AI in immunotherapy

Immunotherapies can help beat cancer. However, many therapies do not have the same effect on all cancer patients or cause extreme side effects. The goal of a current research project is to use artificial intelligence to automatically classify tumors by their individual immune profile and microenvironment in order to recommend the best possible therapy to patients.

Artificial Intelligence in Healthcare Does AI in immunotherapy increase the chances of a cure for cancer? (Photo credit: iStock©twinsterphoto)

The idea of fighting cancer cells with the help of the patient's own immune system is not new. As early as 1867, the Bonn surgeon Wilhelm Busch documented that the tumor of a cancer patient regressed after an intentionally induced infection. Today, more than 150 years later, there has been great progress in the field of immunotherapy. This is impressively documented by the 2018 Nobel Prize in Medicine awarded to the two American and Japanese scientists James P. Allison and Tasuku Honjo.

Digital profiling as part of immunotherapy

Despite the current successes, further fundamental work is needed to be able to determine the patients who will benefit from immunotherapy. A joint research project of Ludwig-Maximilians-Universität München, Technische Universität München and industry partners Definiens AG and Morphisto GmbH aims to develop an analysis method for patients with urinary bladder and prostate cancers that will enable faster and more accurate prediction.

In the ImmunoProfiling process, the cancer is classified as a cold or warm tumor via its individual immune profile and microenvironment by analyzing image data of malignant tissue changes. In contrast to a "cold tumor", a "warm tumor" is characterized by an active immune system. In addition to histopathological images, i.e. images of pathological tissue sections, the research team also uses clinical data. A network architecture is required to process these multimodal input data.

Algorithms from the field of artificial intelligence, and here specifically the so-called "deep neural networks," offer themselves as a suitable tool. They can automatically determine the features relevant for prediction to describe the cellular microenvironment and are advantageous compared to alternative methods such as randomized decision trees. For tumor patients, the research project may mean significant improvements.

ImmunoProfiling using Neural Networks

The collaborative project "ImmunoProfiling using Neural Networks" was funded by the Bavarian Research Foundation as part of the program "High Technologies for the 21st. Century" from May 2017 to October 2019 with a total funding of €487,200.