TL;DR
The proposed VGG16-based approach has potential as a supportive, non-invasive tool for early ASD screening and is deployed as an interactive, Streamlit-based web application that allows users to upload facial images and receive real-time predictions.
Autism Spectrum Disorder (ASD) is a prevalent neurodevelopmental condition for which early screening is essential to support timely intervention. Conventional diagnostic procedures are often time-consuming and resource-intensive, highlighting the need for accessible and cost-effective screening approaches. This study investigates the use of deep learning–based facial image analysis as a supportive tool for ASD screening rather than as a diagnostic solution. A comparative evaluation of three state-of-the-art convolutional neural network architectures EfficientNetB0, MobileNetV2, and ResNet50 is conducted using a unified experimental framework based on transfer learning and fine-tuning. Experiments are performed on a publicly available Autistic Children Facial Image Dataset, with an additional verification step to prevent duplicate images in order to enhance generalization and reduce potential bias. All models are trained and evaluated under identical conditions to ensure a fair comparison. Experimental results indicate that EfficientNetB0 achieves the best overall performance, reaching a test accuracy of 88%, while maintaining a favorable balance between model complexity and generalization ability.
The need to develop large, well‐balanced datasets, the application of explainable AI techniques, standardization and regulatory guidelines for facilitating the clinical translation of ASD detection systems are suggested.
These findings demonstrate the potential of computer vision-based analysis of children’s expressive activities as an effective, non-invasive ASD pre-screening tool and will focus on expanding dataset diversity and integrating multimodal behavioral cues to improve model generalization and clinical applicability.
This study investigates the utilization of deep learning models to recognize ASD among 13-year-old children based on eye movement data collected as participants observed static images and short video sequences, highlighting the potential of deep learning frameworks as objective, data-driven tools for ASD detection in both clinical and research contexts.
Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that affects communication skills, social interaction, and behavioral patterns. Early detection is essential for timely intervention; however, conventional diagnostic methods remain time-consuming and subjective, as they rely heavily on clinical observations and expert judgment. This limitation highlights the need for an automated and objective approach to support early ASD screening. This study aims to analyze the performance, stability, and generalization of ASD classification using geometric features extracted from distances between facial landmarks. By representing facial morphology in terms of quantitative spatial relationships, this approach provides a more interpretable alternative to raw image-based methods. This study contributes by proposing a geometric feature representation based on facial landmark distances, providing a comparative analysis between linear and nonlinear classifiers, and ensuring robust evaluation through cross-validation. The dataset consists of 2,032 facial images, evenly distributed between children with ASD and those with typical development. A total of 68 facial landmark points were detected and used to compute pairwise Euclidean distances as classification features. Two classification algorithms, Logistic Regression and Extra Trees Classifier, were evaluated using 5-fold cross-validation to ensure reliable and unbiased performance estimation. The results show that Logistic Regression achieved an average accuracy of 89.91%, precision of 91.04%, recall of 88.56%, and F1-score of 89.76%. Meanwhile, the Extra Trees Classifier outperformed the linear model, achieving an average accuracy of 91.88%, precision of 92.69%, recall of 90.89%, and F1-score of 91.77%. Overall, both models demonstrated stable and consistent performance across validation folds, with the Extra Trees Classifier showing superior ability to capture nonlinear patterns in the data. These findings indicate that geometric feature extraction based on facial landmark distances is effective for ASD detection and has strong potential to be developed as an objective, interpretable, and efficient early screening tool using children’s facial images.
A comprehensive machine learning framework to classify ASD severity (mild, moderate, severe) is developed and validates by investigating the differential impact of feature engineering and selection, revealing a critical “evaluation paradox” where radical, unguided feature reduction improved geometric cluster cohesion but degraded clinical accuracy.