Cataract Detection using Pupil Patch Classification and Ruled-based System in Anterior Segment Photographed Images

A cataract is an ocular disease affecting the eye’s anterior segment, and it results from clouding of the lens. If left untreated, this condition can cause blindness or vision impairment. The current diagnosis of cataracts involves a series of manual tests that are time consuming, subjective and dependent on the experience of ophthalmologists. Moreover, the medical equipment used for detection and screening is costly. Digital images have been used to develop healthcare applications, including anterior segment images applied to detect ocular diseases. Therefore, this paper presents a cataract detection method using anterior segment photographed images (ASPIs) captured with a smartphone’s camera. The proposed methodology includes a preprocessing step, and cataract detection comprises patch classification and a rule-based system. Using 540 normal and cataract patches, five different pretrained networks were employed to classify the patches. The experimental results showed that the patch classification model with ResNet-50 achieved the highest accuracy, specificity, and area under curve values of 98%, 98.1% and 0.999, respectively. Then, the patch classification model was used to detect cataracts in 20 cataracts and 20 normal ASPIs using a rule-based system. The proposed method achieved high accuracy in cataract detection. The proposed method can be potentially used for cataract detection or screening and can help optometrists or ophthalmologists, particularly in rural areas.