MIT researchers have developed artificial intelligence to improve the life of patients suffering from glioblastoma cancer. Glioblastoma, which is also known as glioblastoma multiforme (GBM) is a malignant tumor in the brain or spinal cord.
It is the most aggressive cancer that begins within the brain and particularly there is no clear way to prevent it. Out of 100,000, about 3 people suffer from this disease in a year, mostly male after 64 years of age.
You will be surprised by knowing that without treatment patient suffering from Glioblastoma can survive for three months only. The most common length of survival following diagnosis is 12 to 15 months, mostly patient live for less than five years.
Those fives years aren’t full of joy, the patient feels tremendous pain in every second he passes. Doctors usually try to minimize the tumor to reduce the pain.
They often prescribe a combination of radiation therapy and the appropriate combination of drugs to shrink the size of the tumor. But such therapies and methods of treatment mostly give trouble to the patient instead of reducing the pain, it has debilitating side effects for patients.
In order to avoid its side effects and to determine the appropriate doses of the drug to effectively shrink glioblastoma patients’ tumors, MIT researchers developed this artificial intelligence.
It analyzes the important metrics based on the clinical records of patients and recommends a regiment designed to shrink brain tumors while lessening the side effects on patients. They plan to present their research at Stanford University’s 2018 Machine Learning for Healthcare conference.
MIT Researchers first created a testing group of 50 simulated glioblastoma patients those who had previously undergone treatment. The team ran 50 new patient profiles through the model to see what the machine would recommend. T
hey asked AI to recommend doses of several drugs typically used to treat glioblastoma.
At first, the AI recommends a quarter or half dosage at various intervals. Sometimes, it projects a tumor can still shrink even if the patient only comes in for treatment less often than the standard 30 days.
Some simulations recommend administering drugs only every six months to shrink tumors. Many times, it skipped doses altogether, scheduling administrations only twice a year instead of monthly.
The AI conducted 20,000 trial-and-error tests for each patient to come up with an optimal treatment plan during the training process. The AI has a train to think in the way how it can get the same or better outcome with much less chemotherapy and radiation for patients.
The model was trained to propose the dosage for four chemotherapy drugs temozolomide, procarbazine, lomustine, and vincristine.
This AI works on a program which consists of an agent that recommend appropriate doses of the drug to shrink the tumor and to avoid its side effects. It works on the reward system.
If it prescribed a tumor-shrinking dosage, the agent is rewarded with greater weighting. Whereas if the AI prescribed the maximum dose all the time, it received a penalty and has to adjust its recommendations.
MIT researchers repeat the program thousands of time. Over time, the AI becomes highly optimized in creating treatments for glioblastoma.
“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,” Shah concluded.
“If all we want to do is reduce the mean tumor diameter, and let it take whatever actions it wants, it will administer drugs irresponsibly,” Shah said.
“Instead, we need to reduce the harmful actions it takes to get to that outcome.”
This AI has only been tested in simulations so far. It still needs to undergo further testing and vetting by the Food and Drug Administration (FDA) before doctors could put it into actual practice.
But if it passes those tests, it could eventually help people with glioblastoma attack their brain tumors without causing them more pain in the process
Read also, you may like it :