A leading pharmaceutical company needed a model to automatically segment the liver and analyze it using radiomics based on CT scans. The premise of the project was that the algorithm to be developed would automatically detect cirrhosis and classify the severity of the disease.
What we did
We started the project with a thorough research in scientific literature. We know how important a scientific approach, based on thorough knowledge of a given field, is when implementing medical projects. Therefore, we checked what the current state of knowledge is in the field of liver cirrhosis. Then we chose the right methods and architecture, on the basis of which we started to work on the algorithm. We developed and trained the algorithm from scratch. We were in possession of studies provided by the client. The studies had not been labeled by clinical experts. We therefore did that ourselves.
The algorithm we have developed segments the liver and indicates whether the liver shows signs of cirrhosis or not. In addition, thanks to the use of radiomics, it classifies and groups data indicating the severity of the disease. The whole process is automatic. Thus, the client is provided with a tool that speeds up the segmentation process and, thanks to the radiomic analysis, assesses the condition of the liver and has access to easily interpretable data.
Image credit: medicaldecathlon.com