Evaluation of target region involvement in neurodegenerative diseases VİA segmentation

Introduction and aim: The loss of work power caused by neurodegenerative diseases and the burden on the health system deserve attention with the aging world population. Therefore, early and accurate diagnosis, to foresee mental destruction by diseases and to follow up at subsequent imaging at anatomical base, to slow down progression with medical treatment are gaining importance for both individual and society. Our aim is detecting the neuronal loss and its natural consequence atrophy at involved regions in brain in neurodegenerative diseases via volumetric calculations based on MRI segmentation and to determine and evaluate role of AI at diagnosis. Material and Method: 57 patients age between 45 and 95 years old (29 women, 28 men) who had obtained MRI examination with a diagnosis of neurodegenerative disease between January 2019 and December 2020 with 3D T1A TSE follow-up images were included to study. All images were assessed and visually scored atrophy rates for relevant age. Then, volumetric information was obtained with the fully autonomous atlas-based segmentation program and volumetric loss calculated in the target regions was compared with the scores given to follow up images. Demographic characteristics, MMSE scores and etiology of the cases were evaluated retrospectively. Results: In the evaluation of the volume data of the patients grouped according to their etiology, it was observed that the whole brain volume and temporal lobe volumes were lower in both first and last application in Alzheimer's patients. In addition, the annual rate of brain atrophy was slightly higher in Alzheimer's patients. In the evaluation of occipital lobe volumes, relative volume loss in the first and last application volumes draws attention in Parkinson's patients. However, the numerical differences defined for the temporal lobe and occipital lobe were not reflected as significant statistical differences due to the small number of patients. No significant differences were observed in reference volumes or annual atrophy rate in other brain regions. 13 In comparison of the groupings made for radiologically defined atrophy rates and percentile curves calculated according to age, a moderate level of agreement was observed in kappa analysis. While the compliance rates in brain regions with severe atrophy or accepted as radiologically normal ranges above 80%, it was observed that the most inconsistency set among the data between the 25th and 50th percentile. Conclusion: In the evaluation of neurodegenerative disease involvement, volumetric data provided by segmentation methods may contribute to the diagnostic process. However, the targeted selection of segmentation methods, the availability of more advanced software that allows detailed analysis, and the evaluation of follow-up imaging in large patient groups will provide a more widespread and more accurate use of AI-supported diagnosis Key words: AI, MRI, Neurodegenerative Diseases, Segmentation