You can download my doctoral thesis from here.
Strabismus Detection with Image Processing and Deep Learning
Approaches
Strabismus, also known as squint, is a common eye disorder affecting 4% of the world's population. In strabismus screening, some clinical tests such as Hirschberg, Covering, Prism Covering and Krimsky areregularly applied to the patient by ophthalmologists. The results of these tests are used to guide the treatment of strabismic patients. Periodic strabismus tests are repeated at intervals of 2-6 months. The fact that the tests used in strabismus screening are performed automatically increases the number of patients treated per unit of time in ophthalmology clinics and allows preliminary examinations to be carried out more quickly. Automating these tests in schools and rural areas is an important step in reducing the incidence of strabismus worldwide, which should be detected and treated at an early age. The continuous development of artificial intelligence techniques makes the previously proposed software models for the automatic performance of strabismus tests inadequate and makes it difficult to use these models effectively. In this sense, a literature review was carried out and the models proposed in this thesis were compared with those proposed in the literature. This thesis proposes software models that can perform the Hirschberg, Covering and Prism Covering tests, which are the most commonly used in strabismus screening, using image processing and deep learning techniques. A total of 165 strabismic patients' data were collected to perform the Hirschberg test automatically. The first of the proposed software models for the Hirschberg test was tested on the images of 88 strabismic patients and correctly identified the degree of strabismus in %90 of the patients. The second software model proposed for the Hirschberg test was tested on high-resolution images of 77 pediatric (0-12 years) strabismic patients and achieved a %97.4 success rate. For 19 of the 77 strabismic patients, the results of the covering test and the results of the automated Hirschberg test were compared, and the test results were verified by the ophthalmologist to be compatible. In this thesis, a medical device has been developed that automatically performs the Covering Test and the Prism Covering Test. This device automatically performs the Covering Test and Prism Covering Test, which are the most commonly used tests in strabismus screening, with the support of artificial intelligence. This device, which performs the covering/uncovering process in the eyes of strabismic patients and examines the patients' eye movements, performs the covering/uncovering tests fully automatically and determines the patients' degree of strabismus within seconds. Each stage of the automatic tests in the thesis was verified by an ophthalmologist and the progress was carried out within the framework of the medical information provided by the ophthalmologist.
Keywords: Strabismus, Hirschberg Test, Cover Test, Prism Cover Test, Image Processing, Deep Learning
0 Yorumlar