Overview / Usage

There are about half a million new cases of kidney cancer each year, and surgery is its most common treatment. Shape, size and morphology of kidney and kidney tumor vary among cases. Surgical outcomes and surgical planning can be related to tumor morphology. Automatic semantic segmentation of kidney and tumor can be used to analyse the tumor morphology. However, the accuracy of segmentation suffers due to the morphological heterogeneity of kidneys and tumors. The goal of this challenge is to accelerate the development of reliable kidney and kidney tumor semantic segmentation methodologies. For the task of semantic segmentation of kidneys and tumors, the organisers have produced ground truth semantic segmentation for arterial phase abdominal CT scans of 300 unique kidney cancer patients who underwent partial or radical nephrectomy at their institution. 210 of these have been released for model training and validation, and the remaining 90 will be held out for objective model evaluation.

Methodology / Approach

We will attempt to achieve the objective of this challenge by implementing 2-D convolution neural networks (CNNs). Combination of Densenet and UNET and different ConvNet will be explored. Training and validation of CNNs will be done slices wise. 

The final leader board will be prepared by the organiser on the basis of the test data set which is not seen by the models. 

Technologies Used

CNNs will be developed using tensorflow, Keras. Preprocessing will involves numpy, scikit-learn in python. 

Other links

• Manuscript
• KiTS 2019 challenge home page

Collaborators