Cancer Data Science Pulse
NCI Funding Helps AI Researchers Gain Global Insight on Cancer
NCI’s Center for Global Health’s Affordable Cancer Technologies (ACTs) Program supports research on tools and technologies, from their initial prototypes to clinical use. Through ACTs, NCI seeks to increase access to quality cancer prevention and care in settings with limited resources found across the globe. This blog features two ACTs grantees working on artificial intelligence (AI) solutions to aid in cancer diagnosis and treatment:
- Ajay Aggarwal, Ph.D., London School of Hygiene & Tropical Medicine, is researching ARCHERY: Artificial Intelligence-based Radiotherapy treatment planning for Cervical and Head and Neck cancer.
- Anant Madabhushi, Ph.D., Winship Cancer Center, Emory University, is examining an AI-enabled Digital Pathology Platform for Multi-Cancer Diagnosis, Prognosis and Prediction of Therapeutic Benefit.
Drs. Aggarwal and Madabhushi give their perspectives on what it’s like developing, managing, and working across multinational borders to find new ways of using technology to address the diagnosis, treatment, and care of people living with cancer.
How did you learn about the ACTs program?
Ajay Aggarwal: I’m from the United Kingdom, so I didn’t hear about the program through traditional sources. Instead, I learned about it from other U.S. researchers. I’ve been collaborating with Dr. Laurence Court, a medical physicist at MD Anderson Cancer Center and an ACTs program grantee. He brought it to my attention as a way to build on our ARCHERY Study. (That project is now funded by the Division of Cancer Treatment and Diagnosis and managed by Dr. Jeff Buchsbaum.) I contacted Dr. Paul Pearlman at NCI’s Center for Global Health, who runs the ACTs program, and had an extremely insightful and informative discussion, which then led me to submit an application.
Anant Madabhushi: Rao Divi, my program officer in NCI’s Division of Cancer Control and Population Sciences, brought this opportunity to my attention. He was aware of the work we were doing in the area of global health. For more than 5 years, I’d been working with Tata Memorial Center (TMC). TMC is not only the largest cancer center in India, but also in Asia. Our team had been investigating breast, oral, and cervical cancers. Also, I’m from India. I grew up in India, and that’s been a big factor in motivating me to think about addressing global health.
Can you briefly explain the aims of your research (under this grant)? How are you using AI and related technologies to advance your research goals?
Ajay Aggarwal: When we talk about cancer, we often focus on the increasing burden of cancer. To truly address this problem, we need to make high-quality care accessible to more people and in a shorter period of time. We can’t do that with our existing workforce alone. We need to think of innovative solutions, and AI, essentially computer algorithms, are showing potential for assisting with two key components of the radiotherapy pathway. This technology can help map the organs to precisely target radiation treatment. These tools also can help maximize the delivery of radiation to the tumor and minimize damage to surrounding healthy tissues.
We’re examining whether these algorithms can do this work with sufficient quality to support implementation. We also want to see if this approach saves time. If the technology saves time for practitioners, radiation oncologists, medical physicists, and radiation therapists, we want to know how this translates to cost savings.
Thus, we have three main objectives: We’re looking at quality, time savings, and cost savings when implementing AI in the radiotherapy planning and treatment pathway. And we want to see how this applies in diverse populations, so we’re looking at this research globally.
We’re examining three cancers: head and neck, cervical, and prostate. It’s almost like three studies in one. We’re working in four countries and six centers, including South Africa, at the University of Cape Town and Stellenbosch University; two cancer center sites in India, Calcutta, and Mumbai; and universities in Jordan and Malaysia. These areas have different populations with different tumor burdens. There’s also diversity from a system’s level, regarding the size and capacity of the treatment sites. This allows us to explore diversity from both a system and population level.
Anant Madabhushi: Our ultimate goals are to develop tools that are affordable, accessible, and equitable. Our specific research aims under this grant focus on the pathology of three different cancers—oral cavity, prostate, and triple-negative breast cancer. Using machine learning (ML) and machine vision (MV) tools, we hope to find subtle patterns in pathology images that we can use to create risk decision support tools. We not only want to be able to predict the outcome of treatment but also identify who will benefit the most from treatment.
We’ll be analyzing and interrogating pathology images from biopsies and developing more tailored population-specific prognostic and predictive models across these three different cancers, specifically within the context of the South Asian population.
We’re particularly interested in features and attributes strongly tethered to the pathobiology of disease. For example, we’re looking at characteristics that relate to immune cells, collagen patterns, multinucleation patterns, and mitotic figures—all hallmarks of cancer biology. We’ll quantitate these features using ML/MV to develop highly interpretable and accurate risk prediction tools.
The other key piece is we want to be sure we tailor these tools to the specific morphology and, thus, the phenotype of the South Asian population.
How will you translate your research into practice? What benefits do you hope will come from that?
Ajay Aggarwal: There’s already a lot of software in this space. Our work essentially serves as a large implementation research study or trial. Through our research, we can answer the question, “Can AI be used in diverse populations and different settings and achieve what we want with respect to quality, time savings, and cost savings?” If the technology meets those endpoints and outcomes, that could lead to four significant advances:
- Accessible. Greater access to healthcare will help to overcome a lack of workforce. We can expand services in places with the most need and not compromise quality.
- Affordable. Offering not-for-profit AI software as a web-based service for the public sector gives hospitals in low-income countries and settings a way to afford the technology.
- Scalable. The tool would be available to both large cancer centers and small community hospitals.
- Applicable. As the technology advances, we aim to look for additional funding to integrate new tumor types into this protocol.
Anant Madabhushi: As a bioengineer, I’ve spent a lot of time thinking about translating research to practice. One of the reasons I started our company, Picture Health, was because I wanted a vehicle that would allow me to translate the technologies we’re creating in the lab into clinical practice.
I think it’s imperative to keep the idea of translation front and center. It can be challenging to maintain your academic responsibilities and find additional time (that you don’t have) to work on ways to move the technology forward. Still, if the goal really is to make an impact, particularly with global heath and cancer, you have to be willing to go above and beyond.
We also need to consider other avenues for translation, so it’s important to take a multipronged approach. One example is laboratory-developed tests. We need to be creative and innovative in how we use these technologies. We need to find new ways to translate and disseminate these findings so that, ultimately, we impact patient care. Because, at the end of the day, that’s really the only thing that matters.
What advice would you give to others looking to apply for an ACTs grant, especially people in the data science field?
Ajay Aggarwal: My general advice is to be very sure of your research question and the need for the research in the first place—the “so what” question. Also, it’s not just about the quality of the data, but the scientists who are on your team. This includes people who add value in terms of their methodological input but also people who can help amplify the findings. Networking is critical, especially in my work, which necessitates collaborating with people from many different countries and settings.
We need to be sure there’s equity in these partnerships. When you first bring together people, many with different skill sets and from varied disciplines, there needs to be coherency among the group. This ensures that everyone understands the research question, how to conduct the research, and the expected output and potential impact of the work. Most importantly, the patient’s voice needs to shine through. Who are we ultimately trying to influence? Whose healthcare outcomes are we trying to improve? The patients and their needs should be a strong theme throughout the application.
Anant Madabhushi: It takes time and intention to develop the relationships needed for global collaboration. It’s not always smooth sailing. There’s a certain degree of skepticism about Western scientists. Researchers in other countries may have the idea that this is simply a “data grab.” That is, they worry that researchers from the United States or United Kingdom will take the data and commercialize or publish the findings, leaving them left out.
I think you have to be very intentional. You have to be persistent and develop relationships with your collaborators founded on trust. You can do all the Zoom calls in the world, but what I’ve found in my recent visit to India is that there’s no substitute for meeting in person. Having lunch, breaking bread, and talking about the research—all help to reinforce the trust.
I advise spending time “on the ground” with your collaborators. Most importantly, be willing to take time to create the collaboration. Just sending an email saying, “Hey, I want to work with you, send me your data,” isn’t sufficient. It takes time and effort to build these relationships.
What gives you the most satisfaction in your work?
Ajay Aggarwal: Finding the ultimate synergy between the right question, the right team, the right funder, and the right time—gives me a lot of satisfaction. Setting up our current research took time, about 3–4 years, and it was extremely hard work, but it was worth it. In some ways, that early work has helped our research go much smoother. Early on, we saw all the challenges that can arise, and we’ve learned ways of addressing those challenges through collaborations, friendships, and creativity.
Anant Madabhushi: One major inflection point for me personally occurred about 20 years ago. I lost my aunt, who lived in India, to triple-negative breast cancer. I realize now that this has been a major motivating factor for me. I intentionally chose my professional journey and career to look for ways to help people with cancer. I want to develop technologies that address the big questions in precision medicine, but always through the lens of affordability and accessibility.
Categories
- Data Sharing (64)
- Training (39)
- Informatics Tools (39)
- Genomics (35)
- Data Standards (34)
- Precision Medicine (32)
- Data Commons (32)
- Data Sets (26)
- Machine Learning (24)
- Seminar Series (22)
- Artificial Intelligence (20)
- Leadership Updates (13)
- Imaging (12)
- Policy (9)
- High-Performance Computing (HPC) (9)
- Jobs & Fellowships (7)
- Funding (6)
- Proteomics (5)
- Semantics (5)
- Awards & Recognition (2)
- Publications (2)
- Request for Information (2)
- Information Technology (2)
- Childhood Cancer Data Initiative (1)
Leave a Reply