Machine Learning Based Techniques for Detection of Renal Calculi in Ultrasound Images

The Ultrasound imaging is a non-invasive procedural technique which is used for detection of kidney diseases in the medical/clinical practice. This work emphasizes on different preprocessing methods to remove the speckle noise, embedded in kidney Ultrasound images. Preprocessing filters like, adaptive median and wiener are applied to both normal and renal calculi US images, evaluated for noise variance ranging from 0.01 and 0.08 against the parameters like Signal to Noise Ratio (SNR), Peak Signal to Noise Ratio (PSNR), Root Mean Square error (RMSE), Mean Squared Error (MAE), to determine the optimum noise variance value to be considered in preprocessing the Ultrasound images. The work also recommends adaptive median filter applied for kidney Ultrasound images with experimental results indicates increase in Peak Signal to Noise Ratio up to 33.51 dB, compared to Weiner filter. The next step is to select the best classifiers like Support Vector Machine with kernels, Multi-Layer Perceptron to preprocessed Ultrasound kidney images to estimate the accuracy, Recall, F1 score and precision. The experimental results obtained by Support Vector machine with poly kernel reaches an accuracy of 81.1% and are compared with results obtained from similar works. � 2021, Springer Nature Switzerland AG.