A recent study found that AI can detect pancreatic cancer up to three years before a human doctor. Despite the fact that cancer is a tough disease to detect at an early stage, AI has made it possible to identify it at an early stage.
Only 12% of pancreatic cancer patients survive for five years, making it a deadly disease. According to the Nature Medicine study, AI-based population screening can detect prospective pancreatic cases.
Furthermore, population screening refers to genetic testing or other assessments that look at the prevalence of a specific trait in a community. AI aids in the initial step of early spot detection.
Using AI on real-world clinical records could lead to a scalable way to find cancer early in the community. This would change the focus from treating cancer in its late stages to treating cancer in its early stages, improve patients’ quality of life, and increase the benefit-to-cost ratio of cancer care, according to the paper.
Chris Sander, a professor at Harvard Medical School, stated that determining who is at high risk for a disease and who would benefit from additional testing is one of the most essential decisions physicians must make every day. More testing can mean more invasive, expensive procedures that come with their own risks.
In addition, at the beginning of the testing phase, researchers examined clinical data from 9 million patients in the United States and Denmark using an AI system. In addition, they developed AI learning models to identify detection codes in patient data and associate them with pancreatic cancer.
The study was carried out in collaboration with the Harvard T.H. Chan School of Public Health, the Dana-Farber Cancer Institute, the VA Boston Healthcare System, and the University of Copenhagen.
Also, at the start of the testing phase, researchers used an AI system to look at clinical data from 9 million patients in the US and Denmark. In addition to this, they developed AI learning models to search patient data for detection codes and link those codes to pancreatic cancer.
The research team used a variety of AI models to achieve the desired results. They tested various versions for six, a year, two, and three years. They ultimately got to the stage where their forecasts were substantially more accurate in predicting who would develop pancreatic cancer than current population-wide estimates of disease incidence.
The study is still in its early stages and must go through several processes. However, the program cannot be used to run screening apps. According to numerous hypotheses, cancer patients have a 12.5% chance of survival, making it a deadly condition.
According to Sander, the actual computing costs for using the software are minor once a surveillance program is in place. The training is what consumes the most processing power. Actual clinical testing to seek for early signs of cancer or to detect cancer when it is very little is quite expensive, far more expensive than, say, mammography.
Soren Brunak, a disease system biology professor at the University of Copenhagen, remarked in a press release that many types of cancers, particularly those difficult to identify and treat early, impose an outsized toll on patients, families, and the health-care system as a whole.
According to him, the aggressive disease pancreatic cancer, which is notoriously difficult to diagnose early and treat quickly when the odds of success are at their maximum, is an opportunity to reverse the disease’s trajectory.
However, some indications and symptoms, such as jaundice, upper back and middle back ache, fatigue, itchy skin, and weight loss, can be used to detect pancreatic cancer. Having a tool that can accurately diagnose at an early stage, however, is critical.