LUMC investigates how AI can optimize DNA testing for endometrial cancer

By: Lennart 't Hart

The characteristics of the tumor play an important role in the treatment of uterine cancer. Pathologists are increasingly being asked to perform DNA testing in order to tailor treatment as closely as possible to the individual patient.

An important indicator (biomarker) is a change (mutation) in the POLE gene. Patients with such a mutation have an excellent prognosis and usually do not need to undergo additional treatment, such as radiation or chemotherapy, after surgery. “So this is good news for patients with a POLE mutation,” says project leader Prof. Dr. Tjalling Bosse. “That is precisely why it is important that we perform these tests to identify all patients with a POLE mutation.”

DNA test not always offered

Currently, only a DNA test can determine whether a tumor contains a POLE mutation. Although this test is reliable, it requires capacity and expertise that are not available everywhere. As a result, the test is not always performed on every patient, both in the Netherlands and internationally. “Consequently, some patients miss out on important information that could directly influence their treatment,” says Bosse.

AI supports the pathologist

Under the leadership of the LUMC, experts from the AIRMEC AI research team developed the POLARIX AI model. Based on digital images of tissue, this model predicts which patients have an increased risk of a POLE mutation. Only these patients will then undergo a confirmatory DNA test. This can significantly reduce the number of DNA tests required, without compromising the quality of the diagnosis.

Research across multiple hospitals

The next step is to investigate whether POLARIX also works safely and reliably in daily practice. The model will therefore be tested in multiple hospitals in the Netherlands and abroad. In doing so, the researchers will examine not only the accuracy of the predictions but also whether the model continues to perform well across different patient groups in various healthcare settings. This active deployment of in-house AI models aligns seamlessly with the LUMC’s ambitions and fits within the “Digitally Connected” program.

Three-quarters of DNA tests perhaps unnecessary

If POLARIX proves successful, it is expected that approximately three-quarters of current DNA tests can be avoided. Only patients with an increased risk of a POLE mutation would then need to undergo an additional DNA test.

The potential impact extends beyond the Netherlands. In countries where DNA diagnostics are limited, POLARIX can help ensure that available capacity is used in a targeted manner. In this way, the project also aligns with the platform for international collaboration, LUMC Global, which recently invested in the implementation of POLARIX in South Africa.

A next step for AI in healthcare

With the Dutch Cancer Sociecty grant the researchers aim to further develop POLARIX for use in daily patient care. If the research is successful, it could contribute to better access to molecular diagnostics and more personalized treatment for women with uterine cancer.

In addition, the project can serve as a model for the responsible implementation of artificial intelligence in pathology. Because, as Bosse notes, “A biomarker only has real value if every patient who can benefit from it actually has access to it.”