Integrating AI into the Clinical Lab

Artificial intelligence (AI) is making great strides in health systems with a multitude of uses such as improving clinical care delivery, refining the scientific understanding of various conditions and providing greater operational efficiency. Organizations with large data sets are well-positioned to take advantage of AI’s capabilities. In a Dec. 2 webinar, Dr. Stan Letovsky, vice president of the Center of Excellence in AI, Bioinformatics, and Data Sciences at Labcorp, shared innovative ways that Labcorp is using AI, and explained how health systems can benefit too.

1 AI can aid in test result prediction

AI can provide physicians with clinical decision support, particularly to determine which tests to order for patients. With AI, information can be extracted from patient histories and specific test results for certain disease conditions. After dividing results into cohorts, the machine learning engine can be used to predict which patients are more likely to be diagnosed with certain diseases. The goal is not to avoid performing future tests, but to identify the risk of patients having an abnormal test result before it’s performed.

2 Use test results to predict disease progression

Deep learning models can be used to understand the progression of disease ahead of time, allowing clinicians to potentially schedule or design longer term treatment plans based on that information. These models can also include a feedback loop, where the clinician’s actions and patient’s outcomes are incorporated back into the algorithm. This reinforcement learning model updates the results and makes it even more accurate going forward.

3 AI can highlight patterns of care in large datasets

AI can be used in the management of population health. For example, primary care practices can proactively target potential medical issues in their community based on risks predicted through AI. Offering early testing, diagnosis and treatment can improve outcomes and lower costs. In another application, natural language understanding was used in the analysis of deidentified patient records to determine risk factors predictive of severe outcomes from Covid-19.

4 AI can provide physician assistance

AI is used to grade biopsy specimens and identify and measure areas of interest in imaging studies and pathology samples in subspecialties like dermatology, oncology and radiology. AI algorithms can often provide this information faster and more accurately than a human, freeing doctors to interpret and perform higher level work. The goal is not to automate physicians’ work, but to provide tools making them more productive and efficient.

5 Help identify patients for clinical trials

The idea of clinical research as a care option is gaining traction in the healthcare community. What’s needed is the ability to bring the right trial to the doctor or patient’s attention at the right time. AI can be used to analyze clinical trial protocols using natural language processing. The algorithm can incorporate trial inclusion criteria and convert it into a database query. Then, AI can analyze patient’s test results and disease conditions to give an educated guess about whether the patient might be a match for the trial. Patients will then be informed about the trial and go through the qualifying process if interested. AI gives a head start in that process, efficiently identifying potential patients and increasing recruitment levels.