Diabetes and Cataracts Development-Characteristics, Subtypes and Predictive Modeling Using Machine Learning in Romanian Patients: A Cross-Sectional Study.
Journal:
Medicina (Kaunas, Lithuania)
PMID:
39859011
Abstract
Diabetes has become a global epidemic, contributing to significant health challenges due to its complications. Among these, diabetes can affect sight through various mechanisms, emphasizing the importance of early identification and management of vision-threatening conditions in diabetic patients. Changes in the crystalline lens caused by diabetes may lead to temporary and permanent visual impairment. Since individuals with diabetes are at an increased risk of developing cataracts, which significantly affects their quality of life, this study aims to identify the most common cataract subtypes in diabetic patients, highlighting the need for proactive screening and early intervention. This study included 201 participants with cataracts (47.6% women and 52.4% men), of whom 105 also had diabetes. With the use of machine learning, the patients were assessed and categorized as having one of the three main types of cataracts: cortical (CC), nuclear (NS), and posterior subcapsular (PSC). A Random Forest Classification algorithm was employed to predict the incidence of different associations of cataracts (1, 2, or 3 types). Cataracts have been encountered more frequently and at a younger age in patients with diabetes. CC was significantly more frequent among patients with diabetes ( < 0.0001), while the NS and PSC were only marginally, without statistical significance. Machine learning could also contribute to an early diagnosis of cataracts, with the presence of diabetes, duration of diabetes, or diabetic polyneuropathy (PND) having the highest importance for a successful classification. These findings suggest that diabetes may impact the type of cataract that develops, with CC being notably more prevalent in diabetic patients. This has important implications for screening and management strategies for cataract formation in diabetic populations.