Myopia is a widespread disease that can lead to a variety of complications, of which myopic maculopathy is an important potential cause of blindness. Fundus photography is a fundamental technique for detecting macular disease, this non-invasive, easily accessible and widely used imaging method provides valuable data for exploring a range of ophthalmic diseases. With the development of machine learning, particularly deep learning, these datasets are now extensively used to analyse ophthalmic images and detect diseases like myopic maculopathy. This research work aims to propose a robust deep learning approach for the automated five-class classification of myopic maculopathy, achieving optimal effectiveness with limited and imbalanced data to address the challenge posed by the Maculopathy Analysis Challenge (MMAC) 2023. The proposed method employed various data augmentation techniques, introduced an efficient hyperparameter optimisation strategy and implemented a ConvNeXt-based model. This method demonstrated notable performance on the MMAC 2023 dataset, achieving the highest Quadratic Weighted Kappa (QWK), highlighting the model’s sensitivity to class order and its potential application value in related fields. Our algorithm achieved a QWK of 0.9004, a macro F1 score of 0.7634, and a macro specificity of 0.9443 on the test set, ranking within the top 3 overall.