Balancing Accuracy and Efficiency in Inferior Alveolar Canal Segmentation from CBCT Scans

Changkai Ji, Yusheng Liu, Yuxian Jiang, Lisheng Wang

Keywords: nnU-Net, Interactive segmentation, Inferior alveolar canal, Computational efficiency

TL;DR: This paper presents an accurate and efficient approach for segmenting the inferior alveolar canal in CBCT scans, analyzing nnU-Net and nnInteractive techniques to achieve high segmentation performance with optimized computational efficiency.

Abstract: Accurate segmentation of the inferior alveolar canal (IAC) in cone-beam computed tomography (CBCT) scans is crucial for dental and maxillofacial applications, yet it remains highly challenging due to its fine-scale structure, anatomical variability, and imaging artifacts. To solve this problem, we utilized both automated and interactive approaches. Specifically, we compared an nnU-Net-based model with an nnInteractive model. While nnInteractive demonstrated promising improvements with minimal user input, our final submission was based on nnU-Net due to its favorable trade-off between accuracy and computational efficiency. To further enhance runtime performance, we incorporated inference acceleration strategies, achieving a speedup without sacrificing segmentation quality. Our method achieved a top-three ranking in the challenge test phase, highlighting its potential for accurate and efficient IAC segmentation. Our code is avaiable at: https://github.com/duola-wa/Toothfairy3.

Submission Number: 14