AI trained to analyse cancer images
Researchers at the Faculty of Information Technology developed a neural computing model, which, according to tests, is able to determine, for example, the T-cell count in cancer tissue based on nothing but a digital image with an error margin of a few percent.
The research data consisted of 5 523 images of intestinal cancer tumours, with the image size of 1 472 x 921 pixels. The T-cell count of each image was previously determined by histopathologists using visualisation software. Computing the cell count was successful in 90 % of the cases.
In a small number of images, errors occurred in the determination process and the neural network was unable to recognize the cells correctly during the computing. This might be due to potential inaccuracies in the data used to train the neural network. Thus, the accuracy of the neural networks will be improved with additional samples and the results might eventually become more accurate compared to the current methods.
With the help of AI, certain traditionally labour intensive processes for humans, such as counting T-cells in examining the properties of tumours, can be sped up. This allows for doctors and pathologists to dedicate more of their time to tasks relevant to the treatment and researching the disease.
In the future, with the help of these methods the utilisation of data collected, for example, in biobanks can be enhanced, thus, assisting in the development of cancer treatment.
- The training of the neural networks took place in a closed AI computing environment built for analysing social and health care data at the Faculty of Information Technology of Ģֱ. There the researchers have access to a powerful IBM Power 9 AI computer. With the help of neural networks, one sample image can be analysed in seconds in the computing environment, meaning that in one day even thousands of samples can be analysed. However, the results from our pilot do still require validation using an independent dataset, states the leader of the research, Adjunct Professor Sami Äyrämö.
- AI based digital pathology offers many new possibilities. Biobanks and hospitals can send their histopathological samples in digital form for analysis to a specialised centre that can provide results very quickly. In the beginning, the centre could serve customers on a national level, and in the future, on an international level, envisages the leader of the research project, Professor Pekka Neittaanmäki.
The research was conducted in collaboration with the Central Finland Health Care District within the research projects “Value From Health Data with Cognitive Computing” and “Watson Health Cloud Finland” with funding from Business Finland. At this point, the research results have been published in the university’s series of scientific publications and will be sent to an international scientific publication for evaluation.
Additional information:
Professor Pekka Neittaanmäki, pekka.neittaanmaki@jyu.fi, Tel: +358 40 550 7005, Faculty of Information Technology, Ģֱ
Adjunct Professor Sami Äyrämö, sami.ayramo@jyu.fi, Tel: +358 50 3255 685, Faculty of Information Technology, Ģֱ