Towards Equitable ASD Diagnostics: A Comparative Study of Machine and Deep Learning Models Using Behavioral and Facial Data

TL;DR

These findings suggest Random Forest's high accuracy and balanced precision-recall metrics could enhance clinical workflows and MobileNet's lightweight structure shows promise for resource-limited environments, enabling accessible ASD screening.

Abstract

Autism Spectrum Disorder (ASD) is often underdiagnosed in females due to gender-specific symptom differences overlooked by conventional diagnostics. This study evaluates machine learning models, particularly Random Forest and convolutional neural networks, for enhancing ASD diagnosis through structured data and facial image analysis. Random Forest achieved 100% validation accuracy across datasets, highlighting its ability to manage complex relationships and reduce false negatives, which is crucial for early intervention and addressing gender biases. In image-based analysis, MobileNet outperformed the baseline CNN, achieving 87% accuracy, though a 30% validation loss suggests possible overfitting, requiring further optimization for robustness in clinical settings. Future work will emphasize hyperparameter tuning, regularization, and transfer learning. Integrating behavioral data with facial analysis could improve diagnosis for underdiagnosed groups. These findings suggest Random Forest's high accuracy and balanced precision-recall metrics could enhance clinical workflows. MobileNet's lightweight structure also shows promise for resource-limited environments, enabling accessible ASD screening. Addressing model explainability and clinician trust will be vital.

Figures (11)

• Figure 1: Visual representation of the key characteristics associated with ASD, highlighting the core aspects such as social interaction difficulties, repetitive behaviors, restricted interests, language delays, eye contact avoidance, communication difficulties, sensory sensitivities, nonverbal challenges, emotional regulation, and need for routine. The icons around the central figure illustrate the common behavioral and sensory traits often associated with ASD.
• Figure 2: Typical CNN architecture.
• Figure 3: Baseline CNN results with SGD optimizer showing combined Training Loss, Validation Loss, Training Accuracy, and Validation Accuracy over 15 epochs.
• Figure 4: Loss results from the second test run for the baseline CNN with the Adam optimizer. The green line represents the validation loss.
• Figure 5: Accuracy results from the second test run for the baseline CNN with the Adam optimizer.