SRL-Net: Application of Convolutional Neural Networks in Lacrimal Duct Obstruction Diagnosis

CT-DCG is a common dacryocystography method used for diagnosing lacrimal duct obstruction (LDO) in patients, whose interpretation relies on the expertise of doctors.In recent years, deep learning technology has seen widespread application in medical imaging. However, due to the relative scarcity of CT-DCG imaging data, challenges remain in applying deep learning network models to diagnose lacrimal duct obstructions. Our goal is to propose a deep learning model called SRL-Net based on CT-DCG imaging data to improve the accuracy of lacrimal duct diagnosis. SRL-Net achieved an accuracy of 86.39%, a recall rate of 67.26%, and a precision of 67.23% on a CT-DCG dataset for lacrimal duct obstructions.Compared to other representative deep learning network models, SRL-Net demonstrates superior diagnostic performance on the CT-DCG dataset.