AI in Cancer Care: Innovations in Detection, Treatment, and Patient Support

By Nicholas Zuk

            Every year, millions of people worldwide are diagnosed with cancer. Because of its high unpredictability, various forms, and difficulty to treat, cancer remains one of the most complex challenges in modern medicine. While there have been groundbreaking advancements in our ability to treat cancer and its many forms, navigating a cancer diagnosis is still highly challenging for both patients and physicians. Delayed diagnosis, ineffective generalized treatments, and the overwhelming forms of cancers can lead to confusion, missteps, and missed opportunities for better patient outcomes. With the introduction of AI, however, novel cancer detection, analysis, and treatment methods are being developed that are revolutionizing cancer treatment. Further, AI brings the possibility of making advanced cancer diagnostic tools available to underserved and rural communities. Learning about how AI is being developed and integrated into the medical field can provide valuable insights into how these technologies are transforming the oncology landscape and paving the way for greater confidence and precision in cancer treatment.

            Detecting cancer in its initial stages is critical for improving treatment success and cancer outcomes. AI’s ability to identify cancers in early stages can significantly improve health outcomes. This can be particularly beneficial for detecting cancers, such as pancreatic and ovarian cancers, that are usually caught late making treatment more difficult. Traditional detection methods, however, can fall short due to the subtlety of initial symptoms and lack of access to timely screenings. If a patient begins to feel ill, it is easy to dismiss the symptoms as a minor illness and forego proper cancer screenings. Additionally, screenings can be arduous to schedule, and results often take up to 3 weeks to be available. AI, however, is being utilized to address many of these issues.

            To reduce the time it takes to receive results from medical imaging, AI is being used to quickly and accurately analyze mammograms, CT scans, and MRIs to identify patterns that may indicate cancer that may not be detectable by the human eye. Tumor analysis is a key part of cancer diagnosis and treatment planning. However, traditional tissue tumor analyses, where a tumor is dissected by a pathologist for evaluation, can be time-consuming and it relies heavily on the evaluator’s expertise. AI-powered image recognition systems are employed to examine tissue samples with extreme accuracy to reduce the time it takes to evaluate tumor scans. PathAI and Paige AI are two models currently being used to identify cancer subtypes, grade tumors, and predict disease progression by analyzing data from pathology slides (PathAI, 2016 & Paige AI, 2018). In a recent groundbreaking study, Google’s AI model, iCAD, used for breast cancer detection, outperformed human radiologists’ diagnosing ability by reducing false negative and false positive diagnoses (McGrath, 2024). The model is commercially available at 7,500 mammography sites worldwide (Park, 2022). Programs can process scans in minutes, what takes human pathologists hours to examine.

            AI is also being used to quickly process genomic and molecular data. AI excels at processing large and complex datasets, making it an invaluable tool for studying genetic and molecular tumor profiles to identify mutations or biomarkers that can indicate the most effective treatments for a patient. An example of this is Tempus, a platform that analyzes patients’ genomic and clinical data to recommend treatments tailored to a patient’s unique cancer profiles that oncologists can then review (Tempus, 2015). This is yet another example of AI being used to analyze data more timely than traditional, human-driven methods allow for. With AI, it is now possible to quickly and accurately determine whether a tumor is more likely to respond to immunotherapy or if better treatment options are available. For instance, AI can analyze the expression of PD-L1 proteins in a type of cancer to determine the effectiveness of immune checkpoint inhibitors (a class of drugs that harnesses the body’s immune system to attack cancer cells. Utilizing AI’s capabilities to answer these questions will be widely available in the near future and is a promising way to reduce misdiagnosis rates, ensure patients get the most effective treatment plans from the beginning, and improve health outcomes worldwide.

Beyond imaging, AI is being utilized to advance non-invasive evaluations through liquid (instead of tissue) biopsies that analyze the biomarkers in a patient’s blood to detect cancer cells or DNA fragments from tumors. Results from a tissue tumor biopsy typically take 2-3 weeks. With AI-assisted liquid biopsies, however, results can be more precise and take only a week to be made available (Connal et al., 2023). Trying to detect cancer, especially at an early stage, can be extremely difficult due to the large amount of data gathered in a screening. The slightest anomaly in a data set or image can be indicative of cancer but it can also be incredibly difficult for a human to notice. With the addition of AI for processing these complex systems, the chances of missing these discrepancies can be reduced, and the chances of effectively catching cancer in its initial stages can be increased (Bi et al., 2019). Further, because of AI’s portability, early detection access can be made available to underserved areas that would otherwise not have the resources or facilities to be screened by specialists (Baron & Haick, 2024). With AI, it may soon be possible for everyone, regardless of community infrastructure or resources, to equally benefit from cutting-edge cancer screening technologies while placing less strain on the healthcare system. 

            In addition to identifying and providing initial treatment recommendations for cancer, AI also has real-time monitoring capabilities that can examine a treatment’s effectiveness and recommend adjustments if necessary. By integrating data from wearable devices, imaging scans, and blood tests, AI models can analyze how patients respond to their treatment (Lipkova et al., 2022). This process ensures that patients receive the most effective care that minimizes the risk of their disease progressing. 

            While AI advancements in a clinical setting are incredible for cancer treatment, AI chatbots such as ChatGPT, Google’s Gemini, Claude, and Perplexity are tools that can help support and educate patients. Receiving a cancer diagnosis can be an overwhelming experience that leaves patients with countless questions about their condition, treatment options, and how to live their life with cancer. AI chatbots are a great tool that can provide personalized information to patients based on the prompts they are given. For example, patients can ask chatbots to explain a diagnosis or how to mitigate the side effects of chemotherapy during treatment in a fraction of the time it would take using a search engine or communicating with a doctor. An additional resource specifically helpful for cancer patients is CancerAcademy. CancerAcademy utilizes AI to curate educational materials, tips, and resources to help prostate cancer patients better understand and navigate their diagnosis. Perhaps the most significant contribution of AI chatbots is their ability to reduce disparities in access to cancer education. AI can offer simplified explanations of complex diagnoses, provide guidance during treatment, and suggest alternative resources. This tool ensures all patients, regardless of socioeconomic status or geographic location, can access the tools they need to navigate their diagnosis. Care should be taken, however, to ensure information is accurate and that sources provided are legitimate.

            AI and its applications to cancer detection, treatment, and education have immense potential, but it is not without flaws. First, AI systems rely on large datasets to train their algorithms. The quality of these datasets determines how accurate or applicable the AI’s results are to a patient. For example, if a dataset is incomplete, unrepresentative, or biased, then an AI model cannot effectively apply the dataset to a patient. This can lead to a misdiagnosis or an incorrect treatment recommendation. Second, integrating AI into existing healthcare systems is logistically and technically challenging. Many healthcare facilities still lack the necessary infrastructure to implement and maintain advanced AI technologies. Additionally, training healthcare providers to use AI programs is expensive and time-consuming. As a result, most institutions will likely not fully incorporate AI into their systems for another 5-10 years (Davenport & Kalakota, 2019). Third, AI needs access to sensitive patient information such as genetic and medical records to analyze data. This raises concerns about ensuring AI complies with U.S. HIPAA and European GDPR regulations for data security and patient privacy. Finally, many healthcare providers remain cautious about AI’s reliability and the possibility of it undermining human expertise (Mennella et al., 2024)). Despite these doubts and limitations, AI is a promising tool for cancer treatment now and in the future. 

While AI should be approached with caution, working to advance its capabilities can revolutionize cancer treatment in the future. AI has already been shown to be highly effective at cancer detection and analysis and helpful in prescribing treatment plans. With ongoing development and innovation into AI, advancements have the potential to close gaps in care, predict disease progression, create novel treatment methods, and overall improve patient outcomes worldwide. Approaching AI as a complement to healthcare providers’ judgment, training, compassion, and empathy can make cancer care more patient-focused, where technology and human expertise work together to provide precise and effective treatments. 

References

Baron, R., & Haick, H. (2024). Mobile Diagnostic Clinics. Retrieved from https://pmc.ncbi.nlm.nih.gov/articles/PMC11217950/ 

Bi, W. L., Hosny, A., Schabath, M. B., Giger, M. L., Birkbak, N. J., Mehrtash, A., … Aerts, H. J. W. L. (2019). Artificial Intelligence in cancer imaging: Clinical challenges and applications. Retrieved from https://pmc.ncbi.nlm.nih.gov/articles/PMC6403009/#:~:text=Although%20conventional%20biostatistics%20and%20machine,accurately%20differentiate%20between%20benign%20and 

Connal, S., Cameron, J. M., Sala, A., Brennan, P. M., Palmer, D. S., Palmer, J. D., … Baker, M. J. (2023). Liquid biopsies: The Future of Cancer Early Detection. Retrieved from https://pmc.ncbi.nlm.nih.gov/articles/PMC9922467/#:~:text=They%20achieved%20an%20overall%20specificity,identify%20five%20types%20of%20cancer 

Davenport, T., & Kalakota, R. (2019). The potential for artificial intelligence in Healthcare. Retrieved from https://pmc.ncbi.nlm.nih.gov/articles/PMC6616181/#:~:text=The%20greatest%20challenge%20to%20AI,to%20work%20alongside%20artificial%20intelligence. 

Lipkova, J., Chen, R., Chen, B., Lu, M., Barbieri, M., Shao, D., … Chen, T. (2022). Artificial Intelligence for Multimodal Data Integration in oncology. Retrieved from https://www.sciencedirect.com/science/article/pii/S153561082200441X 

McGrath, D. E. (2024). Does google AI beat the doctor at detection of breast cancer? Retrieved from https://ezra.com/blog/google-ai-breast-cancer-detection#:~:text=Google’s%20Contribution%20to%20AI%20in,cancer%20that%20the%20AI%20missed 

Mennella, C., Maniscalco, U., De Pietro, G., & Esposito, M. (2024). Ethical and regulatory challenges of AI technologies in Healthcare: A narrative review. Retrieved from https://pmc.ncbi.nlm.nih.gov/articles/PMC10879008/ 

Paige AI. (2018). About Us. Retrieved from https://paige.ai/about/ 

Park, A. (2022). Google’s AI will now be used in mammograms. Retrieved from https://time.com/6237088/mammograms-google-ai/ 

PathAI. (2016). About us | pathai. Retrieved from https://www.pathai.com/about-us/ 

Tempus. (2015). Ai-enabled Precision Medicine. Retrieved from https://www.tempus.com/