Leveraging artificial intelligence to enhance cancer care

Members of the Xing lab

Members of the Xing lab

The Laboratory of Artificial Intelligence in Medicine & Medical Physics is helmed by  Lei Xing, PhD, Stanford Cancer Institute member and director of the medical physics division of radiation oncology department whose research works at the intersection of biomedical physics, oncology, and artificial intelligence (AI) to advance cancer research and healthcare. 

One of the significant overarching projects at Xing’s lab is translating standard AI techniques into clinical practice. Xing explained that current AI technology is often unspecialized for specific fields, rendering the software less functional or applicable to the lab’s needs. He hopes to bring clinical perspectives to the source, advancing AI technology for specialized use in clinical practice. 

“Most AI is developed by computer scientists and engineers who have little or no experience in clinical practice.,” Xing said. “Our faculty in the department of radiation oncology work at the forefront of clinical practice, giving them extensive direct experience in data collection and clinical judgment for truly useful AI research and deployment.”

Xing’s lab pushes the boundaries of AI applications in radiation therapy by working on data-efficient AI solutions to improve healthcare, cutting out the margin of error, and streamlining clinical processes. They leverage cutting-edge AI technologies in medical physics, particularly the latest large language models (AI specializing in comprehending and generating human language) or foundation models (a form of generative AI performing specialized tasks based on human instruction), to enhance imaging, computational procedures, and decision-making processes. 

Xing and his colleagues have created AI that assists in clinical tasks like treatment plan evaluation, tumor segmentation, and even outcome prediction. In imaging, for example, their research spans topics from data-driven image reconstruction and image analysis to deep radiomics and predictive models. With the help of AI, these time-consuming tasks can now be completed in a fraction of the time Xing’s lab pioneered the use of AI in radiation oncology.

“We are still at the beginning of AI healthcare and face many bottlenecks. For instance, data acquisition is slow, expensive, and prone to errors., With our newly developed AI techniques, you can use much less data and get better results, leading to higher quality inference.”

Xing also explained that data provided by physicians can vary significantly due to differences in clinical and educational backgrounds. Standardizing data across the field enables a more collaborative and comprehensive approach to data collection and processing; a task that can be tackled by Xing’s machine-learning techniques.

The research at Xing’s lab extends far beyond the realm of traditional medical physics. Recognizing the importance of high-dimensional data, datasets with a large number of features or attributes, in modern AI and biomedical science, the lab is developing new AI-driven analytics methods and tools for high-dimensional data visualization and interpretable AI to aid in the digital reconstruction of tumors and other relevant structures. They apply analytics tools for pattern discovery in many practical problems, such as clinical trial and molecular data, to advance biomedical data science. 

“While genomic data samples are becoming increasingly indispensable in precision medicine, the data pool size is massive, making it time-consuming or even impossible to find the desired information using existing technologies. Using AI, a data pool can be quickly scanned with a click of a button. The implications could be revolutionary for predicting patient treatment outcomes. When the data is so large and noisy, achieving highly reliable model accuracy becomes problematic. Our new AI model is showing phenomenal improvements in accuracy and efficiency.”

Looking forward, Xing is optimistic about the future of utilizing machine learning techniques to bolster patient care and enhance the research capabilities of laboratories across the Stanford Cancer Institute network.


By Kai Zheng
June 2024