A critical reappraisal of predicting suicidal ideation using fMRI

TL;DR

This paper critically reanalyzes a study claiming high accuracy in predicting suicidal ideation from fMRI data, highlighting issues of overfitting and questioning the reliability of the original findings.

Contribution

It provides a re-evaluation of previous results, emphasizing the importance of rigorous validation to avoid overfitting in neuroimaging machine learning studies.

Findings

Original high accuracy likely overestimated due to overfitting

Re-analysis suggests caution in interpreting fMRI-based predictions

Highlights methodological pitfalls in neuroimaging machine learning

Abstract

For many psychiatric disorders, neuroimaging offers a potential for revolutionizing diagnosis, and potentially treatment, by providing access to preverbal mental processes. In their study "Machine learning of neural representations of suicide and emotion concepts identifies suicidal youth."1, Just and colleagues report that a Naive Bayes classifier, trained on voxelwise fMRI responses in human participants during the presentation of words and concepts related to mortality, can predict whether an individual had reported having suicidal ideations with a classification accuracy of 91%. Here we report a reappraisal of the methods employed by the authors, including re-analysis of the same data set, that calls into question the accuracy of the authors findings. The analysis is a case study in the dangers of overfitting in machine learning.