AUTHOR=Liu Yuliang , Zhang Fenghang , Gao Xizhan , Liu Tingting , Dong Jiwen 

TITLE=Lesion-aware attention network for diabetic nephropathy diagnosis with optical coherence tomography images

JOURNAL=Frontiers in Medicine

VOLUME=Volume 10 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2023.1259478

DOI=10.3389/fmed.2023.1259478

ISSN=2296-858X

ABSTRACT=Diabetes mellitus is a common chronic metabolic disease, which can cause harm to the human body in various ways. Diabetic nephropathy is one of its serious complications, and long-term hyperglycemia can cause damage to kidney function. In order to prevent further progression to end-stage renal disease, early screening for diabetic nephropathy is crucial to detect the disease and take timely treatment measures. Diabetes can also lead to retinal changes to some extent. Since the pathophysiological mechanisms and clinical changes of diabetic nephropathy and diabetic retinopathy are closely correlated, they may be able to predict each other. This provides the potential for non-invasive early screening diabetic nephropathy by analyzing retinal images. Deep learning-based methods can assist in the screening of diabetic nephropathy by analyzing retinal lesions, which is beneficial for early detection and treatment. In this paper, we propose a lesion-aware attention module. It improves the model's attention to retinal regions susceptible to lesions and better captures relevant features. Experimental results on clinically-acquired OCT datasets show that our method can achieve effective assisted screening.