Related Experiment Video
Updated: May 21, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
932
Capsule DenseNet++: Enhanced autism detection framework with deep learning and reinforcement learning-based lifestyle
Ahmed Ibrahim Alutaibi1, Sunil Kumar Sharma2, Ahmad Raza Khan3
1Department of Computer Engineering, College of Computer and Information Sciences, Majmaah University, Majmaah, 11952, Saudi Arabia.
Computers in Biology and Medicine
|March 22, 2025
Summary
This study introduces a novel deep learning framework for early Autism Spectrum Disorder (ASD) detection and personalized lifestyle recommendations. The advanced system achieves high accuracy, aiding timely diagnosis and intervention for children.
Area of Science:
- Neuroscience and Artificial Intelligence
- Developmental Pediatrics
- Machine Learning in Healthcare
Background:
- Autism Spectrum Disorder (ASD) diagnosis relies on subjective methods, impacting cost-effectiveness and outcome uniformity.
- Early identification and tailored interventions are crucial for improving developmental outcomes in children with ASD.
- Increasing ASD prevalence worldwide, particularly in Saudi Arabia, necessitates improved diagnostic tools.
Purpose of the Study:
- To develop and evaluate a two-phase deep learning framework for accurate ASD detection.
- To implement a personalized lifestyle recommendation system using reinforcement learning for ASD management.
- To enhance the efficiency and interpretability of ASD diagnostic approaches.
Main Methods:
- Feature extraction using multiscale statistical techniques and the CosmoNest Optimizer (a hybrid of African Vultures and Butterfly Optimization Algorithms).
- Classification of optimized features using Capsule DenseNet++ for efficient and interpretable ASD identification.
- Personalized lifestyle recommendations via Proximal Policy Optimization (PPO), a reinforcement learning algorithm, for adaptive interventions.
Main Results:
- The deep learning framework demonstrated high performance in ASD detection across two datasets.
- Achieved accuracy rates of 99.2% and 99.3%, precision of 98.5% and 98.7%, sensitivity of 98.7% and 98.9%, and F1-scores of 99.1% and 99.2%.
- The PPO algorithm dynamically adapted lifestyle recommendations for optimized interventions.
Conclusions:
- The proposed two-phase deep learning framework offers a robust and accurate solution for Autism Spectrum Disorder detection.
- The integration of reinforcement learning provides personalized and adaptive lifestyle recommendations, enhancing ASD management.
- This framework shows significant potential for improving early diagnosis and intervention strategies for ASD globally.
