Applying Artificial Intelligence, Machine Learning in Cancer Care

AI in Healthcare

One of the most notable applications of AI/ML in healthcare is drug discovery and development. The drug discovery, research, and development process have become increasingly expensive and competitive over time. In an interview with PharmaNewsIntelligence, Kevin Dondarski, partner at Deloitte Consulting Life Science Strategy Analytics Practice, revealed that it takes roughly $2.3 billion to discover and bring a pharmaceutical asset to market.

Beyond the funding required for preclinical discovery and research, clinical trial budgets must also account for administrative fees, time, sites, data collection and analytics, patient recruitment, the cost of the treatment, follow-up care or monitoring, and personnel. With so many considerations and a statistically high probability of failure, the price tag on pharmaceutical innovation has grown.

According to an article in Drug Discovery Today, AI has accelerated the pharmaceutical lifecycle, with applications in drug discovery, pharmaceutical product development and management, pharmaceutical manufacturing, clinical trial design and monitoring, and quality assurance.

AI in Oncology

Across the oncology space, clinicians have explored the use of AI, ML, and other forms of deep learning for predictive diagnostics, treatment, and patient monitoring.

“There is exploration and application of all forms of AI in the cancer care processes, from operations, finance, and clinical,” Susan Hoang, vice president of oncology intelligence and analytics at McKesson, told PharmaNewsIntelligence. 

The utility of AI in healthcare and its ability to improve patient outcomes and the healthcare system as a whole has led to the rapid integration of AI in this industry.

“While AI’s traction in healthcare is behind other industries, like consumer-based businesses, we should see an increase in AI application in the cancer care process,” Hoang added.  

Cancer Diagnosis and Detection

One of the fastest integrations of AI into oncology is cancer diagnosis and detection. In clinical practice, early and accurate cancer diagnostics are one of the most important indicators of survival. With a deep learning approach, researchers can use big data to train technology for cancer detection.

Imaging is a critical tool in cancer detection and diagnosis. Typically, radiologists manually assess imaging, looking for cancer cells, tumors, lesions, and other indicators of varying cancer types.

Using computers or AI to analyze medical images or components of medical images is called radiomics. According to a publication in Future Science, AI-based cancer imaging may include the following categories:

• Classification
• Detection
• Segmentation
• Characterization
• Monitoring

AI and ML methods can enhance these imaging components, allowing for the rapid detection of mutations and other cancer indicators.

Beyond standard applications to enhance existing imaging techniques, ML approaches may also be used to develop generative adversarial networks (GANs), which can generate new images based on different existing images, alleviating the diagnostic burden on patients, providers, and the healthcare system.

Early Detection and Risk Assessments

In addition to diagnosing patients when cancer is suspected, AI integration into routine screenings, such as mammograms for breast cancer and MRIs or ultrasounds for prostate cancer, may facilitate earlier diagnosis.

Another example is the integration of AI technology into colorectal cancer screenings, which has improved the detection of adenomas. Additionally, AI has been deployed in CT scans to screen for lung cancer and other complications.

AI technology may also predict cancer risk based on gene expression, minimizing the need for expensive and lengthy gene sequencing procedures.

Despite the utility of AI in radiology and pathology, there are still concerns about overdiagnosis, leading to an excessive number of biopsies and an increased burden on patients and pathologists. Validation tests are conducted before launching AI technology to minimize the risk of error.

Cancer Treatment

The robust datasets generated by AI technology reach far beyond cancer risk assessments, providing insights into the best interventions or treatment options.

AI can enhance precision oncology, allowing for more personalized treatment, clinical decision-making, and beneficial patient outcomes.

Providers can use deep learning-based prediction models or ML algorithms to develop decision support systems, assess the probability of the success of a specific treatment, and make more informed treatment decisions.

In addition to clinical decision-making based on population data, AI can improve precision medicine in cancer care by analyzing a patient’s cancer genome, assessing genetic profiling data or biomarkers, and reviewing histology or histopathology information.

For example, McKesson has been leveraging ML to improve advanced care planning for cancer patients. In December 2020, the company implemented the McKesson Advance Care Planning Enrollment eXtended (APEX) mortality risk predictive analytics model to improve prognostic awareness in the OCM population and the timing of the initiation of end-of-life care. 

“We took a design-thinking approach when building this solution by considering how the care team works and interacts with patients. This is important as most AI or ML projects often fail in the field because most models do not account for the complexity and variability of cancer care,” Hoang said.  

“We also co-designed this tool with clinicians. I think the key is to build trust, and we were able to do that,” she added.  

Cancer Research

Beyond existing diagnostic and treatment tools, oncology research has focused extensively on AI and ML integrations. Recently, researchers have used AI/ML tools to assess the efficacy of immunotherapy approaches.

Although assessing risk and treatment data is a vital component of AI/ML in oncology, clinical trial enrollment is an even greater area for AI/ML integration.

Considering the significance of clinical trials in drug development and medical advancements, patient recruitment and diversity are vital to oncology research. However, financial barriers, logistical issues, a lack of resources, and an inability to support enrollment and retention may hinder clinical trial participation.

Only 3% of cancer patients are enrolled in clinical trials. Moreover, 20% of cancer-related clinical trials fail due to insufficient patient enrollment. Low trial accrual impedes patient access to new treatments and limits the applications to the overall population.

AI/ML is uniquely positioned to disrupt the current approach of clinical trials, with applications ranging from patient recruitment to adherence monitoring, data collection, and data analysis. For example, AI-enabled digital transformation can improve patient selection and increase clinical trial effectiveness by mining, analyzing, and interpreting multiple data sources.  

Clinician Burnout

Much like its impacts in other industries, AI/ML may alleviate the burden on workers in the industry. Exacerbated by the COVID-19 pandemic, rates of clinician burnout have continued to rise across the United States. As a result, patient care, including cancer patient care, may be compromised.

AI can play a critical role in streamlining provider workflow, potentially cutting down clinical documentation time and freeing providers to interact directly with patients.