Home Bone marrow transplantion Leveraging machine learning to revolutionize bone marrow transplantation

Leveraging machine learning to revolutionize bone marrow transplantation

by Alex Generous
0 comments

Big innovations often emerge from the intersection of multiple technologies and disciplines. One notable example may be in your hand or pocket right now. It is a combination of communication devices, computers, cameras, entertainment systems, and other electronic equipment that make up a smartphone. Innovations include the combination of chemical capture of light and theatrically created movies, or how peanut butter and chocolate come together to form the famous orange-wrapped candy. Even simpler mixtures can result.

In medicine, the combination of computation and clinical expertise opens the door to better care. Machine learning, the use of computer systems designed to adapt and learn from a set of parameters without following explicit instructions, is a pioneering technology. Scientists at St. Jude Children’s Research Hospital combine machine learning and bone marrow transplant expertise to find machine learning more accurately predicts patient survival at 100 days, 1 year, and 2 years after surgery than ever before I was able to create an algorithm.

“One of the things we always struggle with is how to predict who will have a bad outcome after a transplant,” said Akshay Sharma, MBBS, St. Jude Bone Marrow Transplant and Cell Therapy. said.In recent research Blood Advances. “If we can predict who will have such a bad outcome, then we can do something about it. This study was a proof of concept that machine learning can be used to improve early prediction of poor outcomes after bone marrow transplantation.” .”

“We know very well that transplantation is a major medical procedure and typically requires a lengthy hospital stay of several weeks,” added co-author Dr. Lee Tan of St. Jude Biostatistics. Ta. “Frequent monitoring and periodic testing of transplant patients is also common in most institutions. However, prior to our study, long-term continuous monitoring, electronic medical record integration, , there has been limited exploration of how assimilation of other procedural information can be leveraged to increase accuracy in predicting adverse outcomes in these patients.”

“Our algorithm achieved higher accuracy in predicting short- and long-term survival than existing risk prediction models,” said lead author Ywang Zhou, Ph.D., of St. Jude’s Office of Biostatistics. “That superior performance was driven by our machine learning algorithms and the incorporation of longitudinal measurements from transplant patients.”

Combining bone marrow transplantation and machine learning

Bone marrow transplantation, a treatment for leukemia, can be very dangerous for patients. Patients are treated with chemotherapy to have their own bone marrow harvested and then receive new bone marrow from a donor. This process aims to replace the donor’s cancerous blood cells and remaining normal blood-forming cells with healthy blood-forming cells from the donor. However, there are many potential complications, including graft-versus-host disease, graft failure, and opportunistic infections, all of which can be life-threatening.

Given the significant risks associated with bone marrow transplantation, doctors have created a risk prediction model to identify patients who are at highest risk of developing these complications after transplantation and therefore require additional medical intervention. . Many of these models are less than ideal, with only 50% accuracy in estimating mortality risk. That’s because many of these risk prediction models have fundamental design flaws. That is, it only uses data from a single point in time.

“For most models, that snapshot is taken before porting,” Sharma explained. “No one had accounted for the changes that occurred in the patient after the transplant, which is the most intensive part of this whole treatment. So we have been making predictions without considering the procedure.” St. Jude The difference is that the algorithm looks at multiple time points from 1 month before transplant to 1 month after transplant, creating a longitudinal dataset, allowing the algorithm to find patterns.

“This is something very new for children with cancer,” Zhou says. “This is the first predictive model that incorporates longitudinal data from pediatric patients undergoing allogeneic hematopoietic cell transplantation.”

“Even though the data in the model was from up to 30 days post-transplant, the accuracy of predictions at 1 and 2 years post-transplant was significantly improved compared to considering only baseline values. We were really surprised by what we did,” Sharma said. “This is a testament to our ability to collect and use longitudinal data in a practical way.”

wide variety of variables

In addition to including longitudinal data, the St. Jude model improved its predictions by including far more variables in the analysis than previous models.

“Previous models included only 10 to 20 variables,” says Zhou. “In our method, we started with more than 100 longitudinal measurements from clinical trials, in addition to baseline information. This significantly increased the number of variables considered. .”

To protect transplant patients, doctors regularly take blood samples to monitor their condition. That data is stored in standardized electronic health records and presents an opportunity for machine learning. Unlike earlier statistical models, machine learning models can process the large amounts of data held in these records. Instead of incorporating 10-20 variables, he can use over 100 per day per patient.

By combining an influx of information previously too vast to analyze with the trend-trending nature of longitudinal data, St. Jude’s algorithms can determine whether a patient is at risk of developing complications and dying. I was able to find a pattern that shows this.

Scientists tested the algorithm at both St. Jude and partner institutions. At St. Jude University, his 70% of all information in the analyzed cohort was included in the training of the algorithm. The statisticians then tested their predictions for the remaining 30%. But the real test was whether it would work for patients treated at other hospitals. When applied to a large cohort at Memorial Sloan Kettering Cancer Center, the model continued to successfully predict potential complications.

“This validation shows that our model is very robust,” Zhou explained. “It is reliable and the predictions can be reproduced based on data collected from other institutions.”

A powerful combination: bone marrow transplantation and biostatistics

By combining two distinct fields: machine learning and bone marrow transplantation, St. Jude scientists have created the first version of a powerful predictive tool that could one day improve patient outcomes. A special combination of disciplines was the secret to innovation.

“St. Jude is a unique place, not only because we’re good at transplants and do a lot of transplants, but also because we have a great biostatistics core,” Sharma said. “We can use our brains to develop solutions that no one else has created before.”

“We showed how statisticians and doctors can work together to create something new,” Zhou said. “We encourage researchers in pediatric oncology to consider using machine learning algorithms in their future studies. Databases are becoming increasingly large, and vast amounts of data are collected in electronic medical records. In the future, machine learning techniques will help further clarify these discoveries in many areas of medicine.”

“This is the beginning of our journey, and our plans include digging deeper into this data using cutting-edge computational techniques,” Tan emphasized. “We aim to uncover further insights, piece together a clearer picture, and develop enhanced solutions that benefit both clinicians and patients. This is just the beginning. It is important to emphasize that.”

news source

You may also like

About Us

Welcome to Daily Transplant News, your trusted source for the latest updates, stories, and information on transplantation and organ donations. We are passionate about sharing the inspiring journeys, groundbreaking research, and invaluable resources surrounding the world of transplantation.

About Us

Welcome to Daily Transplant News, your trusted source for the latest updates, stories, and information on transplantation and organ donations. We are passionate about sharing the inspiring journeys, groundbreaking research, and invaluable resources surrounding the world of transplantation.

Copyright ©️ 2024 Daily Transplant News | All rights reserved.