Photo Credit: Courtesy of Adelaide University
Scientific Frontline: Extended "At a Glance" Summary: EndoFusion
The Core Concept: EndoFusion is an artificial intelligence diagnostic tool that rapidly and noninvasively detects signs of advanced endometriosis from a single pelvic scan.
Key Distinction/Mechanism: Unlike the current standard of visual identification through invasive, costly, and slow keyhole surgery, EndoFusion utilizes machine learning to integrate and analyze data from both magnetic resonance imaging (MRI) and transvaginal ultrasound scans. This circumvents the limitations of relying on single imaging methods and the variabilities of operator experience.
Major Frameworks/Components:
- Diagnostic Datasets: The system was trained using information gathered from four datasets, comprising more than 9,000 female pelvic MRI scans and over 800 transvaginal ultrasound sliding scans.
- Algorithmic Efficacy: The framework achieves an 83% accuracy rate in distinguishing between positive and negative cases of endometriosis, outperforming all competing models, and produces results in just 18 milliseconds.
- Interdisciplinary Collaboration: The ongoing development includes partnerships with Flinders University, Benson Radiology, Omni Ultrasound and Gynecological Care, the University of Surrey, the McMaster University Medical Center, and the Mohamed bin Zayed University of Artificial Intelligence.
Branch of Science: Artificial Intelligence (in Medicine), Gynecology, Radiology, and Pathology.
Future Application: The dataset will be expanded to include additional endometriosis markers to improve classification accuracy. The technology also presents potential applications for researching other conditions that require multimodal imaging, including prostate cancer, breast cancer, and fetal abnormalities.
Why It Matters: Endometriosis is a chronic condition affecting more than 190 million women globally, characterized by the painful growth of uterine tissue outside the uterus. By providing an accurate, noninvasive early diagnostic method, EndoFusion has the potential to eliminate the typical seven-year wait for a surgical diagnosis, thereby reducing patient suffering, alleviating anxiety, and lowering healthcare costs.
Less than a second. That is how long it takes for an emerging AI tool to detect signs of endometriosis through one simple scan. It is a much faster process than the current seven-year wait for surgery, and it is less invasive.
The AI tool, called EndoFusion, is a recent development from IMAGENDO®, an ongoing collaborative study led by Adelaide University researchers. In this latest study, they found the framework was able to accurately identify two major indicators of advanced endometriosis in pelvic scans, producing the results in just 18 milliseconds.
Researchers say it is a significant development, as MRI and ultrasound imaging are often better at detecting one sign of the condition over the other.
“Current scanning methods each have their own strengths when it comes to detecting two common markers that indicate the likelihood of endometriosis, and patients will often only have access to one of them,” said study author Associate Professor Jodie Avery, research co-lead of chronic reproductive conditions in the Endometriosis Research Group at Adelaide University’s Robinson Research Institute.
“This means that some patients could be disadvantaged if they are scanned by the less optimal option for their particular signs. Some of the imaging tools also rely on operator experience and can be costly.
“Our AI tool can help address these shortcomings by combining data from both imaging tools, giving the framework the knowledge it needs to detect both signs of endometriosis through a single scan more effectively and efficiently.”
The AI framework is still in the early stages of development. It works by using information gathered from four datasets containing more than 9,000 female pelvic MRI scans and more than 800 transvaginal ultrasound sliding scans.
“We looked at how well EndoFusion was able to distinguish between positive and negative cases of endometriosis and found it was able to provide a correct diagnosis 83% of the time, which is more accurate than all competing models,” said lead author Dr. Yuan Zhang from Adelaide University’s Robinson Research Institute and the Australian Institute for Machine Learning.
“This is a positive step forward and moves us closer to a future where an AI tool can help clinicians provide a faster, more accurate diagnosis without the need for surgery.”
There are more than 190 million women worldwide who have endometriosis. The chronic condition occurs when uterine tissue grows outside of the uterus, causing symptoms including abdominal pain, heavy periods, bloating, anxiety, fatigue, and infertility.
It is notoriously difficult to diagnose, and diagnosis often relies on identifying lesions visually through surgery, a process that is slow, costly, and risky.
“The development of accurate, noninvasive early diagnostic methods is critical to shorten the diagnostic timeline and reduce associated costs,” said Associate Professor Avery.
Researchers collaborated with Flinders University, Benson Radiology, Omni Ultrasound and Gynecological Care, the University of Surrey, the McMaster University Medical Center, and Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), with the results recently published in Artificial Intelligence in Medicine.
The next step will involve expanding the dataset to include additional endometriosis markers to improve the classification accuracy of the tool.
“We envision that clinicians will be able to use these tools to help make a determination about the presence of endometriosis from a single scan,” said Dr. Zhang.
“It could also potentially provide insights for research into other diseases that require the use of multimodal imaging, such as gynecological disorders, prostate and breast cancers, and fetal abnormalities.”
Published in journal: Artificial Intelligence in Medicine
Title: Unpaired multi-modal multi-label learning for detecting endometriosis signs
Authors: Yuan Zhang, Hu Wang, Yutong Xie, Minh-Son To, Steven Knox, Mathew Leonardi, George Condous, Jodie C. Avery, M. Louise Hull, and Gustavo Carneiro
Source/Credit: Adelaide University
Edited by: Scientific Frontline
Reference Number: ai091626_01
