Real-time gene profiling could improve brain cancer surgery

A new machine learning tool can profile the genetics of brain tumour cells in real-time, informing rapid treatment decisions during surgery.

A tool to predict brain tumour genetics during surgery has been developed by researchers at Harvard Medical School in Massachusetts. The tool, called CHARM – 'Cryosection Histopathology Assessment and Review Machine' – can predict the genetic profile of glioma cells, giving surgeons important information on the aggressiveness of the cancer and the expected response to treatment. This allows decisions on tumour removal and drug treatments to be made during surgery, instead of waiting several days for genetic profiling.

'Right now, even state-of-the-art clinical practice cannot profile tumours molecularly during surgery. Our tool overcomes this challenge by extracting thus-far untapped biomedical signals from frozen pathology slides', said Dr Kun-Hsing Yu, co-author of the research published in Med.

Currently, genetic profiling requires samples to be taken, frozen and examined under a microscope to determine the characteristics of the cancer. CHARM can provide information on the aggressiveness of the cancer and the likelihood it can be treated with drugs. Surgeons use this information to decide how much tissue to remove during surgery; an important decision which affects the cognitive function of the patient and the chance of the cancer coming back. Having genetic information available in real-time can help surgeons to remove the optimal amount of tissue and achieve the best outcome for the patient.

CHARM uses machine learning to predict genetically-important mutations in tumour cells. Tissue samples are cut into thin slices, called cryosections, which are examined by the tool and used to predict which genetic mutations are present in the cells. CHARM showed 93 percent accuracy when predicting mutations, and can identify three major types of glioma. It examines cell density in the tissue sample and uses this information to predict the aggressiveness of the cancer.

While CHARM was trained on glioma cells, the researchers highlight that the same approach could be used to create genetic profiles for other cancer types.

CHARM will now be tested in real-world clinical settings. The research team hope to demonstrate its usefulness in practice and have the tool approved by the FDA, so that clinicians can start to use it in surgical practice.