Medical Gospel? MIT develops AI to ensure that cancer patients receive adequate treatment
Glioblastoma is a malignant tumor in the brain or spinal cord that usually occurs less than 5 years after diagnosis. These five years may be painful – in order to minimize tumors, doctors usually develop a combination of radiation therapy and drugs that can cause side effects and weak patients. Now, researchers at the Massachusetts Institute of Technology Media Lab have developed artificial intelligence (AI) that can determine the minimum dose needed to effectively reduce tumors in patients with glioblastoma. They plan to present their research at Stanford University's 2018 Machine Learning Medical Conference. Two-pronged approach To create an artificial intelligence that would determine the optimal drug treatment for patients with glioblastoma, MIT researchers turned to a training technique called Reinforcement Learning (RL). First, they created a test group of 50 patients with simulated glioblastoma who were based on large amounts of data that had been previously treated. They then asked their artificial intelligence to recommend several drugs that are commonly used to treat xylozolomide (TMZ) glioblastoma and regularly inject perzine and loce for each patient (weeks or months). Mstin and vincristine (PVC). After the artificial intelligence has prescribed a dose, it examines a computer model that predicts how a dose will shrink the tumor. When artificial intelligence specifies the dose of tumor shrinkage, it is rewarded. However, if artificial intelligence only specifies the maximum dose, it will be punished. According to the researchers, this requires a balance between the goals and the consequences of the behavior – in this case, the reduction in tumors and the quality of life of the patients – unique in the RL field. Other RL models only work toward one goal; for example, DeepMind's AlphaZero only focuses on winning a game. "If what we want to do is reduce the average tumor diameter and let it take whatever action it wants, it will manage the drug irresponsibly," lead researcher Pratik Shah told MIT News. “On the contrary, we say that we need to reduce the harmful actions taken to achieve this result.†Personal test Artificial intelligence performed approximately 20,000 tests for each simulated patient to complete their training. Next, the researchers tested artificial intelligence on a group of 50 new simulated patients and found that it can reduce both dose and frequency while reducing tumor size. It also takes into account specific patient information such as tumor size, medical history and biomarkers. We asked the "model": "Do you have to give the same dose to all patients?" It said: "No. I can give this person a quarter dose, half to this person, maybe we will jump for this person. After a dose," Shah said. “This is the most exciting part of this work, and we can experiment with one person by using an unorthodox machine learning architecture to produce precise medications.†Artificial intelligence is still subject to further testing and review by the US Food and Drug Administration (FDA) in order for doctors to implement it. But if it passes these tests, it may eventually help people with glioblastoma to attack their brain tumors without causing more pain. Gelatin Size 4 Empty Capsule,Color Gelatin Empty Capsules,Medical Gelatin Empty Capsules,Size 4 Empty Pill Capsule Shell Ningbo Jiangnan Capsule Co., Ltd. , https://www.jncapsule.com