Led by Dr. Wei Shao, an assistant professor of quantitative health in the Department of Medicine and Assistant Director of AI Imaging for the University of Florida’s Intelligent Critical Care Center (IC3), researchers at the UF MIRTH AI lab are developing a model to be used by physicians for early detection and localization of clinically significant prostate cancer. They recently disseminated their preliminary findings through Cornell University’s arXiv open-access preprint server.
In their article, posted May 31, 2023, UF MIRTH AI Lab researchers discuss a relatively new type of advanced ultrasound called micro-ultrasound that provides higher resolution than regular ultrasound, making it useful for outlining and diagnosing prostate cancer. Early detection of prostate cancer improves 5-year survival rates. Moreover, precise outlining of the prostate is important for various tasks; but, it is also challenging due to unclear boundaries between organs. The traditional approach to detect and localize prostate cancer, specifically image-guided biopsy including magnetic resonance imaging (MRI)-ultrasound fusion, while helpful in cancer detection, can be costly. Similar to MRI, micro-ultrasound scans can be produced at a lower cost than other imaging technologies, and is used by urologists for prostate cancer detection and monitoring, prostate biopsy, and active surveillance of prostate cancer.
To improve the usefulness and efficiency of micro-ultrasound, the MIRTH AI Lab researchers present MicroSegNet, a computer model utilizing micro-ultrasound and designed to accurately outline the prostate through a special approach that focuses on difficult areas. In this article, first-author graduate assistant, Hongxu Jiang, and coauthors state that their new model outperformed other methods and even human experts in testing, thereby achieving high accuracy.
Also shared by the group via arXiv on March 31 is a paper proposing a method to align in vivo micro-US images with ex vivo histopathology images. The researchers’ method utilizing mirco-ultrasound imaging was tested on 18 patients and achieved accurate alignment. This study, first-authored by post-doctoral associate Dr. Muhammad Imran, showcases the potential for aligning micro-ultrasound and histopathology images.
“By implementing our proposed AI system,” says Shao, “our team of researchers has the potential to significantly improve the health and quality of life of men impacted by prostate cancer. Our methods can help in reducing the cost and burden of both invasive biopsy as well as more common prostate cancer detection imaging. Our ultimate goal at the UF MIRTH AI Lab is to lessen the healthcare burden on individuals and our nation.”
MIRTH AI is an acronym for Medical Imaging Research for Translational Healthcare with Artificial Intelligence coined by Shao to reflect their hope in creating personalized, effective, and less expensive healthcare solutions. The UF MIRTH AI Lab researchers publicly share their model, code, and data for further research; links to these resources can be found in each respective article.