표제지
목차
용어집 7
논문개요 8
I. 서론 9
II. 연구 대상 및 방법 14
A. 데이터 획득 및 전처리 14
B. 합성곱 신경망 기반의 충치 분할법 17
B.1. DeepLabV3+ 17
B.2. DeepLabV3+를 이용한 충치 분할법 18
B.3. 영상증강 20
C. 평가방법 23
III. 결과 24
IV. 고찰 29
V. 결론 31
참고문헌 32
ABSTRACT 36
Table 1. Dice similarity coefficient of training and test data for DeepLabV3+ based caries auto segmentation 26
Figure 1. Periapical and bitewing X-ray machine (a), Periapical X-ray image (b), Bitewing X-ray image (c), Panoramic X-ray machine (d), and... 10
Figure 2. Caries in bitewing X-ray image (red region) 15
Figure 3. Bitewing X-ray image (a) and caries image (b) 16
Figure 4. The cropped bitewing X-ray images (a-d) and caries images (e-h) used for Deep-learning based caries auto segmentation. 17
Figure 5. Diagram of DeepLabV3+ model 18
Figure 6. Dice similarity coefficient for training and test data during the training of DeepLabV3+ 20
Figure 7. Original bitewing image (a), caries image (b) and augmented bitewing image (c) and caries image (d) 21
Figure 8. Original bitewing image (a), caries image (b) and augmented bitewing image (c) and caries image (d) 22
Figure 9. Histogram of dice similarity coefficient for the training (a) and test data (b) 24
Figure 10. Bitewing X-ray images (a, d, g), ground truth images (b, e, h), and predicted images by DeepLabV3+ (c, f, i) for caries auto segmentation 25