May 22nd, 2024 — Realyze Intelligence technology has powered a groundbreaking research study that will inform more personalized care pathways and improve the quality of life for breast cancer patients. The research was published today in the Journal of Clinical Oncology Clinical Cancer Informatics, an American Society of Clinical Oncology (ASCO) journal. The study demonstrates the utility of Realyze’s validated artificial intelligence (AI) software in unlocking real-world data and empowering healthcare researchers to improve patient health outcomes. The same Realyze technology is already helping patients more efficiently and accurately match to clinical trials at NCI designated cancer centers.
Realyze CEO Aaron Brauser shares, “At Realyze Intelligence, we are committed to moving beyond the hype of AI in healthcare to deliver real, tangible benefits to patient care. Our technology is built to enhance the capabilities of clinicians and researchers. Ultimately, our goal is to improve the lives of patients. With this publication, we’re demonstrating how our platform can be utilized to identify a highly specific population of patients, to de-escalate unnecessary surgical intervention, and to improve quality of life outcomes.”
“This research informs the overall goal of moving away from a one-size-fits-all approach for treating breast cancer,” said lead author Neil Carleton, graduate student in Pitt’s Medical Scientist Training Program. “Instead, our focus is tailoring care so that treatment is at the right level for each patient, which includes doing less surgery when it’s unlikely to have a benefit.”
The objective of the study was to examine whether the Society of Surgical Oncology’s ‘Choosing Wisely‘ recommendations – which advise against sentinel lymph node biopsy (SLNB) for certain breast cancer patients aged 70 and over – could be extended into younger patients. Because there is an increased risk of lymphedema post-procedure, the goal is to minimize SLNB procedures in patient populations where the risk outweighs the benefit. A group of like-minded researchers from UPMC Hillman Cancer Center, Magee Women’s Institute and the University of Pittsburgh collaborated to use Realyze Intelligence’s technology to scour the electronic health records of more than 6000 patients to answer this complex research question.
Realyze’s patented, clinician-trained AI software identified 925 previously diagnosed, early-stage, estrogen receptor positive (ER+), aged 55+ breast cancer patients who had a sentinel lymph node biopsy as part of their care from 2015-2017. The researchers then compared rates of lymph node-positivity, which indicates that the cancer has spread, with rates of lymphedema identified in an average 5.5-year follow-up period. Using natural language understanding (NLU), the Realyze software filtered hundreds of thousands of clinical notes, radiology reports, pathology reports, etc. Notably, the Realyze technology can ‘read’ scanned PDF documents where the majority of this critical information is buried.
“There’s a lot of valuable data in the electronic health care record that is hidden within unstructured physician notes,” said Carleton. “The power of Realyze is that it goes into these notes and pulls out information by looking at the language itself and, importantly, the context of that language. So, it allowed us to look not only for records of lymphedema but also where that lymphedema was located. That’s important because arm and breast lymphedema are likely to be related to lymph node biopsy, whereas leg lymphedema is likely due to other conditions.”
In comparison to manually verified records, Realyze’s NLU pipeline was 95% accurate in identifying cases of breast or arm lymphedema, indicating the tool was effective at extracting this valuable information. which isn’t routinely documented in a predictable way in patient records. Realyze empowered breast cancer researchers to ask this complex question and get answers quickly from real-world data found in patient records.
Brauser adds that the use case of Realyze technology isn’t limited to this specific study, “The same, powerful artificial intelligence technology used for this project can be used to accelerate clinical trial screening and revolutionize the speed of answering many more research questions.” Indeed – Realyze has demonstrated the utility of its platform for cancer data element extraction with Memorial Sloan Kettering in a poster presented earlier this year at AACR. Brauser underscores, “We’re always looking for new partners who want to deploy clinician-trained AI and take patient care to the next level. If you’re ready to move beyond hype and put AI into practice for the betterment of your patients, we’re here to help.”
Other authors on the study were Gilan Saadawi M.D., Ph.D., of Realyze Intelligence; and Priscilla McAuliffe, M.D. Ph.D., Atilla Soran, M.D., M.P.H., Steffi Oesterreich Ph.D., and Adrian Lee Ph.D., all of Pitt and UPMC.
This research was supported by the National Cancer Institute (5F30CA264963-02), the Breast Cancer Research Foundation and Susan G. Komen.
The UPMC press release expands more on this project.
The original article titled, ‘Use of Natural Language Understanding to Facilitate Surgical De-Escalation of Axillary Staging in Patients With Breast Cancer‘ can be downloaded from Journal of Clinical Oncology – Clinical Cancer Informatics.
Reach out to the Realyze Intelligence team here.