The aim of the study conducted by the University Hospital Bonn was to compare automated measurements of brain atrophies with manual measurements in patients with confirmed, advanced Huntington's disease. For this purpose, the results of the deep-learning-based software mdbrain for automatic volumetry were compared with the manual measurement methods and segmentations used in clinical practice for various anatomical regions. The comparison showed a high agreement between the automatically and manually determined volumes of the caudate nuclei, with an average relative deviation of only −2.3 ± 5.5%.

The study confirmed that the volume of the caudate nuclei automatically determined by mdbrain allowed a clear distinction between the groups. This was supported by comparing the results with established thresholds, such as the ratio of frontal horn width to the distance of the caudate nuclei. Additionally, significantly lower volumes in several brain regions of the Huntington group compared to the control group were found, particularly in the caudate nucleus, putamen, and globus pallidus (all p < 0.0001).

A unique feature of the AI software is the output of percentiles and their classification in relation to standard deviations from the stored reference collective. The normative values are calculated individually for each patient based on characteristics such as gender and age. This allows the evaluation of the determined volumes of individual cases even without a control group in everyday clinical practice.

"This independent study confirms the precision and reliability of our software solution in the analysis of brain volumes," says Andreas Lemke, CEO of mediaire. "We are proud that mdbrain provides valuable support not only for common but also for rare issues in neurodiagnostics.”