Enhancing Autism Spectrum Disorder Diagnosis through Behavioral Pattern Recognition with Deep Learning Approaches
Syed Farzana, Ramkumar DevendiranIntroduction:
Autism Spectrum Disorder (ASD) diagnosis increasingly benefits from automated behavioral analysis, particularly for identifying self-stimulatory behaviors that are critical clinical indicators. This proposed patent-oriented methodology presents a deep learning-based behavioral pattern recognition framework designed to detect and classify self-stimulatory actions using a curated Self-Stimulatory Behaviour Dataset. The proposed model learns discriminative temporal and spatial behavioral features to improve diagnostic reliability. Experimental evaluation demonstrates strong classification performance, achieving an accuracy of 95.8%, precision of 94.9%, recall of 95.2%, and an F1-score of 95.0%. The results indicate robust generalization across behavioral variations and highlight the framework’s ability to distinguish subtle repetitive actions associated with ASD. These findings support the use of patent-driven deep learning behavioral analytics as a promising assistive tool for early screening and objective assessment in clinical environments.
Methods:
A Convolutional Neural Network (CNN) architecture was developed and trained using the SSBD dataset, which contains comprehensive behavioural data of individuals with and without ASD. The model was designed to identify subtle behavioural cues and non-linear relationships that may not be evident through traditional assessment methods. Performance metrics, including accuracy, precision, recall, and F1-score, were computed and compared with results from conventional diagnostic approaches.
Results:
The proposed CNN model demonstrated a notable improvement in diagnostic accuracy and efficiency over traditional clinical methods. The model effectively recognized distinctive behavioural indicators associated with ASD, achieving high classification performance across all evaluation metrics. The deep learning approach successfully captured complex behavioural dependencies, minimizing diagnostic subjectivity and variability.
Discussion:
The findings indicate that deep learning-based behavioural analysis can serve as a robust and scalable alternative to manual diagnostic assessments. By leveraging large-scale behavioural datasets, the model offers clinicians data-driven insights, enabling earlier and more objective detection of ASD. This approach also highlights the potential of artificial intelligence in bridging existing gaps in neurodevelopmental diagnostics.
Conclusion:
This study presents a novel CNN-based framework for automated ASD diagnosis using behavioural pattern recognition. The model’s superior accuracy and efficiency suggest its potential for clinical integration, allowing earlier interventions and improved therapeutic outcomes for individuals with ASD. Future research will explore model generalization across diverse populations and real-time behavioural monitoring systems.