To efficiently integrate artificial intelligence (AI) tools into community cancer care, technical, clinical, economic, regulatory and ethical challenges must first be adequately addressed, according to Kashyap Patel MD, ABIM, BCMAS.1
“AI generates information based on thousands of resources,” Patel, the chief executive officer of Carolina Blood and Cancer Care Associates in Rock Hill, in South Carolina, explained in an interview with OncLive® during the inaugural
Overcoming AI Challenges in the Community Setting: Key Takeaways
- AI has the potential to transform community practice through screening, diagnostic, risk stratification and monitoring, and treatment personalization applications.
- Its current implementation at scale is limited by technical, clinical, economic, and regulatory and ethical challenges.
- In order to combat these challenges, pilot AI programs are being tested in specific screening applications, data integration infrastructure is being built out, and providers are being trained in the use of AI applications.
How Can AI Enhance Community Practice?
Patel began his presentation by summarizing the potential benefits of AI in cancer management, first focusing on its ability to aid in treatment selection and in providing personalized care for patients. He noted that the technology allows for enhanced treatment selection via its ability to quickly analyze multiple clinical factors simultaneously, including disease genotype, patient prognosis, molecular genetics, and therapeutic targets. This approach can lead to reduced adverse effects for patients by eliminating ineffective treatments, improved outcomes through optimal treatment selection, the integration of tissue diagnosis with next-generation sequencing (NGS), and personalized therapeutic protocols based on individual patient characteristics, he added.
Regarding diagnostic accuracy, Patel noted that AI-based tools have the capability to examine complex microscopic features in blood and tissue samples, determine subtle patterns that cannot be detected by humans and reduce diagnostic variability in order to improve consistency. Additionally, AI can be used to integrate historical data from flow cytometry, fluorescence in situ hybridization, cytogenetics, and NGS, he added.
AI can also be used in cancer screening, offering the potential to improve screening accuracy and reduce the numbers of false positives and unnecessary procedures, Patel said. In lung cancer screening, AI is able to analyze low-dose CT scans for suspicious nodules and elucidate the likelihood of a tumor based on location, type, and appearance patterns. In terms of breast cancer screening, AI-enhanced mammography can identify how lesion locations, appearance characteristics, and size correlate with the probability of a malignancy.
There are currently multiple AI-based tools that have been cleared by the FDA to aid radiologists in detecting breast cancer from mammograms.2 The regulatory agency has also given the green light to AI algorithms that can help to interpret MRIs and ultrasounds. These tools are being prospectively evaluated in breast cancer care.
“During a regular screening mammography, AI can look at the likelihood of a spot being malignant, then we can do a proactive biopsy,” Patel explained in the interview. “Once a patient is diagnosed with cancer, AI can help us distinguish in detail between the likelihood of it being very aggressive or slow growing. [It can] also help us filter which patients may be eligible for immunotherapy, who would be best treated with chemotherapy, or who may be best for targeted therapy.”
He also explained that AI is useful in risk stratification and patient monitoring.1 The technology is able to identify high-risk patients, monitor for minimal residual disease for early recurrence detection, predict treatment resistance and enable proactive therapy modifications, and optimize follow-up schedules based on personalized patient risk profiles.
What Are the Present Challenges With AI in Community Practice and How Can They Be Addressed?
Patel then spotlighted the current challenges that are being faced in terms of implementing AI into community practice. He noted that some of the technical challenges with the technology include integrating data across multiple platforms and institutions, ensuring that AI models perform consistently across patient populations and disease settings, building out the necessary computing resources, and integrating the systems with existing workflows.
In terms of clinical challenges, physician acceptance, the need for training, clinical validation of AI-based tools, and liability concerns stand as significant barriers to the wide adoption of AI at the community level. He also noted that data privacy, algorithmic bias, and explaining to patients how AI contributes to clinical decision making could hinder the fast uptake of the technology. The significant up-front investment the technology requires along with maintenance expenses and potential workflow disruptions represent economic barriers to AI’s adoption in community practice, he added.
In order to address the challenges posed by the widespread uptake of AI at the community level, Patel proposed short-terms actions. Within the next 1 to 2 years, he advocated for the piloting of AI programs in specific cancer screening applications, the development of data integration infrastructure for multimodal analysis, the establishment of partnerships with AI technology vendors, and the initiation of provider training programs. In the medium term, within the next 3 to 5 years, he envisions the full deployment of AI-enhanced diagnostic workflows, the integration of AI-driven treatment selection protocols, the implementation of comprehensive AI patient monitoring systems, and the establishment of outcome measurement and quality assurance programs. This will lead to the long-term development of comprehensive AI-driven precision medicine programs with continuous learning and adaptation capabilities, he explained.
Patel noted that emerging AI technologies which are able to integrate multiomics data, monitor patients in real-time through wearable devices, and transparent AI models represent future opportunities for the technology in community practice. In order to achieve these goals, he added that collaborative AI training across institutions with an emphasis on preserving patient privacy will be needed.
“As we move [more] into personalized medicine, there are so much data that are evolving,” Patel said. “There are hundreds of new mutations being discovered every few months, and there’s so much research coming out about new targeted therapies. For [us] to continue to learn, evolve, and implement these changes, [time is a] barrier because we have a limited capacity of absorbing information. AI can help us streamline [our processes in terms of] what testing is appropriate [and] what treatments are available.”
References
- Patel K. Overcoming AI hurdles and challenges in the community setting. Presented at: MiBA Community Summit; September 27-28, 2025; Scottsdale, Arizona.
- Lotter W, Hassett MJ, Schultz N, Kehl KL, Van Allen EM, Cerami E. Artificial intelligence in oncology: current landscape, challenges, and future directions. Cancer Discov. 2024;14(5):711-726. doi:10.1158/2159-8290.CD-23-1199