Researchers which focuses on computer engineering and , built two machine learning algorithms that can detect pancreatic cancer at a higher threshold than current diagnostic standards. The two models are combined to form the PRISM neural network. It is designed to specifically detect pancreatic adenocarcinoma (PDAC), the most common form of pancreatic cancer.
Current standard PDAC screening criteria capture approximately 10 percent of cases in patients seen by specialists. By comparison, MIT’s PRISM was able to identify PDAC cases 35 percent of the time.
When using not an entirely new feat, MIT’s PRISM stands out because of how it was developed. The neural network was programmed based on access to a variety of real electronic health records from healthcare facilities in the United States. It was fed data from more than 5 million patients’ electronic health records, which the team’s researchers said “exceeded the scale” of the information given to the AI model in this particular area of research. “The model uses routine clinical and laboratory data to make its predictions, and is a significant improvement over other PDAC models that typically limit the diversity of the US population to specific geographic regions, such as a few healthcare centers in the US,” said Kai Jia, PhD, MIT CSAIL, lead author of the paper. .
MIT’s PRISM project began more than six years ago. The motivation for developing an algorithm that can detect PDAC early has a lot to do with the fact that most patients are diagnosed at a later stage of cancer development – especially since about eighty percent are diagnosed too late.
AI works by analyzing patient demographics, previous diagnoses, current and previous medications in care plans, and lab results. Collectively, the model works to predict the likelihood of cancer by analyzing electronic health record data in tandem with factors such as a patient’s age and certain risk factors evident in lifestyle. Still, PRISM can still help diagnose as many patients as AI can reach the masses. Currently, the technology is closed and select patients in the United States. The logistical challenge of scaling AI will involve feeding the algorithm with more diverse datasets and perhaps global health profiles to increase accessibility.
However, this is not MIT’s first foray into creating an artificial intelligence model that can predict cancer risk. In particular, he developed a way to train models to make predictions among women using mammogram records. In this line of research, the MIT experts confirmed that the more diverse the datasets, the better the AI at diagnosing cancer. and population. The continued development of AI models capable of predicting the likelihood of cancer will not only improve patient outcomes if malignancy is detected earlier, but also reduce patient workload. The market for artificial intelligence in diagnostics is so ripe for change that it’s attracting the interest of big tech ads. a year ago, he attempted to create an artificial intelligence program that could detect breast cancer.
