-qi-.png)
Various fields are increasingly adopting artificial intelligence to augment human capabilities, and the health-care sector stands out as one of the most promising areas. AI has the potential to significantly reduce health disparities and improve diagnostic tools, especially in treating underrepresented and underfunded areas, such as ovarian cancer.
This is why professor Qianru (Cheryl) Qi received a Telfer SMRG Research Development Grant for a project titled “Enhancing medical foundation models for underrepresented diseases.” This groundbreaking research aims to improve medical foundation models, specifically CT image analyses used in detecting metastasis associated with ovarian cancer. Accurate detection is vital to guide decisions regarding the best sequence of chemotherapy and surgery, and so it has a direct impact on patient outcomes.
The overarching goal of the project is to enhance diagnostic accuracy for ovarian cancer and develop scalable solutions that can be extended to other underserved health conditions. A central challenge lies in adapting and fine-tuning existing models to effectively detect disease while addressing issues like data bias and scarcity. The research will contribute to creating comprehensive guidelines for using foundation models in clinical settings, which would improve the diagnosis of underrepresented conditions and ultimately advance health care equity.
The methodology used in this project involves evaluating current foundation models, which fail to detect ovarian abnormalities, and training a new model using annotated pelvic CT images. Prompt engineering will then enhance both general and ovarian cancer-specific models to improve prediction accuracy. Finally, the researchers will generate data and test it to further refine the models.
This research has the potential to significantly improve patient care by equipping clinicians with more accurate tools to detect ovarian cancer metastases, leading to earlier and more informed treatment decisions. It will also streamline clinical workflows, reduce diagnostic time, and contribute to the ongoing advancement of ovarian cancer research and management.
