Application of deep convolutional neural network for clinical decision support in diagnosis benign pathological lesions of the jaw

Name: Wiwiek Poedjiastoeti

keyword: Benign lesions of the jaw

; Deep learning

; Convolutional neural network

Abstract: As general dental practitioner or dental students, its difficult to compare benign pathological lesions in the jaws, because of their similarities in the clinical and radiographic appearance. To have the true diagnosis of these lesions, we need to perform many examination before we conclude the true diagnosis of those oral lesions. Early diagnosis of the disease will increase the chances of recovery dramatically. The aim of this study is to develop and evaluate whether CNN has a role in identifying and distinguish Ameloblastomas and Keratocystic Odontogenic Tumors (KCOTs) in digital panoramic images. Materials and methods: The training and validation data comprised 200 ameloblastomas images and 200 KCOT images. The test data comprised 50 ameloblastoma images and 50 KCOT images. To overcome the limitation of the small training dataset, we applied the data augmentation method to increase the number of training datasets. Second, a VGG-16 (16-layer CNN) was pre-trained in ImageNet, and refined with our secondary dataset training. . A separate test data set with known biopsy results was evaluated to compare the performance of the developed CNN with that of board certified oral and maxillofacial specialist. Results: The diagnostic performance of CNN was relatively high, with an accuracy 83%, sensitivity 81.8%, specificity 83.3%, PPV 85.2% and NPV 80.5%. The Oral and Maxillofacial Specialist demonstrated an accuracy 82.9%, sensitivity 81.1%, specificity 83.2%, PPV 86.2% and NPV 80.7%. The AUCs of the deep learning system and of the Specialist were 0.88 and 0.90, at the cutoff point of the operating threshold of 0.43, respectively. The total calculation time taken by CNN to analyze all the images was 38 seconds. The average time for the Oral and Maxillofacial Specialist to evaluate all the images of the test datasets was 23.1 minutes. Conclusion: We develop an algorithm which detects and diagnose Ameloblastoma and KCOT from panoramic digital X-ray images at a level exceeding practicing Oral and Maxillofacial Specialist, thus the CNN as deep learning system could be a useful method for diagnostic support

Thammasat University. Thammasat University Library

Address: BANGKOK

Email: preserv@tu.ac.th

Name: Siriwan Suebnukarn

Role: advisor

Created: 2019

Modified: 2022-11-22

Issued: 2022-11-22

วิทยานิพนธ์/Thesis

application/pdf

DOI: 10.14457/TU.the.2019.1571

eng

DegreeName: Doctor of Philosophy

Descipline: Oral Health Science

Grantor: Thammasat University

Level: Doctoral Degree

©copyrights Thammasat University

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	9442wiwiek.pdf	1.55 MB

ใช้เวลา

0.021391 วินาที