Scientists at the Institute of Neuropathology and the Berlin Institute of Health at Charité University Hospital in Berlin, Germany, have developed an artificial intelligence (AI) model that can identify more than 170 tumor types from all organs in the human body, with particular focus on difficult-to-diagnose brain tumors.
Recently published in the journal Nature Cancer and discussed in a paper titled “crossNN is an explainable framework for cross-platform DNA methylation-based classification of tumors,” the new AI model, called crossNN, can analyze certain modifications in the genetic material of a tumor, called the epigenetic characteristics, and diagnose the type of tumor to determine the best treatment option available.
“Hundreds of thousands of epigenetic modifications act as on and off switches for individual gene sections,” explained Philipp Euskirchen, MD, head of the study and scientist at the Berlin site of the German Cancer Consortium (DKTK) and the Institute of Neuropathology at Charité.
“Their patterns form a unique, unmistakable fingerprint. In tumor cells, the epigenetic information is altered in a characteristic way. Based on their profiles, we can differentiate between tumors and classify them.”
Currently, the diagnosis of tumors is done using a combination of AI models—with the ability to compare large, complex datasets in short amounts of time, different DNA sequencing methods, and epigenetic analyses that remain limited to known genetic patterns of individual tumors.
While in many cases, this combination of diagnostic procedures helps physicians form comprehensive treatment plans for patients, sometimes, physicians are confronted with tumors that are very difficult to diagnose. When it comes to brain tumors, for instance, conventional microscopic diagnostics can be tricky, as they require surgery to extract tissue from the tumor. The researchers’ new AI model, on the other hand, can use cerebrospinal fluid for diagnostics instead, which is obtained more easily.
“[…] Our aim was to develop a model that accurately classifies tumors, even if they are only based on parts of the entire tumor epigenome or the profiles were collected by means of different techniques and varying degrees of accuracy,” explained Sören Lukassen, PhD, bioinformatician and Head of the Medical Omics working group at the Berlin Institute of Health at Charité.
crossNN is based on a simple neural network and was trained with more than 2,800 epigenetic samples from 82 tumor types, and subsequently tested on over 5,000 tumors.
“Our model allows a very precise diagnosis of brain tumors in 99.1 percent of all cases and is more accurate than the AI solutions at work to date,” explained Euskirchen. “In addition, we were able to train an AI model in the same way that can differentiate between over 170 tumor types from all organs, while achieving accuracy of 97.8 percent. This means that it can be used for cancers of all organs, in addition to the relatively rare brain tumors.”
Lukassen added: “Although the architecture of our AI model is far more simple than previous approaches and therefore remains explainable, it delivers more precise predictions and therefore greater diagnostic certainty.”
In the near future, the research team, in cooperation with the DKTK, wants to run clinical trials across all eight DKTK locations in Germany to test the accuracy of crossNN further and, one day, introduce the AI model into routine cancer care to make the diagnosis of tumors and establishment of precision treatment plans simpler for physicians.
