Anzeige
Erschienen in:
2020 | OriginalPaper | Buchkapitel
A Multi-task Learning Method for Direct Estimation of Spinal Curvature
verfasst von : Jiacheng Wang, Liansheng Wang, Changhua Liu
Erschienen in: Computational Methods and Clinical Applications for Spine Imaging
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Abstract
In scoliosis diagnosis and treatment, estimation of spinal curvature plays an important role. Compared with the traditional method, which is time-consuming and unreliable, automated estimation has been more and more popular. But it remains to be such a great challenge that direct estimation has poor precision due to the lack of information. To meet this challenge, we propose a Multi-Task learning method with pyramidal feature aggregation. Our method is one-stage. It means that we can directly estimate the angles without detecting landmarks. To enhance the feature extraction and collect more information, we make the fusion of the pyramidal features and extend the base model by adding an extra branch for spinal segmentation. We evaluate our method on the validation set from the challenge (Accurate Automated Spinal Curvature Estimation, MICCAI 2019) and obtain a symmetric mean absolute percentage error of 12.97.