Multi-class motion-based semantic segmentation for ureteroscopy and laser lithotripsy

Kidney stones represent a considerable burden for public health-care systems. Ureteroscopy with laser lithotripsy has evolved as the most commonly used technique for the treatment of kidney stones. Automated segmentation of kidney stones and laser fiber is an important initial step to performing any automated quantitative analysis of the stones, particularly stone-size estimation, that helps the surgeon decide if the stone requires more fragmentation. Factors such as turbid fluid inside the cavity, specularities, motion blur due to kidney movements and camera motion, bleeding, and stone debris impact the quality of vision within the kidney and lead to extended operative times. To the best of our knowledge, this is the first attempt made towards multi-class segmentation in ureteroscopy and laser lithotripsy data. We propose an end-to-end CNN-based framework for the segmentation of stones and laser fiber. The proposed approach utilizes two sub-networks: HybResUNet, a version of residual U-Net, that uses residual connections in the encoder path of U-Net and a DVFNet that generates DVF predictions which are then used to prune the prediction maps. We also present ablation studies that combine dilated convolutions, recurrent and residual connections, ASPP and attention gate. We propose a compound loss function that improves our segmentation performance. We have also provided an ablation study to determine the optimal data augmentation strategy. Our qualitative and quantitative results illustrate that our proposed method outperforms SOTA methods such as UNet and DeepLabv3+ showing an improvement of 5.2% and 15.93%, respectively, for the combined mean of DSC and JI in our invivo test dataset. We also show that our proposed model generalizes better on a new clinical dataset showing a mean improvement of 25.4%, 20%, and 11% over UNet, HybResUNet, and DeepLabv3+, respectively, for the same metric.