Optical coherence tomography (OCT) is an advanced tool used to view the layers of the retina with precision. Consequently, unhealthy retinal layers can be detected in OCT images. The automated detection of unhealthy OCT images can significantly impact eye disease control applications. This work proposes a machine learning approach that utilizes simple features related to the symmetries and smoothness of the inner limiting membrane (ILM) and retinal pigment epithelium (RPE) layers to detect unhealthy images. Several classifier models were tested, with the Light Gradient Boosting Machine (LGBM) yielding the highest accuracy of 96% and a low false negative rate of 6%, while maintaining a very high precision of 98% and a recall of 94%. The proposed method outperforms a baseline method that relies on vertical line profile features, improving accuracy, precision, recall, and F1-score by 5%, 8%, 3%, and 5%, respectively. Additionally, it reduced the baseline method’s false negative rate of prediction by 3%. The excellent results of this work demonstrate its effectiveness in classifying unhealthy cases versus healthy cases, thereby reducing the workload for ophthalmologists and quickening the patient prescreening process.