Automatic classification of kidney stone components based on smartphone microscopy and the GoogLeNet model

Key Points

• The research aims to create a system for classifying kidney stone compositions using a smartphone microscope and a deep learning model.
• Developed a smartphone-based microscope for capturing images of kidney stones.
• Applied the GoogLeNet deep learning model for image classification.
• Evaluated classification accuracy across different stone types, including uric acid and magnesium ammonium phosphate hexahydrate.
• Achieved an overall classification accuracy of 85.7%.
• Showed strong performance specifically in classifying uric acid and magnesium ammonium phosphate hexahydrate stones.
• Demonstrated practicality for use in resource-limited regions due to low cost and portability.