MIT researchers use AI model to determine dosage for cancer patients

MIT researchers are using an artificial intelligence (AI) model that would help determine the correct drug dosage and, in turn, reduce debilitating side effects for brain cancer patients.

According to a release from MIT, researchers are employing novel machine-learning techniques to “improve the quality of life for patients by reducing toxic chemotherapy and radiotherapy dosing for glioblastoma, the most aggressive form of brain cancer.” The model could make the dosage regimens less toxic, but still effective.

"Powered by a 'self-learning' machine-learning technique, the model looks at treatment regimens currently in use, and iteratively adjusts the doses," the release said. "Eventually, it finds the optimal treatment plan, with the lowest possible potency and frequency doses that should still reduce tumor sizes to a degree comparable to that of traditional regimens."

Researchers conducted simulated trials on 50 patients using the machine-learning model designed treatment. In the the trials, the model either reduced the potency of the dosage or skipped it altogether, while still maintaining the same tumor-shrinking potential.

“We kept the goal, where we have to help patients by reducing tumor sizes but, at the same time, we want to make sure the quality of life—the dosing toxicity—doesn’t lead to overwhelming sickness and harmful side effects,” principal investigator Pratik Shah said in the release.

Researchers used a technique called reinforced learning (RL) in which a model learns to favor a certain behavior that leads to a desired outcome.

“The technique comprises artificially intelligent ‘agents’ that complete ‘actions’ in an unpredictable, complex environment to reach a desired ‘outcome,’” the release said. “Whenever it completes an action, the agent receives a ‘reward’ or ‘penalty,’ depending on whether the action works toward the outcome. Then, the agent adjusts its actions accordingly to achieve that outcome.”

Researches then adapted an RL model for glioblastoma treatments and used a combination of drugs, which were administered over the course of two weeks.

“As the model explores the regimen, at each planned dosing interval—say, once a month—it decides on one of several actions. It can, first, either initiate or withhold a dose. If it does administer, it then decides if the entire dose, or only a portion, is necessary,” the release said. “At each action, it pings another clinical model—often used to predict a tumor’s change in size in response to treatments—to see if the action shrinks the mean tumor diameter. If it does, the model receives a reward.”

Researchers later tested the model on 50 new simulated patients and compared results to those of a conventional regimen. They also designed the model to treat each patient individually. When the model wasn’t penalized, it gave the same dosage amount as human physicians. When it was penalized, it cut the dosage frequency and potency, while still reducing tumor sizes.

Researchers believe the model offers a major improvement over the eye-balling technique for administering doses, observing how patients respond and adjusting accordingly.

“We said [to the model], ‘Do you have to administer the same dose for all the patients? And it said, ‘No. I can give a quarter dose to this person, half to this person, and maybe we skip a dose for this person,’” Shah said in the release. “That was the most exciting part of this work, where we are able to generate precision medicine-based treatments by conducting one-person trials using unorthodox machine-learning architectures.”