Researchers underscored the importance of addressing these barriers, as artificial intelligence may offer improvements in clinical decision making, broaden access to care, and improve clinical efficiency.
As uptake of artificial intelligence (AI) in cancer care continues to grow, researchers highlighted barriers that remain and laid out what they envision for the future of the technology.
In a recent paper, researchers underscored the importance of addressing these barriers, as the potential of AI in the space could offer improvements in clinical decision making, broaden access to care, and improve clinical efficiency.
“The inherent organizational complexity of cancer care delivery, the need to interpret and synthesize vast amounts of data from different sources, the narrow therapeutic window of treatment, and the heterogeneity of cancer make oncology a challenging, yet ideal area to develop and implement AI tools,” authors wrote.
To date, AI in oncology has shown particular promise in cancer imaging, including use for digital pathology, radiographic imaging, and clinical photographs. These tools can also be used to accurately estimate a patient’s risk of various outcomes—promoting precision oncology—and aid in cancer prevention efforts.
However, various challenges and questions remain, researchers explained. These include:
“The challenges facing AI in oncology are formidable and span the entire ecosystem of oncology care,” researchers said. “Yet, these challenges are surmountable and can be addressed methodically and systematically.”
In addition to offering a framework for addressing these barriers, authors pushed for training and educating oncology providers so they can be adopters of AI-based clinical decision support systems and demonstrate the effectiveness of these tools. Researchers also recommended the continuation of standardizing oncology terminology to promote structured data elements in reports.
Other recommendations included formalizing standards for external and continuous validation of AI models, and increasing research on algorithm fairness to minimize AI bias.
Creating a consensus around a framework that fairly assigns liability resulting from AI-related error in an effort to build trust around AI-based tools—while ensuring quality and safety—and running randomized controlled trials to demonstrate AI-based tools improve patient outcomes are also steps necessary to better incorporate AI into the field.
Reference:
Chua IS, Gaziel-Yablowitz M, Korach ZT, et al. Artificial intelligence in oncology: path to implementation. Cancer Med. Published online May 7, 2021. doi:10.1002/cam4.3935