According to new research led by Massachusetts General Hospital, it has been found that magnetic resonance imaging (MRI) and artificial intelligence (AI) can be used to detect early signs of tumour cell death in response to novel virus-based cancer therapy.

This research has been published in the Nature Biomedical Engineering Journal. Recently, a promising therapeutic virus that selectively kills cancer cells while sparing normal tissue has sparked hope for treating aggressive brain tumours. To further optimise the virus-based therapy, frequent non-invasive monitoring of the treatment response must be performed. This monitoring is crucial for understanding the interactions between the virus and cancer cells, such as the extent of virus spread within the tumour and therapeutic response.

The researchers used quantitative molecular MRI images to measure multiple tissue properties, including tissue pH and protein concentration that are altered with cell death. This method allowed therapeutic response monitoring much earlier than with previous techniques. The treatment responses were visible just 48 hours after viral therapy, long before any changes in tumour volume were observed.

"We programmed an MRI scanner to create unique signal "fingerprints" for different molecular compounds and cellular pH. A deep learning neural network was then used to decode the fingerprints and generate quantitative pH and molecular maps," said Christian Farrar, PhD, an investigator and faculty at the Athinoula A. Martinos Center for Biomedical Imaging. "The MRI molecular fingerprinting method was validated in a mouse brain tumour study where the tumours were treated with a novel virus-based therapy that selectively killed cancer cells," Farrar added.

To maximise the efficiency of this treatment approach, the researchers developed a method for the detection of tumour cell death caused by the virus. This has allowed for the early and rapid detection of treatment-responsive tumour regions. Recently, researchers have implemented this method to quantify cellular pH and molecular compounds in the healthy human brain. Future investigation of this approach in human brain tumour patients would help to optimise these virus-based therapies.

"This study demonstrates the strength and promise of implementing computerised AI-based technology in medicine for the noninvasive investigation of biological processes that underlie disease," said Or Perlman, PhD, a research fellow at the Athinoula A. Martinos Center for Biomedical Imaging.

"One of the most interesting and key components for the success of this approach was the use of simulated molecular fingerprints to train the machine learning neural network. This concept could potentially be expanded and investigated for solving other medical and scientific challenges," Perlman added.

This study described a new method for detecting tumour cell death non-invasively using MRI. The capacity to do this could be useful for non-invasive monitoring of cancer treatment, potentially improving patient care and tailoring the treatment to an individual patient. The same approach might also be beneficial for detecting and characterising other medical conditions where elevated cell deaths occur, such as stroke and liver disease. While the study was mainly validated using a brain tumour model, the researchers have demonstrated the ability to use the same method for producing quantitative pH and molecular maps in rat stroke models and healthy humans. In the future, they plan to further explore the applicability of this non-invasive imaging approach in patients with brain tumours and stroke.