Combining Optical Sensing with Deep Learning for Enhanced Non-Invasive Kidney Function Measurement

The difficulty of non-invasive kidney function monitoring is addressed in the paper by using a deep learning model that uses optical sensors to estimate Glomerular Filtration Rate (GFR). The model showed impressive performance by predicting GFR with a Mean Absolute Error (MAE) of 2.3. It also registered 0.84 F1 Score, 0.78 precision, 0.92 recall, and 0.85 accuracy. These measurements demonstrate how well the model divides renal disease and predicts GFR. With an accuracy of less than $\mathbf{5 0 \%}$ and an MAE of 4.7, the proposed method significantly outperforms the baseline technique. Furthermore, the model outperforms earlier methods in terms of accuracy and precision, as shown by a comparison with earlier studies. The findings support the strong efficacy of the suggested approach in offering a trustworthy, non-invasive substitute for evaluating kidney health, with possible clinical setting applicability.