overall process of ASD-detection

ASD detection system built on the ABIDE neuroimaging dataset — multi-atlas feature extraction, tangent connectivity, and an MLP classifier achieving 76% accuracy and 0.818 AUC. Active development.

Machine learning model for early Autism Spectrum Disorder (ASD) screening using behavioral data and predictive analytics.

ASD early screening in children using ML & ANN — 9 classifiers (MLP-ANN, Random Forest, SVM, KNN, QDA, Naïve Bayes, LR), SMOTE, 7-metric evaluation, overfitting analysis. Streamlit web app deployment. MSc ISE thesis — Purbanchal University, Nepal.

Graduate clinical AI system for ASD detection from sMRI — NIfTI pipeline, Hybrid CNN-ViT (AUC 0.997), Grad-CAM + LIME explainability, MC-Dropout uncertainty, LLM clinical narrative, deployed on Hugging Face Spaces.

B.Tech 2023 capstone — CNN vs Vision Transformer for ASD detection from structural MRI (ABIDE-I, 1,067 subjects). Established CNN baseline (AUC 0.978) and documented ViT failure from scratch. Foundation for graduate Phase 2.

This project is a multi-modal Flask web app that uses quiz, image, and text-based machine learning models to assist in Autism Spectrum Disorder (ASD) detection.