About 2-3% of the worldwide population develop kidney stones. The stone can be treated with extra corporeal shock wave lithotripsy (ESWL). For the treatment the stone has to be positioned in the focus of the shock wave source. Patient positioning is critical, because misalignment will lead to ineffective treatment and, in the worst case, destroy healthy tissue. The alignment of the shock wave focus with the treatment volume is coordinated with X-ray and/or ultrasound imaging. Due to breathing and other movements of the patient, the positioning must be monitored periodically, often with X-rays. Care must be taken not to overexpose the patient beyond the permissible dose. Ultrasound diagnosis and imaging on the other hand is harmless but needs to have trained operators. The goal is to ease and improve ultrasound diagnosis by using the spatial and temporal information contained in the ultrasound images. Based on spatial and temporal image information from ultrasound, a Convolutional Neural Network (CNN) stabilizes the image and assists the operator to identify the stone by highlighting the kidney and stone as an overlay in the image. This research aims to prove our theoretical assumptions that the aggregation of spatial and temporal information in sequential ultrasound images in combination with CNNs leads to more stable and better recognizable ultrasound images. At the same time, the CNN should enable a segmentation of kidney and stone.