Machine Learning Approaches for Kidney Diagnostics Using SPECT Imaging
Abstract
Purpose
A gamma camera is used as part of a standard process at nuclear medicine centers or hospitals to measure the glomerular filtration rate (GFR) in chronic kidney disease (CKD) cases. However, the gamma camera is not able to offer correct disease stages. Consequently, the purpose of this study was to use an artificial neural network (ANN) to determine whether CKD is in a normal or pathological stage, depending on the value of GFR.
Methods
To determine the renal test, a gamma camera was used to scan the 250 (188 Training, 62 Testing) kidney patients who had ultrasonography at our nuclear medical centre. A 99mTc-DTPA injection was given to the patients before to the scanning process. The gamma camera was used to calculate the radioactive counts before and after the syringe was inserted into the patient's vein. Depending on the value of GFR in the output layer, the artificial neural network diagnoses CKD as normal or abnormal using the softmax function with cross-entropy loss.
Results
According to the findings, the suggested ANN model's accuracy for K-fold cross-validation was 99.20%. The results showed a sensitivity of 99.10% and a specificity of 99.20%. 0.9994 was the area under the curve (AUC).
Conclusion
The suggested model can distinguish between the normal and abnormal of CKD using an artificial neural network. The proposed model may be upgraded to detect normal or abnormal of chronic kidney disease (CKD) with an acceptable GFR value after it has been used clinically.