Preliminary Study of Retinal Lesions Classification on Retinal Fundus Images for the Diagnosis of Retinal Diseases

Abstract

Retinal diseases are the principal causes of blindness and vision loss worldwide. Early detection can diminish the incidence of total vision impairment and blindness due to retinal diseases. Thus, the present study was conducted to study retinal lesions classification on retinal fundus images for early detection of retinal diseases. The indicative confirmation of various retinal diseases relies on the recognition and segmentation of dark and bright retinal lesions structures. Bright lesions are mainly categorized into exudates and cotton wool spots and dark lesions into microaneurysms and haemorrhages. Nevertheless, variations present in retinal fundus images make it challenging to distinguish varying types of lesions in the incidence of landmark structures (healthy structures), such as retinal blood vasculature and optic disk. Therefore, it is vital to eliminate false responses due to healthy structures before the segmentation of dark and bright retinal lesions. Additionally, to design a robust automated retinal disease diagnostic method, a retinal image dataset comprising fundus images having varying attributes is essential. Putting these facts into consideration, the detailed preliminary study of various retinal lesions and their classification is carried out in this chapter to design an effective computer-aided diagnostic method for the detection of various retinal diseases.