An implementation of regional descriptor and line roi in development of semi-automated strabismus detection system
The strabismus (squint) is one of children's most common vision disorders. It can cause discomfort and have a significant detrimental effect on daily life. A timely diagnosis is needed to prevent it from getting worse. However, the traditional diagnosis screening is usually done manually and re...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://eprints.uthm.edu.my/10966/1/24p%20NUR%20SYAZLIN%20%20ZOLKIFLI.pdf http://eprints.uthm.edu.my/10966/2/NUR%20SYAZLIN%20%20ZOLKIFLI%20COPYRIGHT%20DECLARATION.pdf http://eprints.uthm.edu.my/10966/3/NUR%20SYAZLIN%20%20ZOLKIFLI%20WATERMARK.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The strabismus (squint) is one of children's most common vision disorders. It can cause discomfort and have a significant detrimental effect on daily life. A timely diagnosis is needed to prevent it from getting worse. However, the traditional diagnosis screening is usually done manually and requires expertise, time and high cost due to the equipment. Thus, the proposed semi-automated strabismus detection using computer-aided diagnosis can help to reduce the time for the ophthalmologist to diagnose the strabismus and the misalignment measurement. This research aims to propose the image processing approach for detection and diagnosis of strabismus. This research proposes three phases: pre-processing, feature extraction, and classification. Initially, the image in pre-processing undergoes Viola Jones algorithm, red channel extraction, contrast adjustment and median filtering to reduce the noise and enhance the image. In feature extraction, binarization and morphological operations are implemented to identify the location of the iris and the misalignment measurement. Finally, the classification is divided into two, where the coordinates of the iris and misalignment measurement are carried out using regional descriptor and line ROI, while the strabismus and non-strabismus are classified using Convolutional Neural Network (CNN). The experimental results have proven that the proposed method has successfully detected the strabismus and the misalignment measurement with an average accuracy for the Eye Disease dataset (0.9167), Google Images (0.9217), CAVE (0.9167), and SiblingsDB (0.9167). In conclusion, by utilizing the image processing approach, this system will be able to assist the ophthalmologist and health care practitioners as strabismus pre-screening tools |
|---|