University of Melbourne and Western Health researchers have developed a new artificial intelligence tool to prevent cancer patients from receiving incorrect doses of chemotherapy.
Chemotherapy is typically dosed by calculating the patient's body surface area using their height and weight. But, researchers say this method is highly inaccurate as it doesn't factor in body composition.
The researchers have now developed an AI algorithm which uses image recognition technology and machine learning to accurately predict the precise amount of chemotherapy a patient requires based on their body make-up.
Currently, around 60 percent of colorectal cancer patients who undergo chemotherapy are either overdosed or underdosed.
Professor Justin Yeung, who is also a consultant colorectal surgeon at Western Health, said overdosing can cause a raft of side-effects ranging from mild to severe.
"Major side-effects include immunosuppression, heart attacks and chest infections," Professor Yeung said.
"Many incorrectly dosed patients stop treatment too early due to these debilitating side-effects, but if there's no other cancer treatment available for them, this decision can drastically reduce their chance of survival."
Professor Yeung's team utilised data including CT scans from a cohort of more than 1,000 colorectal cancer patients at Western Health to help train and test the algorithm.
The algorithm analysed the scans and found the patients' body compositions (percentages of fat, bone and muscle) determined how the chemotherapy drug was metabolised and stored in their bodies.
Using these findings, the algorithm can now calculate tailored chemotherapy doses for patients using their body compositions.
"Our algorithm was able to produce accurate chemotherapy dosing for 84 percent of those patients which is a significant improvement over current methods of dosing," he said.
"It doesn't make logical sense for two patients with significantly different fat and muscle ratios to be given the same chemotherapy dose just because they have similar body surface areas.
"For example, a sumo wrestler and a body builder who have comparable body surface areas, would theoretically be given the same chemotherapy doses, however as their body compositions are vastly different, they would likely develop different degrees of toxicities."
Recognising the clinical need for a patient tailored chemotherapy dosing solution, the team has formed a startup called 'PredicTx Health' to translate their research into a product.
The team recently secured $499,760 in grant funding through Australia's Economic Accelerator program to fund an observational trial, the development of an AI algorithm and a prototype UI, as well as technology to integrate the solution into future health systems.
The observational trial only applies to colorectal cancer patients.
The final step will be a clinical trial run at several health services including Western Health, to compare the technology with current chemotherapy dosing methodology.
