Title Page
Contents
국문초록 8
ABSTRACT 10
I. Introduction 12
II. Materials and Methods 15
III. Results 22
IV. Discussion 24
V. Conclusion 28
References 29
Table 1. MRI sequence parameters for training set 35
Table 2. Demographic data and analysis results for study population 36
Table 3. Diagnostic predictive values of DLA for segmentation of glands and tumors 37
Table 4. Diagnostic predictive values of DLA for tumor classification 38
Fig. 1. Flowchart of case-enrollment process 39
Fig. 2. A representative case of gland segmentation. The Dice score for the gland segmentations was 0.94. 40
Fig. 3. A representative case of tumor segmentation with GS7(4+3). The Dice score for the tumor segmentations was 0.78. 41
Fig. 4. Graphs showing precision recall curve (A) and receiver operating characteristic (ROC) curve (B B) of deep learning algorithm... 42
Fig. 5. A representative case of misclassification: over-estimation of GS6 as GS7. 43
Fig. 6. A representative case of misclassification: under-estimation of GS7 as GS6. 44