แจ้งเอกสารไม่ครบถ้วน, ไม่ตรงกับชื่อเรื่อง หรือมีข้อผิดพลาดเกี่ยวกับเอกสาร ติดต่อที่นี่ ==>
หากไม่มีอีเมลผู้รับให้กรอก thailis-noc@uni.net.th
หากไม่มีอีเมลผู้รับให้กรอก thailis-noc@uni.net.th
Shrestha, Raisha. Classification of treatment requirement for posterior capsular opacification using digital image processing, unsupervised machine learning and deep learning. Master's Degree(Engineering and Technology). Thammasat University. Thammasat University Library. : Thammasat University, 2020.
Classification of treatment requirement for posterior capsular opacification using digital image processing, unsupervised machine learning and deep learning
Name: Shrestha, Raisha
keyword:
Posterior capsular opacification
Abstract:
Posterior Capsular Opacification (PCO) is an after cataract complication which has been under continuous research over past couple of decades. Studies so far have been limited to its grading, calculating severity and rate of occurrence. The grading can be more useful in practical world, if its severity level determined cases with requirement of treatment. PCO patients are treated only when PCO is in central visual axis and patient has difficulty in eyes, resulting in need of regular check-ups. This research presents an approach of PCO treatment classification, with two methods: intensity-based (method 1) and k-means clustering based (method 2), to classify the requirement of treatment. When cross-checked to classifications made by expert ophthalmologists, methods 1 and 2 showed accuracy of 94.57% and 96.74%, with F1 score of 0.9669 and 0.9811 respectively. Also, the classifications made by method 1 and 2 were 83% and 87% correlated to the classification made by expert ophthalmologists. Another approach we use is deep learning based (method 3) to semantically segment and classify images of Posterior Capsular Opacification (PCO) into treatment required and not yet required cases to make the treatment easier for both patients and medical staffs. To train the model, we prepare a training dataset with labels obtained from two strategies: (i) manual labelling and (ii) automated labelling using unsupervised machine learning method. We then compare classification accuracy of models trained with the two strategies to see if automated labeling can replace manual labeling. Using both strategies, the accuracy of models was found to be greater than 94% and loss less than 0.15 in our experiments. Comparison to clinical classification shows 0.97 and 0.96 F2-scores for outputs from automated and manual labels, respectively
Thammasat University. Thammasat University Library
Address:
BANGKOK
Email:
preserv@tu.ac.th
Name: Waree Kongprawechnon
Role:
advisor
Created:
2020
Modified:
2023-01-26
Issued:
2023-01-26
วิทยานิพนธ์/Thesis
application/pdf
eng
DegreeName: Master of Science
Level: Master's Degree
Descipline: Engineering and Technology
Grantor: Thammasat University
©copyrights Thammasat University
RightsAccess:
ใช้เวลา
0.032146 วินาที
Shrestha, Raisha
| Title | Contributor | Type |
|---|---|---|
| Classification of treatment requirement for posterior capsular opacification using digital image processing, unsupervised machine learning and deep learning
มหาวิทยาลัยธรรมศาสตร์ Shrestha, Raisha | Waree Kongprawechnon | วิทยานิพนธ์/Thesis |
Waree Kongprawechnon