Machine learning classification of functional neurological disorder using structural brain MRI features 2024 Westlin, Perez et al

Abstract

Background
Brain imaging studies investigating grey matter in functional neurological disorder (FND) have used univariate approaches to report group-level differences compared with healthy controls (HCs). However, these findings have limited translatability because they do not differentiate patients from controls at the individual-level.

Methods
183 participants were prospectively recruited across three groups: 61 patients with mixed FND (FND-mixed), 61 age-matched and sex-matched HCs and 61 age, sex, depression and anxiety-matched psychiatric controls (PCs). Radial basis function support vector machine classifiers with cross-validation were used to distinguish individuals with FND from HCs and PCs using 134 FreeSurfer-derived grey matter MRI features.

Results
Patients with FND-mixed were differentiated from HCs with an accuracy of 0.66 (p=0.005; area under the receiving operating characteristic (AUROC)=0.74); this sample was also distinguished from PCs with an accuracy of 0.60 (p=0.038; AUROC=0.56). When focusing on the functional motor disorder subtype (FND-motor, n=46), a classifier robustly differentiated these patients from HCs (accuracy=0.72; p=0.002; AUROC=0.80). FND-motor could not be distinguished from PCs, and the functional seizures subtype (n=23) could not be classified against either control group. Important regions contributing to statistically significant multivariate classifications included the cingulate gyrus, hippocampal subfields and amygdalar nuclei. Correctly versus incorrectly classified participants did not differ across a range of tested psychometric variables.

Conclusions
These findings underscore the interconnection of brain structure and function in the pathophysiology of FND and demonstrate the feasibility of using structural MRI to classify the disorder. Out-of-sample replication and larger-scale classifier efforts incorporating psychiatric and neurological controls are needed.

Open access, https://jnnp.bmj.com/content/early/2024/07/20/jnnp-2024-333499

 

how does this work, given their longtime reliance on the software/hardware analogy?

 

SND.

 

This study is a step in the right direction, though none of the areas under the curve are clinically useful as too many patients are misclassified.

 

Whether or not Jon Stone is an author is a usually decent proxy for how “useful” an FND study is. Checks out here.

 

Perez is just as awful and biased, but the data about differentiating are pretty bad. Not sure if this is anything beyond "there are non-specific changes", which is not super useful but does invalidate the entire premise of the ideology.

But a faith-based ideology doesn't adapt to facts, so it's not as it matters much. We are talking about people who have no issues with insisting that all the symptoms are caused by deconditioning, but will excuse the failure of an exercise program on the basis that they are active enough anyway. They don't care about facts, and neither does the rest of the profession, happy to let them spew bullshit as if it somehow frees the entire profession from responsibility here.

 

well, David Perez at Harvard is the US version of Stone/Carson. And the three often are co-authors together.

 

But I still don't get how they can argue there could be structure-related biomarkers and still argue that it's all a software issue. Why don't they address that contradiction? At some point, don't they need to acknowledge that something is physically wrong besides "software" and "brain network" issues?

 

Was when I posted, looks like they changed it.

 

Wasn't Frackowiak's taxi driver paper shown to be garbage data?

It doesn't, as you say.

They don't seem to explain why the psychiatric controls are different from healthy? And if there is a difference what is the justification for putting all psychiatric patients in one group?
And so on.

 

https://twitter.com/user/status/1815418056321175668