Autism Spectrum Disorder Detection of Facial Images Using VGG Model

Autism Spectrum Disorder Detection of Facial Images Using VGG Model

V.Hemanth Kumar Reddy, B.Jaipal Reddy, K.Vamsi Krishna, Y.Gopinath Reddy,Dr.Koteswara Rao Anne

Department of Computer Science and Engineering Kalasalingam Academy of Research and Education Krishnankoil-626126,TamilNadu,

India

ABSTRACT:

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterized by challenges with social interaction, communication, and repetitive behaviors. Early detection and intervention are crucial for improving outcomes for individuals with ASD. In this study, we propose a machine learning approach for ASD prediction using convolutional neural networks (CNNs) and transfer learning. We leverage the VGG16 architecture, a pre-trained CNN model, to classify images into "Autistic" and "Non-Autistic" categories. Our pipeline involves data preprocessing, model construction, and training, incorporating techniques such as data augmentation and regularization to improve generalization performance. We evaluate the model's performance using metrics such as accuracy, precision, recall, and area under the receiver operating characteristic curve (AUC-ROC). Our findings demonstrate the potential of deep learning techniques for automated ASD prediction from image data, offering a promising avenue for early diagnosis and intervention strategies.

Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition characterized by a wide range of symptoms, including social communication deficits, repetitive behaviors, and sensory sensitivities. Early detection and intervention are critical for improving outcomes and facilitating appropriate support for individuals with ASD. In this study, we present a novel approach to ASD prediction leveraging deep learning techniques applied to image data.

Our methodology centers around the utilization of convolutional neural networks (CNNs), a class of deep learning models well-suited for image classification tasks. Specifically, we employ the VGG16 architecture, a widely-used CNN pre- trained on the ImageNet dataset, as the backbone of our

predictive model. Transfer learning is utilized to adapt the pre- trained VGG16 model to the task of ASD classification, allowing us to leverage the learned features from a diverse range of images.

The pipeline begins with comprehensive data preprocessing, including image resizing and normalization. We meticulously curate a dataset comprising images of individuals diagnosed with ASD ("Autistic") and neurotypical controls ("Non- Autistic"). Data augmentation techniques, such as rotation, shifting, and flipping, are employed to augment the training set, enhancing model generalization and robustness.

The constructed model consists of a VGG16 base followed by a custom classifier. We fine-tune the classifier layers while keeping the VGG16 base frozen to prevent overfitting and exploit the powerful features learned by the pre-trained network. Dropout regularization is incorporated to mitigate the risk of overfitting, enhancing the model's ability to generalize to unseen data.

The performance of our model is rigorously evaluated using a range of metrics, including accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic curve (AUC-ROC). Experimental results demonstrate the efficacy of our approach in accurately classifying individuals as "Autistic" or "Non-Autistic" based on image data.

Keywords – Autism Spectrum Disorder, Data Preprocessing, VGG16, Transfer Learning, Model Training,