MIT researchers have announced a new system based on artificial intelligence that helps reduce the toxic dose of chemotherapy and radiotherapy in glioblastoma, the most aggressive form of brain tumor.
These are automated learning techniques for the treatment of glioblastoma, a malignant tumor that occurs in the brain or in the spinal cord, with a prognosis for adults not older than five years and taking maximum doses of drugs that cause debilitating side effects in patients.
How Is Cancer treated with Deepmind Artificial intelligence?
Making dose regimens less toxic but still effective will be presented at Stanford University’s Conference 2018 Machine Learning for Healthcare next week. The model analyzes the treatment regimens currently used and iteratively adjusts the doses until an optimal treatment plan is found, with the lowest possible dose strength and frequency, that should reduce the size of the tumors to a level comparable to conventional dosing schedules.
It was simulated with 50 patients and the potency was reduced to a quarter or half of all doses, while the tumor shrinkage potential remained constant.
The technique is called Reinforced Learning (RL), a method inspired by behavioral psychology that performs “actions” in an unpredictable and complex environment to achieve a desired “result”. Each time an action is completed, the agent receives a “reward” or “punishment”, depending on whether the action is working towards the result or not. The processor then adjusts the actions accordingly to achieve this result.
DeepMind was used for the training that made headlines in 2016 to defeat one of the world’s best human players in the Go game, but it was also used in the autonomous automotive sector, e.g. when entering traffic or parking.
Artificial intelligence is becoming more and more ubiquitous in our daily lives, although we often do not perceive it.