Daniel Dabil and Kamyar Saidabadi have been friends since childhood and share a passion for science and technology. The rapid development of artificial intelligence (AI) in recent years has made them interested in exploring how machine learning can improve healthcare systems.
“Today, AI is becoming more and more accessible and we are seeing it being implemented in almost every aspect of our lives,” said Saeedabadi, who studies industrial engineering. Masu. “So we thought, why not introduce something that can help make a big impact on our current healthcare system?”
Two TMU students are currently spearheading a project to use machine learning to detect and classify brain tumors. Their tools aim to achieve his three main goals: reduce wait times for radiologists, reduce healthcare costs, and improve healthcare outcomes in hospitals and medical settings.
“Canadian Association of Radiologists” (PDF file) Public data (External link) The results show that by 2022, it will take the average person 89 days to get an MRI scan, costing the industry around $5 billion in waiting time.'' Psychology student said Mr. Dabil. “Our idea was to increase efficiency and reduce waiting times, which in turn increases cost savings. By using this AI-powered tool, MRI scans will be faster and more accurate. There are many beneficial results.”
April Kademi, TMU's Medical Imaging AI Canada Research Chair and professor of biomedical engineering, says AI tools for medical imaging hold promise for improving diagnostic accuracy and turnaround time for patients suffering from neurological diseases.
“Using AI in the diagnosis and management of brain tumors has the potential to enable more personalized treatment plans and ultimately improve quality of care. Additionally, AI tools can reduce diagnosis time and improve patient It reduces stress,” Kademi said.
Easy-to-use technology that accelerates adoption
Using this tool, medical professionals can upload images or MRI scans and perform image analysis to classify tumors into four categories: meningioma, glioma, pituitary, and non-tumor (i.e., clean MRI scan). ) can be accurately classified as one of the following.
In designing the tool, Dabir and Saeedabadi used a deep learning algorithm called a convolutional neural network, which analyzes an input image and processes it using mathematical operations.
“Convolutional neural networks are widely available and people are using them for a variety of purposes. It has many implications in medicine, and people are now starting to use it,” Dabil said.
“A report from the Canadian Association of Radiology recommends that AI or deep learning machines should be built to assist radiologists, whether it's scanning lungs, broken bones, or internal bleeding. Most of the time it's just classifying images, and that's the whole principle of radiology, so the possibilities are endless. But we put a unique twist on it by using it for MRI scans. .”
Kademi added that AI will revolutionize the way doctors practice by providing them with new quantitative indicators of disease for different treatments. “We are at this stage of rapid technology evolution, but it is important that physicians are involved in the design as the ultimate end users,” she said.
I'm working on it
To make the tool as accessible and easy to use as possible, students adapted their model to Google's Teachable Machine. (External link) So anyone can input images from their own scans and get the results. However, there are some limitations that make it easier.
“We need good datasets to provide completely accurate results, but there are variations in tumor location, shape, and size that can skew results and diagnoses,” Saedabadi said. Ta. “However, this technology is still a great tool that can assist in the diagnostic process and flag certain his MRIs. As with any technology, there is always room for improvement when it comes to design and testing.”
The students are currently prioritizing improved accuracy and working to expand the dataset to enhance the tool. They are also drafting proposals for potential research partnerships with medical institutions.
“We are optimistic about the potential partnership and the impact our project could have on the medical diagnostic field. We are exploring potential collaborations and opportunities to deploy our AI-based systems in the United States,” Dabil said.
