Related Experiment Video
Updated: Jul 17, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
Abnormal maxillary sinus diagnosing on CBCT images via object detection and 'straight-forward' classification deep
Peisheng Zeng1, Rihui Song2, Yixiong Lin1
1Hospital of Stomatology, Guanghua School of Stomatology, Sun Yat-sen University and Guangdong Research Center for Dental and Cranial Rehabilitation and Material Engineering, Guangzhou, China.
Journal of Oral Rehabilitation
|September 4, 2023
Summary
A new deep learning model accurately screens maxillary sinus abnormalities on CBCT scans, improving diagnostic performance for dental implant procedures. This AI tool aids dentists in identifying potential issues before surgery.
Area of Science:
- Artificial Intelligence in Dentistry
- Medical Imaging Analysis
- Cone-Beam Computed Tomography (CBCT)
Background:
- Maxillary sinus abnormalities can compromise dental implant success, leading to treatment failure.
- Diagnosing these abnormalities on CBCT images is challenging, particularly for less experienced dentists.
- Accurate diagnosis is crucial for successful sinus lift and implant procedures.
Purpose of the Study:
- To develop a deep learning (DL) model for screening maxillary sinus abnormalities using CBCT images.
- The model aims to improve the accuracy and efficiency of diagnosing these conditions.
- To assist dentists, especially those in primary care, in identifying potential risks.
Main Methods:
- An object detection technique was used to define the region of interest, reducing background noise.
- A 'straight-forward' classification strategy was implemented with tuning methods for clinical adaptation.
- The model was trained to classify images as either normal or abnormal, focusing on typical features.
Main Results:
- The DL model achieved high performance with an AUROC of 0.953 and AUPRC of 0.887.
- The model demonstrated over 90% accuracy at the optimal cut-off point.
- Comparison tests showed the model outperformed dental students in diagnostic accuracy, with high consistency to ground truth.
Conclusions:
- The developed DL model effectively screens maxillary sinus abnormalities on CBCT images.
- The combination of object detection and straightforward classification yields satisfactory predictive performance.
- This AI-driven approach can enhance diagnostic capabilities in dental practice.
