase 1: Motion data analysis . . . . . . . . . . . . . . . . . . 43 2.2 Phase 2: Analysis of image data using deep learning . . . . . . 47 2.3 Phase 3: Multimodal learning via late fusion . . . . . . . . . . 51 3 Validation Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 3.1 Dataset Preparation . . . . . . . . . . . . . . . . . . . . . . . 53 3.2 Model Architectures . . . . . . . . . . . . . . . . . . . . . . . 53 3.3 Experimental Settings . . . . . . . . . . . . . . . . . . . . . . 56 3.4 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . 57 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Conclusion and future work 60 Bibliographie 75 Côte titre : MAI/0965 Machine Learning and Deep Learning for Autism Spectrum Disorder (ASD) detection [document électronique] / Nour elhouda Mahdaoui ; kaouther Belkefoul, Auteur ; Semcheddine,Moussa, Directeur de thèse . - [S.l.] : Setif:UFA, 2025 . - 1 vol (75 f .) ; 29 cm. Langues : Anglais (eng) Catégories : Thèses & Mémoires:Informatique Mots-clés : Autism Spectrum Disorder Early detection Deep learning Facial images Movement Index. décimale : 004 Informatique Résumé : Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by difficulties with social interaction, communication, and behavior. Early identification ensures timely intervention and improves quality of life. The present study proposes a two-modality binary classification approach for ASD detection based on facial image data and movement data. The image dataset consists of facial photos of children labeled as ASD or non-ASD. The movement data, captured using a Kinect v2 sensor, comprises 1,259 features per subject derived from 3D joint positions and gait metrics, with a total of 800 samples. A DenseNet121 convolutional neural network (CNN) was employed for images, achieving 89% accuracy. Grad-CAM was used to provide visual explanations by highlighting important regions in the images. For the movement modality, a Multi-Layer Perceptron (MLP) trained on features learned via an autoencoder achieved 99.38% accuracy, with Shapley Additive exPlanations (SHAP) applied to identify key features influencing model decisions. Finally, a late fusion mechanism combining both models was evaluated, resulting in 88.12% accuracy. Results highlight the effectiveness of unimodal solutions, particularly the MLP with autoencoder, and indicate that multimodal fusion requires further optimization for enhanced overall performance Note de contenu : Sommaire List of Figures iv List of Tables v List of abbreviations vi General Introduction 1 Chapter 1 : Autism spectrum disorder (ASD) 3 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1 Definition and overview . . . . . . . . . . . . . . . . . . . . . . . . 4 1.1 Historical perspective on autism . . . . . . . . . . . . . . . . . 5 2 Prevalence and epidemiology of ASD . . . . . . . . . . . . . . . . . 6 3 Causes and risks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3.1 Genetic factors . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3.2 Environmental risk factors . . . . . . . . . . . . . . . . . . . . 7 3.3 Parental age and premature birth . . . . . . . . . . . . . . . . . 7 4 Signs and symptoms of autism spectrum disorder (ASD) . . . . . . . 8 5 Diagnosis of autism spectrum disorder . . . . . . . . . . . . . . . . . 9 6 Importance of early diagnosis . . . . . . . . . . . . . . . . . . . . . 11 7 Comorbidities of autism spectrum disorder (ASD) . . . . . . . . . . . 11 7.1 Common comorbidities . . . . . . . . . . . . . . . . . . . . . 11 7.2 Psychiatric conditions . . . . . . . . . . . . . . . . . . . . . . 12 8 Interventions and treatments for ASD . . . . . . . . . . . . . . . . . 12 9 The role of technology in autism spectrum disorder . . . . . . . . . . 13 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Chapter 2 : Background and Literature Review 15 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1 Machine learning (ML) . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.1 Definition and basic concepts of ML . . . . . . . . . . . . . . 16 1.2 Types of machine learning . . . . . . . . . . . . . . . . . . . . 17 2 Deep learning (DL) . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.1 Definition and basic concepts of DL . . . . . . . . . . . . . . . 19 2.2 Evolution of deep learning models . . . . . . . . . . . . . . . 20 2.3 Types of artificial neural networks . . . . . . . . . . . . . . . . 21 2.4 Evaluation metrics for deep learning and machine learning models 25 2.5 Deep learning vs. traditional machine learning . . . . . . . . . 26 3 ML and DL for ASD detection . . . . . . . . . . . . . . . . . . . . . 26 3.1 Advantages of ML/DL over traditional methods . . . . . . . . 27 3.2 Analyzing complex patterns in large datasets . . . . . . . . . . 27 4 Machine learning techniques for ASD detection . . . . . . . . . . . . 27 4.1 Neuroimaging-based approaches . . . . . . . . . . . . . . . . 28 4.2 Behavioral data and feature-based models . . . . . . . . . . . . 30 4.3 Hybrid and ensemble learning approaches . . . . . . . . . . . 31 5 Deep learning techniques for ASD detection . . . . . . . . . . . . . . 32 5.1 Convolutional neural networks for ASD classification . . . . . 33 5.2 Transfer learning for ASD detection using facial features . . . . 34 5.3 Hybrid and multi-modal learning approaches . . . . . . . . . . 36 6 Data sources for machine learning and deep learning in autism spectrum disorder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 7 Challenges and limitations in ML/DL for ASD detection . . . . . . . 37 7.1 Unavailable MRI neuroimaging datasets of ASD patients . . . 38 7.2 Challenges in artificial intelligence algorithms in diagnosing autism spectrum disorder . . . . . . . . . . . . . . . . . . . . 38 7.3 Challenges in hardware . . . . . . . . . . . . . . . . . . . . . 38 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Chapter3 : Contribution,experiments and results 40 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 1 Development environment . . . . . . . . . . . . . . . . . . . . . . . 41 1.1 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 1.2 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 2 Proposed methodology . . . . . . . . . . . . . . . . . . . . . . . . . 43 2.1 Phase 1: Motion data analysis . . . . . . . . . . . . . . . . . . 43 2.2 Phase 2: Analysis of image data using deep learning . . . . . . 47 2.3 Phase 3: Multimodal learning via late fusion . . . . . . . . . . 51 3 Validation Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 3.1 Dataset Preparation . . . . . . . . . . . . . . . . . . . . . . . 53 3.2 Model Architectures . . . . . . . . . . . . . . . . . . . . . . . 53 3.3 Experimental Settings . . . . . . . . . . . . . . . . . . . . . . 56 3.4 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . 57 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Conclusion and future work 60 Bibliographie 75 Côte titre : MAI/0965

Exemplaires (1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
MAI/0965	MAI/0965	Mémoire	Bibliothèque des sciences	Anglais	Disponible Disponible

---

1 (1 - 1 / 1)