A critical reappraisal of predicting suicidal ideation using fMRI
This highlights risks of overfitting in machine learning for psychiatric diagnosis, potentially affecting researchers and clinicians relying on neuroimaging for mental health assessments.
The paper critically reappraises a prior study that claimed 91% accuracy in predicting suicidal ideation from fMRI data using a Naive Bayes classifier, finding that methodological issues like overfitting undermine the reported results.
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.