Machine learning classifies premenstrual dysphoric disorder in terms of grey matter structure

Researchers from Uppsala University and Umeå University, led by SciLifeLab Fellow Erika Comasco (Uppsala University) have used data-driven methods, such as machine learning, to analyse grey matter measurements from women with premenstrual dysphoric disorder (PMDD), and compared them to healthy controls. The approach allows to objectively classify neuropsychiatric disorders.

Psychiatry is hampered by the lack of neurobiological diagnostic and treatment markers and is still relying on patients’ self-reports of symptoms. The limited accessibility of the tissue of interest (brain), is circumvented by the employment of in vivo neuroimaging techniques, such as magnetic resonance imaging (MRI), which allows to measure brain structure.

Grey matter is the part of the brain where bodies of neural cells reside. Structural MRI (sMRI) captures the brain morphology, and combined with voxel-based morphometry (VBM), it estimates regional volumes of grey matter, while surface-based morphometry (SBM) enables the reconstruction of the cortical surface and extraction of brain architecture metrics (i.e. cortical thickness, gyrification index, sulcal depth, and cortical complexity).

In a recent study, SciLifeLab Fellow Erika Comasco (Uppsala University) and her team evaluated grey matter properties in women with premenstrual dysphoric disorder (PMDD) and in healthy women. The disorder affects a woman’s ability to work or go to school, and relationships are often harmed by the emotional lability. Symptoms of PMDD peak during the premenstrual phase of a woman’s menstrual cycle and resume upon menstruation onset. Profound irritability or anger, mood swings, anxiety and depression, that are severe enough to negatively affect usual life activities, are key to the diagnosis.

Earlier this year, the researchers provided the first evidence of brain morphological characteristics (i.e., grey matter surface and volume) being related to PMDD symptomatology. Interestingly, the severity of mental and physical PMDD symptoms were related to amygdala grey matter volume, with this key region in emotion processing being smaller the more severe the symptoms are.

In support of this quantitative relationship, the researchers demonstrated a qualitative association depicted by smaller volume of the amygdala in women with PMDD compared to the healthy controls. Additionally, cortical thickness and folding of corticolimbic regions, all regions expressing sex hormone receptors and of relevance to cognitive-affective functions, were related to PMDD symptomatology.

In collaboration with Professor Marie Bixo and her team from Umeå University, the researchers combined their finely characterized neuroimaging datasets and used a machine learning approach to test whether the brain morphological differences would enable them to distinguish women with PMDD from controls.

“Based on grey matter probability maps, the grey matter volume of specific brain areas contributed to the most to the classification accuracy” says SciLifeLab Fellow Erika Comasco (The Department of Women’s and Children’s Health, Uppsala University).

“These data-driven findings are expected to help to move the field toward precision psychiatry. Thus, we plan to assess the accuracy of a neuroimaging-based classifier not only for the automated detection of PMDD but also to predict treatment response, and ultimately contribute to advance the field of women´s mental health”, she concludes.


Differential grey matter structure in women with premenstrual dysphoric disorder: evidence from brain morphometry and data-driven classification
Dubol M, Stiernman L, Wikström J, Lanzenberger R, Neill Epperson C, Sundström-Poromaa I, Bixo M, Comasco E.
Transl Psychiatry. 2022 Jun 15;12(1):250. doi: 10.1038/s41398-022-02017-6.

Grey matter correlates of affective and somatic symptoms of premenstrual dysphoric disorder
Dubol M, Wikström J, Lanzenberger R, Epperson CN, Sundström-Poromaa I, Comasco E.
Sci Rep. 2022 Apr 9;12(1):5996. doi: 10.1038/s41598-022-07109-3.


Last updated: 2022-11-09

Content Responsible: Johan Inganni(