# Two distinct trajectories of brain volume loss in myotonic dystrophy type 1 via machine learning

## TL;DR

Machine learning reveals two distinct brain volume loss patterns in myotonic dystrophy Type 1, with one linked to cognitive decline.

## Contribution

The study identifies two subtypes of brain volume loss in myotonic dystrophy Type 1 using machine learning and explores MRI as a potential biomarker.

## Key findings

- Two subtypes of brain volume loss were identified: cortical and subcortical.
- The subcortical subtype showed cognitive impairment and larger ventricle volume.
- MRI morphometry may serve as a biomarker for cognitive decline in the disease.

## Abstract

Myotonic dystrophy Type 1 is a disorder that affects multiple systems, including the muscles and the CNS. Previous studies have primarily used voxel-based morphometry to examine areas of brain volume reduction and their correlation with symptoms; however, consistent findings have not been obtained. Subtype and stage inference is an unsupervised machine learning algorithm that elucidates disease progression and subtypes from cross-sectional data. In this study, we used Subtype and Stage Inference to analyse the morphometric MRI data of patients with myotonic dystrophy Type 1 to reveal the detailed trajectories of brain volume loss and to explore the potential of morphometric MRI as a biomarker for myotonic dystrophy Type 1. We examined 60 patients with myotonic dystrophy Type 1 and 50 age- and sex-matched controls. The patients with myotonic dystrophy Type 1 had a median age of 44 years (range 20–67 years) and included 32 males. Using three-dimensional T1-weighted MRI images, we analysed the subtypes of brain involvement and their respective trajectories of brain volume loss with subtype and stage inference. Additionally, we examined the differences and correlations in clinical and brain morphological indicators between the identified subtypes and controls. Subtype and stage inference revealed two subtypes: cortical and subcortical. In the cortical subtype, volume reduction began in the precentral gyrus and spread primarily to the cerebral cortex. In the subcortical subtype, it progressed early in the putamen, thalamus, hippocampus and amygdala. Examination of clinical indicators showed that despite the younger age of the subcortical subtype compared to the cortical subtype, mini-mental state examination scores were significantly lower in the subcortical subtype and negatively correlated with subcortical probability. The total intracranial volume, a marker of maximal brain growth, was significantly smaller in the cortical subtype; however, it was not smaller in the subcortical subtype than in controls. Furthermore, the subcortical subtype showed a larger total ventricle volume than both the controls and the cortical subtype. In contrast, its total brain parenchymal volume was lower than that of the controls, similar to the cortical subtype. These results suggest early childhood brain development differences between the two subtypes. Using Subtype and Stage Inference, we identified two subtypes of myotonic dystrophy Type 1 and demonstrated the potential of morphological MRI as a biomarker for cognitive impairment and brain developmental disorders. Machine learning can aid in stratifying myotonic dystrophy Type 1 in clinical settings and contribute to the elucidation of its complex pathophysiology.

Linking brain imaging findings to the symptoms of myotonic dystrophy Type 1 is challenging. Imokawa et al. used machine learning to identify two brain volume loss trajectories, cortical and subcortical, with the latter linked to cognitive impairment. This study demonstrated the potential of MRI as an indicator of cognitive decline.

Graphical Abstract