Content
summary Summary

MIT (Massachusetts Institute of Technology) announced two new AI models called "PRISM". They are designed to detect pancreatic cancer earlier than conventional methods.

Conventional diagnostic methods for pancreatic cancer detect only about ten percent of cases at an early stage. The PRISM system significantly improves this rate.

While the standard screening criteria detect approximately ten percent of pancreatic ductal adenocarcinoma (PDAC) cases with a fivefold increase in the relative risk threshold, the PRISM model can detect 35 percent of PDAC cases with the same threshold. Increasing the early detection rate is important because early diagnosis can be life-saving.

Two statistical models working together

The researchers developed two models: the PRISM neural network and a logistic regression model. Both models analyze electronic health records, including patient demographics, diagnoses, medications, and lab results, to assess PDAC risk.

Ad
Ad

The PRISM model uses neural networks to recognize complex patterns in these data points and calculate a risk score for the likelihood of PDAC. The logistic regression model uses a simpler analysis to generate a probability score for PDAC from these features.

In total, PRISM was trained on data from more than five million patient records. This large dataset enabled the algorithms to identify patterns that human physicians might miss.

MIT has experience developing AI models for cancer diagnosis, such as predicting the risk of breast cancer. These projects show that greater diversity in data sets can lead to more accurate diagnoses.

Research on PRISM began more than six years ago to improve the detection of pancreatic cancer, which is diagnosed at a late stage in 80% of patients.

While the PRISM models are promising, there are still some aspects that need further development. Currently, the models are based only on U.S. data, which needs to be tested and adapted for global use.

Recommendation

In the future, the team plans to expand the applicability of the models to international datasets and integrate additional biomarkers for more refined risk assessment. The researchers also want to facilitate the implementation of the models in routine health care.

The vision is for the models to work seamlessly in the background of the healthcare system, automatically analyzing patient data and alerting physicians to high-risk cases without increasing their workload.

"Despite the promise of the PRISM models, as with all research, some parts are still a work in progress," the researchers write.

Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Bank transfer
Summary
  • Researchers at the Massachusetts Institute of Technology (MIT) have developed PRISM, two AI models that improve early detection of pancreatic cancer. PRISM detects 35 percent of cancer cases at an early stage, compared to only 10 percent with conventional methods.
  • The system has been trained on data from more than five million patient records, enabling more accurate pattern recognition.
  • PRISM could improve the medical diagnosis of pancreatic cancer. The earlier the cancer is detected, the better the chances of recovery. However, the researchers also point out that the systems still need to be improved and that their use is limited to the U.S. because of the training data used.
Sources
Michelangelo works as a news reporter for THE DECODER. His background is in electrical engineering.
Co-author: Matthias Bastian
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.