Detection and analysis of kidney location and kidney stones in medical abdominal images using image processing methods

In this study, a method that performs automatic kidney and kidney stone detection in abdominal Computed Tomography (CT) images without user interference is proposed by combining Aggregate Channels Features (ACF), thresholding and morphology methods. ACF method performs automatic kidney detection by training with less labelled data. In addition, ACF does not require any pre-processing or image enhancement. By applying thresholding and morphological operations to the kidney regions detected with ACF, stones are detected only in the kidney regions. In this way, stone-like objects outside the kidney boundary in abdominal CT images are not evaluated. The ACF method was applied to a dataset consisting of 150 abdominal CT images. The performance of the method is evaluated by confusion matrix. Experimental results show that ACF, thresholding and morphology methods successfully detect kidneys and kidney stones in abdominal CT images.