Where Artificial Intelligence is Taking Medical Imaging

Nearly weekly, news stories and the popular media report that artificial intelligence is revolutionizing some aspect of healthcare. Some of the more recent examples include a Google algorithm that predicted the death of a patient with breast cancer more accurately than her doctors [1], a study from Canadian researchers that found an AI algorithm could predict risk of Alzheimer’s disease based on brain imaging, genetics, and clinical data up to 5 years before patients become symptomatic [2], and findings from an international study in which AI-based machines differentiated malignant melanoma (skin cancer) from benign moles more accurately than senior doctors [3]. The emerging role of AI in medical imaging is even giving medical students a reason to avoid specializing in radiology, as they worry their jobs will be largely taken over by computers [4].

Many in the field suggest that the AI and machine learning revolution is mostly (at this point) hype and point to well-publicized examples like IBM Watson’s missteps at MD Anderson Cancer Center in Texas [4] or a more recent study out of Mount Sinai, which found the health system’s deep learning algorithms used to identify patients with pneumonia from chest x-rays weren’t as accurate when it was tested on images from Indiana University and the National Institutes of Health [5]. As with many evolving technologies we expect to see success and failure – But successes like those described above suggest that the technology is beginning to find its place in healthcare, particularly in medical imaging.

San Francisco-based Arterys is a startup working on integrating deep learning and artificial intelligence with medical imaging. Their FDA-cleared, cloud-based platform integrates within existing clinical workflows, providing analysis for cardiac, lung, and liver imaging within minutes and making recommendations about specific areas to review or treatment options [7]. The AI system works alongside the clinician, reducing the time they have to spend on calculations and analyzing radiological images, so they can focus on other tasks. When integrated within the existing clinical workflow, it’s not difficult to see how a deep learning system like those of Arterys will change radiology.

Join me forepisode 15of the MoneyBall Medicine Podcastwhere I talk to Fabien Beckers, CEO of Arterys, where we discuss how the confluence of artificial intelligence, cloud computing, and medical imaging is creating a paradigm shift in healthcare. We’ll touch on the role of data-driven medicine to support value-based care and the importance of understanding the complexities of healthcare for technology companies entering the space.

References:

  1. Clark, S. 2018. Google AI predicted cancer diagnosis better than doctors. The Stack. https://thestack.com/big-data/2018/06/19/google-ai-predicted-cancer-diagnosis-better-than-doctors/
  2. Rohman, M. 2018. AI could predict risk of Alzheimer’s on MRI 5 years before symptoms emerge. Health Imaging. https://www.healthimaging.com/topics/artificial-intelligence/ai-could-predict-risk-alzheimers-mri-5-years-symptoms-emerge
  3. Donnelly, L. 2018. Robots are better than doctors at diagnosing some cancers, major study finds. The Telegraph (UK). https://www.telegraph.co.uk/news/2018/05/28/robots-better-doctors-diagnosing-cancers-major-study-finds/
  4. Walter, M. 2018. Rise of AI keeping some medical students away from radiology. Radiology Business. https://www.radiologybusiness.com/topics/leadership/ai-medical-students-radiology-medical-imaging
  5. Jaklevic, M. C. 2017. MD Anderson Cancer Center’s IBM Watson project fails, and so did the journalism related to it. HealthNewsReview.org. https://www.healthnewsreview.org/2017/02/md-anderson-cancer-centers-ibm-watson-project-fails-journalism-related/
  6. Sullivan, T. 2018. Mount Sinai finds deep learning algorithms inconsistent when applied to outside imaging sets. Healthcare IT News. https://www.healthcareitnews.com/news/mount-sinai-finds-deep-learning-algorithms-inconsistent-when-applied-outside-imaging-data-sets
  7. Combs, V. 2018. AI platforms aim to ease information overload in healthcare and improve patient care. TechRepublic. https://www.techrepublic.com/article/ai-platforms-aim-to-ease-information-overload-in-healthcare-and-improve-patient-care/