Artificial Intelligence

   

Machine Learning Alternatives for the Diagnosis of Adhd from Functional Connectivity and Phenotypic Information

Authors: Amrit Baveja

Current estimates are that 5-10% of school age children (including the author) suffer from ADHD, costing the US healthcare system alone over $36B. However, factors such as a revenue-motivated healthcare system and researcher confirmation bias make ADHD overdiagnosis a very real issue, and today, there continues to be no reliable technique for automated ADHD diagnosis in clinical use. This work’s objective was to improve upon previous attempts to develop machine learning models for automated ADHD diagnosis. For this project, the author used the ADHD200 dataset which was generated for an international competition in 2011. The author extended the competition’s approach in several ways: First, the author combined high coverage phenotypic features with the functional connectome features used previously. Next, the author used a random 20% test/train split cross-validated five times to avoid overfitting, rather than the previously used fixed test/train split. Third, the author used a broad range of newer models such as neural networks and deep forest. Finally, the author used newer hyperparameter optimization techniques to identify the best model parameters. The best model explored was the more recent gcForest model with automated optimization -- it improved the previous best ADHD F1 from 0.32 to 0.52, a substantial improvement in binary diagnostic performance.

Comments: 43 Pages.

Download: PDF

Submission history

[v1] 2019-10-22 23:10:12

Unique-IP document downloads: 13 times

Vixra.org is a pre-print repository rather than a journal. Articles hosted may not yet have been verified by peer-review and should be treated as preliminary. In particular, anything that appears to include financial or legal advice or proposed medical treatments should be treated with due caution. Vixra.org will not be responsible for any consequences of actions that result from any form of use of any documents on this website.

Add your own feedback and questions here:
You are equally welcome to be positive or negative about any paper but please be polite. If you are being critical you must mention at least one specific error, otherwise your comment will be deleted as unhelpful.

comments powered by Disqus