Data leakage in feature selection invalidates reported classification accuracy for autism diagnosis from fMRI
Semyon Kurkin, Alexander Hramov · 2026
Read Data leakage in feature selection invalidates reported classification accuracy for autism diagnosis from fMRI by Semyon Kurkin, Alexander Hramov (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13382443.txt.
Description
Semyon Kurkin and Alexander Hramov wrote a 2026 letter challenging a study that claimed very high accuracy in diagnosing autism from brain scans. The original study used deep learning on a shared scan data set and reported accuracy above ninety-eight percent. The letter writers checked the authors' publicly shared code and found a critical flaw.
Feature selection, which narrows thousands of scan measures to a smaller set, was run on the whole data set before the data were split for validation. That leaks information from the test cases into training and inflates the reported scores. The letter sets out the code evidence line by line and argues the published performance figures cannot be trusted.
Catalog record
| CREATOR | Semyon Kurkin, Alexander Hramov |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 7640 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |