Researchers at Radboudumc have developed a system that can recognize the aggressiveness of prostate cancer with the help of artificial intelligence (AI). According to the researchers, the system can better determine the degree of aggressiveness of the tumor than pathologists.
Normally a pathologist examines pieces of tissue from the prostate and then assesses how aggressive the cancer is. But this process is subjective, the researchers say: "Whether and how a patient is treated may depend on the pathologist who assesses the piece of tissue."
With the algorithm it is possible in the future to determine aggressiveness without the subjective gaze of a pathologist. The researchers trained the system through deep learning with 5,759 pieces of tissue.
Deep learning is a technique that is very similar to the way people learn. The computer receives a number of sample photos with information - in this case, how aggressive the cancer is - and learns from it. After a certain number of examples, the algorithm can determine the aggressiveness itself based on a new image.
'System is comparable to the work of experienced pathologists'
After the training, the algorithm was compared to the work of fifteen pathologists. According to researcher Wouter Bulten, the system was able to better recognize the aggressiveness of prostate cancer than ten of these pathologists. "Our system is comparable to highly experienced pathologists."
The algorithm is not yet used. The researchers want to organize an international competition in collaboration with Swedish researchers from the Swedish medical university Karolinska Institutet and Kaggle, a subsidiary of Google.
In addition, other researchers are challenged to beat the existing algorithm. That information can then be used to improve the technology.