Artificial Intelligence can be trained to determine location, size of tumours

Algorithms are already being used to analyse medical imaging data; it can be made more refined 

Artificial Intelligence can be trained to analyse images generated by positron emission tomography (PET) and computed tomography (CT) in a way that they can determine the location and size of tumours.

This is the result of AutoPET, an international competition in medical image analysis, where researchers of Karlsruhe Institute of Technology (KIT) were ranked fifth. The seven best autoPET teams report in the journal Nature Machine Intelligence on how algorithms can detect tumor lesions. Imaging options such as PET and CT are widely used for the diagnosis of cancer. To decide the right therapy, precision in identifying the contours and location of a tumour is very important. Cancer patients can have hundreds of lesions.

“Automated evaluation using an algorithm would save an enormous amount of time and improve the results,” explains Professor Rainer Stiefelhagen, Head of the Computer Vision for Human-Computer Interaction Lab at KIT.

AI is already being used for the diagnosis of cancer using imaging techniques but it is still a nascent technology in the field. However this new paper holds out prospects for making it more efficient and targetted to better treatment outcomes. 

As the researchers explain in their publication, an ensemble of the top-rated algorithms proved to be superior to individual algorithms. The ensemble of algorithms is able to detect tumor lesions efficiently and precisely. 

“While the performance of the algorithms in image data evaluation partly depends indeed on the quantity and quality of the data, the algorithm design is another crucial factor, for example with regard to the decisions made in the post-processing of the predicted segmentation,” explains Stiefelhagen.