# Skeleton-based action analysis for ADHD diagnosis

**Authors:** Yichun Li, Yi Li, Rajesh Nair, Syed Mohsen Naqvi  
**arXiv:** [2304.09751](https://arxiv.org/abs/2304.09751)  
**Date:** 2023-04-14  
**Summary page:** https://www.summarizepaper.com/en/paper/2304.09751v1/

## Key points

- ADHD is a prevalent neurobehavioral disorder that affects individuals worldwide
- Most machine learning methods for ADHD diagnosis rely on expensive equipment, making them inaccessible to many patients
- There is a need for low-cost diagnostic methods based on the action characteristics of ADHD
- Skeleton-based action recognition has gained attention due to its action-focused nature and robustness
- Researchers propose a novel ADHD diagnosis system using a skeleton-based action recognition framework
- The proposed approach shows significant performance improvement while maintaining cost-efficiency, making it more accessible for initial ADHD diagnoses across a broad range of patients
- The experiment results demonstrate that the proposed method outperforms conventional methods in accuracy and AUC
- The researchers note that their approach is widely applicable for mass screening purposes.

## Summary

Attention Deficit Hyperactivity Disorder (ADHD) is a prevalent neurobehavioral disorder that affects individuals worldwide. While machine learning methods have been extensively studied for ADHD diagnosis, most rely on expensive equipment such as MRI machines and EEG patches, making them inaccessible to many patients. As a result, there is a need for low-cost diagnostic methods based on the action characteristics of ADHD. Skeleton-based action recognition has gained attention due to its action-focused nature and robustness. In this study, Yichun Li, Yi Li, Rajesh Nair, and Syed Mohsen Naqvi propose a novel ADHD diagnosis system using a skeleton-based action recognition framework. The researchers utilize a real multi-modal ADHD dataset and state-of-the-art detection algorithms to develop their method. Compared to conventional methods, the proposed approach shows significant performance improvement while maintaining cost-efficiency, making it more accessible for initial ADHD diagnoses across a broad range of patients. The experiment results demonstrate that the proposed method outperforms conventional methods in accuracy and AUC. Additionally, the researchers note that their approach is widely applicable for mass screening purposes. Overall, this study presents an innovative solution for diagnosing ADHD using skeleton-based action recognition technology. By providing an accessible and cost-effective alternative to traditional diagnostic methods, this research has the potential to improve early detection rates of ADHD among diverse populations globally.