Detection of Autism Spectrum Disorder

🏷️ Early Detection, Empowered Futures: Leveraging AI for Autism Spectrum Disorder Diagnosis

Autism Spectrum Disorder (ASD) Detection is a machine learning-based approach that analyzes images and structured data to identify ASD traits, aiding in early diagnosis and assessment. ASD is a developmental disorder affecting communication and behavior. It is developed as part of Mini Project II during the VI semester-BE at Sri Siddhartha Institute of Technology, Tumakuru, focusing on image-based and structured-data-based approaches for ASD diagnosis using machine learning. Early detection is crucial for timely intervention, which can significantly improve the quality of life for individuals with ASD. This project aims to develop a system for ASD detection using two different approaches:

1. Image-based detection: Utilizes deep learning techniques to analyze facial features from images to identify ASD characteristics. This approach leverages Convolutional Neural Networks (CNNs) to extract patterns that may indicate ASD.
2. CSV-based detection: Uses structured data containing behavioral and demographic attributes to classify ASD cases. Machine learning models are applied to analyze key features and make predictions based on clinical data.

By comparing both methods, this project provides insights into the effectiveness of image-based and structured data-based approaches in ASD diagnosis. The goal is to explore how artificial intelligence can assist in ASD detection and potentially support healthcare professionals in their assessments.

✅ Mini-Project II at SSIT | Project Cycle Closed | Official Guideline

PROJECT DOCUMENTATION, you can find the complete project documentation here.

PROJECT PRESENTATION, you can find the complete project presentation here.

Methodology :

1. Image-based Approach :

CODE

• Preprocessing of images (resizing, normalization, augmentation)
• Model: Convolutional Neural Networks (CNNs) trained on ASD image datasets
• Evaluation: Accuracy, precision, recall, and F1-score

IMAGE DATASET : Contains unlabeled images for ASD and non-ASD individuals.

2. CSV-based Approach :

CODE

• Data preprocessing (handling missing values, encoding categorical variables, feature selection)
• Model: Machine Learning classifiers (Logistic Regression, XGBClassifier, SVC)
• Evaluation: Performance metrics and validation techniques

CSV DATASET : Tabular data with features relevant to ASD diagnosis.

Technologies Used :

1. Programming Languages: Python

2. Machine Learning & Deep Learning: TensorFlow, Keras, Scikit-learn

3. Data Processing & Visualization: Pandas, NumPy, Matplotlib, Seaborn

4. Image Processing: OpenCV, PIL

5. Model Deployment (Optional): Flask, Streamlit

Team Members :

1. Neha Acharya

2. Prathuasha K B*

Suggestions and project improvement are invited!

Prathuasha K B