The primary documented case of pancreatic cancer dates back to the 18th century. Since then, researchers have undertaken a protracted and difficult odyssey to grasp the elusive and deadly disease. Thus far, there is no such thing as a higher cancer treatment than early intervention. Unfortunately, the pancreas, nestled deep throughout the abdomen, is especially elusive for early detection.
MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) scientists, alongside Limor Appelbaum, a staff scientist within the Department of Radiation Oncology at Beth Israel Deaconess Medical Center (BIDMC), were desperate to higher discover potential high-risk patients. They got down to develop two machine-learning models for early detection of pancreatic ductal adenocarcinoma (PDAC), probably the most common type of the cancer. To access a broad and diverse database, the team synced up with a federated network company, using electronic health record data from various institutions across the USA. This vast pool of knowledge helped make sure the models’ reliability and generalizability, making them applicable across a big selection of populations, geographical locations, and demographic groups.
The 2 models — the “PRISM” neural network, and the logistic regression model (a statistical technique for probability), outperformed current methods. The team’s comparison showed that while standard screening criteria discover about 10 percent of PDAC cases using a five-times higher relative risk threshold, Prism can detect 35 percent of PDAC cases at this same threshold.
Using AI to detect cancer risk will not be a recent phenomena — algorithms analyze mammograms, CT scans for lung cancer, and assist within the evaluation of Pap smear tests and HPV testing, to call a number of applications. “The PRISM models stand out for his or her development and validation on an in depth database of over 5 million patients, surpassing the dimensions of most prior research in the sphere,” says Kai Jia, an MIT PhD student in electrical engineering and computer science (EECS), MIT CSAIL affiliate, and first writer on an open-access paper in outlining the brand new work. “The model uses routine clinical and lab data to make its predictions, and the range of the U.S. population is a major advancement over other PDAC models, which are often confined to specific geographic regions, like a number of health-care centers within the U.S. Moreover, using a novel regularization technique within the training process enhanced the models’ generalizability and interpretability.”
“This report outlines a strong approach to make use of big data and artificial intelligence algorithms to refine our approach to identifying risk profiles for cancer,” says David Avigan, a Harvard Medical School professor and the cancer center director and chief of hematology and hematologic malignancies at BIDMC, who was not involved within the study. “This approach may result in novel strategies to discover patients with high risk for malignancy which will profit from focused screening with the potential for early intervention.”
Prismatic perspectives
The journey toward the event of PRISM began over six years ago, fueled by firsthand experiences with the restrictions of current diagnostic practices. “Roughly 80-85 percent of pancreatic cancer patients are diagnosed at advanced stages, where cure isn’t any longer an option,” says senior writer Appelbaum, who can also be a Harvard Medical School instructor in addition to radiation oncologist. “This clinical frustration sparked the thought to delve into the wealth of knowledge available in electronic health records (EHRs).”
The CSAIL group’s close collaboration with Appelbaum made it possible to grasp the combined medical and machine learning points of the issue higher, eventually resulting in a way more accurate and transparent model. “The hypothesis was that these records contained hidden clues — subtle signs and symptoms that might act as early warning signals of pancreatic cancer,” she adds. “This guided our use of federated EHR networks in developing these models, for a scalable approach for deploying risk prediction tools in health care.”
Each PrismNN and PrismLR models analyze EHR data, including patient demographics, diagnoses, medications, and lab results, to evaluate PDAC risk. PrismNN uses artificial neural networks to detect intricate patterns in data features like age, medical history, and lab results, yielding a risk rating for PDAC likelihood. PrismLR uses logistic regression for a less complicated evaluation, generating a probability rating of PDAC based on these features. Together, the models offer an intensive evaluation of various approaches in predicting PDAC risk from the identical EHR data.
One paramount point for gaining the trust of physicians, the team notes, is best understanding how the models work, known in the sphere as interpretability. The scientists identified that while logistic regression models are inherently easier to interpret, recent advancements have made deep neural networks somewhat more transparent. This helped the team to refine the 1000’s of probably predictive features derived from EHR of a single patient to roughly 85 critical indicators. These indicators, which include patient age, diabetes diagnosis, and an increased frequency of visits to physicians, are routinely discovered by the model but match physicians’ understanding of risk aspects related to pancreatic cancer.
The trail forward
Despite the promise of the PRISM models, as with all research, some parts are still a piece in progress. U.S. data alone are the present weight loss plan for the models, necessitating testing and adaptation for global use. The trail forward, the team notes, includes expanding the model’s applicability to international datasets and integrating additional biomarkers for more refined risk assessment.
“A subsequent aim for us is to facilitate the models’ implementation in routine health care settings. The vision is to have these models function seamlessly within the background of health care systems, routinely analyzing patient data and alerting physicians to high-risk cases without adding to their workload,” says Jia. “A machine-learning model integrated with the EHR system could empower physicians with early alerts for high-risk patients, potentially enabling interventions well before symptoms manifest. We’re desperate to deploy our techniques in the actual world to assist all individuals enjoy longer, healthier lives.”
Jia wrote the paper alongside Applebaum and MIT EECS Professor and CSAIL Principal Investigator Martin Rinard, who’re each senior authors of the paper. Researchers on the paper were supported during their time at MIT CSAIL, partially, by the Defense Advanced Research Projects Agency, Boeing, the National Science Foundation, and Aarno Labs. TriNetX provided resources for the project, and the Prevent Cancer Foundation also supported the team.