TL;DR:
- Montana State University researchers are utilizing artificial intelligence to advance cancer diagnosis and treatment.
- By detecting visual patterns in biopsy scans, the team aims to provide more consistent and accurate diagnoses.
- The project combines expertise in biomedical engineering, computational geometry, computer vision, machine learning, and pathology.
- Integrating topological data analysis with machine learning has propelled the project forward.
- AI technology holds significant potential for assisting doctors in recommending effective treatment protocols.
- While not aiming to replace doctors, the project aims to create tools that enhance the diagnostic process.
- The application of AI in cancer diagnosis signifies a critical turning point for personalized and improved patient care.
Main AI News:
In recent times, the astounding linguistic capabilities of ChatGPT and other chatbots have been grabbing headlines worldwide. Yet, amidst the buzz, a group of dedicated researchers at Montana State University has been quietly pushing the boundaries of artificial intelligence in a different, more profound way—one that holds the potential to save lives. Their focus: using AI to diagnose and treat prostate cancer.
By employing novel computing methods to identify visual patterns in biopsy scans, these scientists, in collaboration with medical experts, are forging ahead with groundbreaking technology that could revolutionize the way doctors recommend effective treatments.
The significance of their work lies in the realization that three different pathologists might examine the same biopsy slide and arrive at varying conclusions about the presence and severity of cancer. “Our goal is to develop a computer-assisted process that can enhance the consistency and, hopefully, the accuracy of diagnoses,” explained John Sheppard, a distinguished professor in the Gianforte School of Computing at MSU’s Norm Asbjornson College of Engineering.
Sheppard, who himself battled prostate cancer, joined the project about four years ago. Following a blood test that indicated the disease, he underwent robot-assisted surgery to remove his prostate. Prostate cancer is the most prevalent form of cancer and ranks as the second leading cause of cancer-related deaths among men in the United States, according to the National Cancer Institute.
“When I heard about the project, I immediately knew I wanted to be a part of it,” Sheppard revealed. “Having experienced the personal journey of diagnosis and treatment, I am motivated to assist others in finding the most appropriate course of action.“
The project was initiated in 2015 by Brittany Fasy, an associate professor of computer science at MSU, in collaboration with experts from Tulane University and medical practitioners in Bozeman and New Orleans. Generously funded by the National Science Foundation and the National Institutes of Health, the research endeavors have been propelled forward with $420,000.
Through partnerships with medical professionals, the researchers gained access to real prostate scans—microscope images of tissue samples from cancerous tumors—along with anonymized patient data encompassing diagnoses and outcomes. Sheppard, Fasy, and their interdisciplinary team developed sophisticated computer techniques to identify visual patterns that correlate with the grade of cancer in the tumors. These patterns could equip physicians with valuable insights to determine optimal treatment protocols. The project involved a team of seven graduate students and four undergraduates.
Combining their expertise in biomedical engineering, computational geometry, computer vision, machine learning, and pathology, this collaborative effort stands in a league of its own when it comes to improving the analysis of scanned images. Leading MSU’s involvement in the project, Fasy emphasized the importance of an open dialogue between different disciplines, which drives the development and successful application of cutting-edge techniques. Her particular expertise lies in topological data analysis (TDA), a field that studies and employs algorithms to identify shapes and patterns within datasets. The integration of TDA with machine learning represents a noteworthy achievement that has propelled the project forward.
Sheppard, with his vast experience in machine learning—a subset of AI that utilizes algorithms and statistical models to adapt to patterns in data—was brought onto the team to contribute his expertise. Reflecting on the distinction between traditional programming and machine learning, Sheppard explained, “With traditional programming, you already know how to approach finding the answer to a problem and use computer code to get it. Machine learning, however, begins with the data and processes it to determine what the models should be. That’s the ‘intelligent’ part of the process.“
While today’s chatbots, powered by what are known as large language models, optimize over a billion parameters as they scour vast amounts of existing text to construct lifelike responses, the model used by the MSU team contains only thousands. Nevertheless, the project team has successfully demonstrated that machine learning algorithms can discern nuanced visual patterns within prostate images. Further work is necessary to integrate these patterns with specific treatment recommendations in a manner that can be readily employed by physicians, according to Fasy.
David King, who served as the clinical research manager at Bozeman Health and collaborated with the researchers, stressed the importance of maintaining the personal doctor-patient relationship while embracing computer-assisted diagnosis. He acknowledged the significant potential of such technology, stating, “It’s truly exciting. Prostate cancer can be elusive. If we can teach the machine to recognize the subtleties, patients can benefit from what could be called an AI consultation as part of their treatment.“
Identifying the aggressiveness of tumors poses the most significant challenge when it comes to prostate cancer diagnosis. Determining whether immediate invasive surgery is required or if a monitoring approach is suitable is crucial. “This is where the technology could be a game-changer for clinicians,” King asserted. Additionally, this tool holds promise for improving care in the realm of breast cancer and, ultimately, other medical conditions.
As Sheppard underscored, it is essential to consider the broader context of doctors’ knowledge, refined over decades of experience, as well as their relationships with patients. This guiding principle has remained at the forefront of the project—a testament to the responsible application of AI in any field. “We’re not trying to replace doctors; rather, we aim to create tools that assist them,” Sheppard concluded.
Conclusion:
The use of artificial intelligence in cancer diagnosis and treatment represents a significant milestone in the healthcare market. Montana State University’s pioneering research in this field has the potential to revolutionize the way doctors diagnose and recommend treatments, leading to more consistent and accurate outcomes. The integration of AI technology in medical practices holds promise for enhanced patient care and personalized treatment approaches, marking a transformative shift in the market’s landscape.
