Hypothalamic atrophy in primary lateral sclerosis, assessed by convolutional neural network-based automatic segmentation

Primary lateral sclerosis (PLS) is a motor neuron disease (MND) which mainly affects upper motor neurons. Within the MND spectrum, PLS is much more slowly progressive than amyotrophic laterals sclerosis (ALS). `Classical` ALS is characterized by catabolism and abnormal energy metabolism preceding onset of motor symptoms, and previous studies indicated that the disease progression of ALS involves hypothalamic atrophy. Very limited weight loss is observed in patients with PLS, which raises the question of whether there are also less hypothalamic alterations. The purpose of this study was to quantitatively investigate the hypothalamic volume in a group of PLS patients and to compare it with ALS and controls. Recently, we have introduced automatic hypothalamic quantification method based on the use of convolutional neural network (CNN) to reduce human variability and enhance analysis robustness. This CNN of U-Net architecture was applied for automatic segmentation of the hypothalamus and intracranial volume (ICV) to allow adjustments of the hypothalamic volume between subjects with different head sizes respectively. Automatic segmentation and volumetric analysis were performed in high resolution T1 weighted MRI volumes (acquired on a 1.5 T MRI scanner) of 46 PLS patients in comparison to 107 healthy controls and 411 `classical` ALS patients, respectively. Significant hypothalamic volume reduction was observed in PLS (818 ± 73 mm³) when compared to controls (852 ± 77 mm³); significant hypothalamic volume reduction was also confirmed in ALS (823 ± 84 mm³), in support of previous studies. No significant differences were found in normalized hypothalamic volumes between ALS patients and PLS patients at the group level. This unbiased CNN-based hypothalamus volume quantification study demonstrated similarly reduced hypothalamus volume in PLS and ALS patients, despite the clinical phenotypic differences.