Second Mesiobuccal Canal Segmentation With Yolov5 Architecture Using Cone Beam Computed Tomography Images

Introduction

The study aimed to detect second mesiobuccal (MB2) canals in maxillary molars using a deep-learning model based on the You Only Look Once v5 (YOLOv5x) architecture. MB2 canals are crucial in endodontic treatment success, but their detection can be challenging.

Materials and Methods

- 922 axial sections from CBCT images of 153 patients were used.

- MB2 canals were manually labeled using polygonal techniques.

- Preprocessing steps included image resizing and enhancement techniques.

- Training utilized a NVIDIA V100 GPU with 16 GB VRAM and 5120 CUDA Cores.

 Results

- The deep-learning model achieved a sensitivity of 0.92, precision of 0.83, and F1 score of 0.87 for MB2 canal segmentation.

- The area under the curve (AUC) in the ROC graph was 0.84.

- The mean average precision (mAP) value at 0.5 IoU was 0.88, indicating high accuracy in MB2 canal detection.

Discussion

- The study demonstrates the potential of AI models in improving endodontic treatment outcomes by identifying root canal variations like MB2 canals.

- YOLOv5 architecture showed promise in efficiently detecting intricate dental structures, enhancing treatment success and saving clinicians' time.

Conclusion

- The deep-learning algorithm successfully detected MB2 canals in maxillary molars, offering a reliable and speedy method for clinicians.

- Utilizing AI in CBCT images allows for three-dimensional examination of canals, enhancing diagnostic capabilities in endodontics.