A standard cerebral artery coordinate has been developed to be able to evaluate patient's blood vessel shapes quantitatively. During the construction of a standard cerebral coordinate, the extraction of centerline data is required. Then, blood vessel part centerlines need a manual labeling for each artery name. Since the construction of a standard cerebral coordinate requires a huge amount of labeled data, the huge amount of manual labeling process is time-consuming. Therefore, the development of automatic labeling program is an urgent issue. In automatic labeling of cerebrovascular sites, In conventional studies, an existence probability atlas is constructed and automatic labeling based on this atlas is performed. They used a same number of patients for each artery site to construct the existence probability atlas. However, the artery shapes cannot be obtained in some parts because of image quality and individual variation of the artery structure. Therefore, sometimes the number of arteries in the MR data set varies. Especially in the small data set, the difference sample size often has a large influence on existence probability based on superimposing reconstructed artery regions. Hence, to moderate this influence, we propose a new algorithm of atlas with normalization for automatic labeling in the construction of a standard cerebral coordinate and evaluate its accuracy and precision.The cerebral artery shapes were obtained from nuclear Magnetic Resonance Angiography (MRA) images of patients with aldosteronism. And the centerline and the diameter of main cerebral arteries were obtained, and centerlines were labeled manually. To construct the existence probability atlas, based on the obtained centerline and blood vessel diameter data, the blood vessel region for each patient and each site was reconstructed in a voxel space. The existence probability atlas was created by superimposing these vessel regions. Three normalization methods were compared: dividing by the number of patients, by the number of patients and the length of blood vessels, and by the number of patients and volume, respectively. Summation of Probability and Labeling Method (SPLM) is conducted to calculate the macro averages for the accuracy and the precision, respectively.As the results, accuracy of all parts was higher than 85% in all normalization methods. Whereas, the precision varies among the parts in all methods. In some parts, improvements are observed by the normalization methods. By the normalization, the relative value of the existence probability among the parts changes. The confusion matrices imply that the tendency of miss-classification varies among three normalizations. Therefore, the normalization changes the tendency of the miss-classifications and results in changes of precision.In conclusion, the automatic labeling algorithm, SPLM is used to compare the normalization methods for existence probability atlas. The accuracy maintains the values higher than 85%. And normalization improves the precision in some artery parts.

Evaluation of Normalization Methods in a Cerebral Artery Atlas for Automatic Labeling

Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2021/volume/85543

ASME 2021 International Mechanical Engineering Congress and Exposition (IMECE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.