Infrared Imaging With AI Used to Personalize Colon Cancer Therapy

Researchers at the Centre for Protein Diagnostics PRODI at Ruhr University Bochum, Germany, are using artificial intelligence (AI) in combination with infrared imaging to optimally tailor colon cancer therapy to individual patients. The label-free and automatable method can complement existing pathological analyses, according to a report led by Professor Klaus Gerwert in the European Journal of Cancer.

The PRODI team has been developing a new digital imaging method over the last years: the so-called label-free infrared (IR) imaging measures the genomic and proteomic composition of the examined tissue, i.e. provides molecular information based on the infrared spectra. This information is decoded with the help of artificial intelligence and displayed as false-color images. To do this, the researchers use image analysis methods from the field of deep learning.

In cooperation with clinical partners, the PRODI team was able to show that the use of deep neural networks makes it possible to reliably determine the so-called microsatellite status, a prognostically and therapeutically relevant parameter, in colon cancer. In this process, the tissue sample goes through a standardized, user-independent, automated process and enables a spatially resolved differential classification of the tumor within one hour.

In classical diagnostics, microsatellite status is determined either by complex immunostaining of various proteins or by DNA analysis. “15 to 20 per cent of colon cancer patients show microsatellite instability in the tumor tissue,” says Professor Andrea Tannapfel, head of the Institute of Pathology at Ruhr University. “This instability is a positive biomarker indicating that immunotherapy will be effective.”

With the ever-improving therapy options, the fast and uncomplicated determination of such biomarkers is also becoming more and more important. Based on IR microscopic data, neuronal networks were modified, optimized, and trained at PRODI to establish label-free diagnostics. Unlike immunostaining, this approach does not require dyes and is significantly faster than DNA analysis. “We were able to show that the accuracy of IR imaging for determining microsatellite status comes close to the most common method used in the clinic, immunostaining,” says PhD student Stephanie Schörner. “Through constant further development and optimization of the method, we expect a further increase in accuracy,” adds Dr. Frederik Großerüschkamp.

The PRODI researchers were able to access the ColoPredict Plus 2.0 molecular registry, a non-interventional, multi-center registry study for patients with early-stage colorectal cancer, to develop this diagnostic approach. “The ColoPredict registry also enables a more targeted therapy for patients through the targeted analysis of biomarkers. Thus, the registry recently serves as a study platform for precision onc approaches,” says Anke Reinacher-Schick. In addition to providing tissue samples, the registry offers a sound database of prognostically and therapeutically relevant baseline characteristics. “In such a project, it is of immense importance to be able to draw on an excellent cohort and pathological expertise,” emphasizes Klaus Gerwert. “Our work on the classification of microsatellite status in colon cancer patients is based on one of the largest cohorts we have published to date and clearly demonstrates the potential for use in translational cancer research,” says Andrea Tannapfel.