Skip to content

lakshyeahh/Autism-Detection

Repository files navigation

Autism Detection Model

Overview

This project aims to develop a model for autism detection using a modified version of Simon's Game to assess specific parameters related to Sensory Perception, Attention to Detail, and Multitasking Abilities. These parameters are crucial in identifying traits associated with Autism Spectrum Disorder (ASD). The game provides metrics such as Average Reaction Time (in milliseconds) and Accuracy Rate (in percentage), which are mapped to the responses recorded from the child during gameplay.

Dataset Description

The model is trained on a dataset consisting of 17 data points, where responses to Autism Quotient (AQ) questions are labeled as either 0 (not close to autism) or 1 (close to autism). Specifically, questions AQ2, AQ4, AQ6, and AQ8 are identified as relevant to Sensory Perception, Attention to Detail, and Multitasking Abilities.

Game Implementation

The assessment game is based on Simon's Game, enhanced with multiple levels and features to measure reaction time and accuracy. The collected gameplay data is correlated with AQ parameters to predict the likelihood of autism traits.

Screenshot of Simon's Game

Threshold Values

The following threshold values were identified from research and data analysis to classify responses:

AQ Parameter Average Reaction Time (ms) Accuracy Rate (%) Classification
AQ2 2500 70 1 (Close to ASD) if below threshold values, otherwise 0
AQ4 2200 78 1 (Close to ASD) if below threshold values, otherwise 0
AQ6 2000 75 1 (Close to ASD) if below threshold values, otherwise 0
AQ8 1800 80 1 (Close to ASD) if below threshold values, otherwise 0

Usage

Screenshot of Simon Game Metrics

To use the model:

  • Ensure the game is set up to record Average Reaction Time and Accuracy Rate.
  • Feed the recorded metrics into the model to obtain predictions for AQ parameters.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published