Transformer-Enhanced U-Net with Transfer Learning for Accurate Detection of Eye Coloboma

Eye coloboma is a type of congenital eye abnormality that causes developmental abnormalities of the eye which may culminate in eye disorders that fundamentally derange eyesight. Fractional and timely diagnosis remains highly important to its proper managing and treating. This work proposes a model with a Transformer-Enhanced U-Net design implemented with Transfer Learning to enhance the ability of the model to detect and segment eye coloboma from medical imaging data. This work builds upon the benefits of U-Net which has shown promising results especially in the biomedical image segmentation problem through addition of transformer modules that introduce long range dependencies and contextual information. This enhancement makes helps the model to pay more attention to such minor aspects as coloboma and general diagnostic accuracy consequently. Thanks to Transfer Learning with a pretrained backbone, for example Efficient Net the system is provided with Feature Representations recovered from large datasets that are already available and training time is drastically cut short while the performance of the system is not compromised at all.The approach includes fine-tuning of Transformer-Enhanced U-Net model using a selected set of eye images, labelled for coloboma existence. By including the self-attention mechanism, the model is capable of focusing on the critical regions in the image, increasing its responsiveness to any form of coloboma. Preliminary tests show that this scheme performs better than the standard convolutional networks for segmenting complex regions by minimizing false negatives. Further, the model’s flexibility to different imaging conditions demonstrates its suitability in realistic clinical applications. The objective of this research is to come up with a dependable, Auto-Generated diagnostic tool that will help the ophthalmologists in the early diagnosis of the eye related illnesses hence making increased positive outcomes for eye patients. The Transformer-Enhanced U-Net is an enhancement that brings a paradigm shift in utilizing state-of-art deep learning technology for accurate identification of eye coloboma.