Mark Lambrecht, PhD, is director of the health and life sciences global practice for SAS, a data analytics firm. Here, he weighs in on artificial intelligence's role in healthcare and how it is affecting gastroenterology.
Note: Responses have been lightly edited for style and clarity.
Question: What do you think are the key AI trends to watch right now?
Dr. Mark Lambrecht: AI is getting a tremendous amount of attention, but to move from AI hype to AI science, more work is needed at many levels. AI (e.g., deep learning and natural language understanding) is just a subfield of analytics. There are applications where AI alone can drive positive results, but there are many more applications where the mix of AI, analytics, process change and physician interaction with technology will be the catalyst. To bring AI’s predictive and diagnostic power to the individual patient will require integration of clinical, financial, demographic, genetic and physiological data —- an enormous undertaking.
One trend is the delivery of AI and machine learning technologies into a clinician’s reality. But an AI system with great sensitivity or specificity is meaningless if it does not actually improve clinical reality and ultimately patient outcomes. This is not unlike new and innovative medicines that have a great efficacy in a test phase, but do not find the uptake in the real world with both clinicians and patients.
Another area to watch is the necessity to better manage risk, safety, inherent bias and regulation of AI methods in development. AI engineers need to abandon "the science experiment," and starting in the design phase, build algorithms that respect privacy and have validated clinical evidence.
Q: What are the current applications for AI (including machine learning and neural networks) in gastroenterology? What is SAS' contribution to AI in gastroenterology?
ML: Machine learning has been shown to be very effective for the development and testing of new medicines, studying large populations of patients for trends or patterns and predicting which patients are likely to be readmitted to a hospital. SAS has been used for decades to aid healthcare providers and payers in ensuring that healthcare resources are deployed effectively, minimizing waste and cost.
Manual examination of tumors and lesions is both time-consuming and subjective for radiologists. Machine vision excels at recognizing and interpreting medical images with greater accuracy and speed. Deep learning AI methods, like computer vision diagnostic techniques, can now help gastroenterologists evaluate whether polyps found during colonoscopy are precancerous or an adenoma.
Colorectal cancer is the third most common cancer worldwide, and in about half of patients, the cancer spreads to the liver. SAS is partnered with Amsterdam University Medical Center in the Netherlands on an AI medical-imaging initiative to help patients with colorectal liver cancer live longer, more productive lives. SAS’ AI-trained models help AUMC physicians identify patients who respond well to chemotherapy and become candidates for surgery.
Medical image applications using computer vision and predictive analytics provide evaluation criteria (called RESIST - response evaluation criteria in solid tumors) that are more objective, accurate and automated than the current manual ones.
Q: What do you think AI's potential impact on patients (and healthcare as a whole) will be?
ML: Of course, AI technology has applications beyond response assessment in colorectal liver cancer. From a clinical standpoint, it can be used to assess many other types of solid tumors, such as breast and lung cancers. A surgeon can spend more time making complex and logical decisions on a treatment path and communicating this to a patient.
Healthcare is transitioning from the era of disruption to transformation, impacting how clinical trials are run or how patients in the hospital are evaluated. With the application of new AI techniques —- from lab tests, wearables, CT scans, genomics and EHR information —- data volumes and clinical information will explode. AI can help decrease physician burnout by managing repetitive or standard tasks (e.g., reporting normal lab results to patients). Computers are much better at rapidly and accurately evaluating images, sound, EHR or text information and offer the opportunity for physicians to spend more time on high-value clinical tasks and interaction with patients.
Q: What GI technologies are you excited for in the future?
ML: Next to AI-driven optical biopsy and the ability to integrate multiple data sources, we are excited about the ability of AI to power up non-invasive techniques such as sound and ultrasound analysis. In developing countries that lack modern hospital infrastructure and medical equipment, affordable mobile devices equipped with AI algorithms can provide healthcare access to remote areas and reduce costs. AI will not only revolutionize telemedicine, but also generate enormous amounts of structured information that can catalyze further research and delivery of new AI methodologies for application in GI medicine and beyond.