{"id":19208,"date":"2016-11-18T16:24:37","date_gmt":"2016-11-18T21:24:37","guid":{"rendered":"https:\/\/digital.hbs.edu\/platform-rctom\/submission\/doctor-watson-will-see-you-now\/"},"modified":"2016-11-18T16:24:37","modified_gmt":"2016-11-18T21:24:37","slug":"doctor-watson-will-see-you-now","status":"publish","type":"hck-submission","link":"https:\/\/d3.harvard.edu\/platform-rctom\/submission\/doctor-watson-will-see-you-now\/","title":{"rendered":"Doctor Watson will see you now"},"content":{"rendered":"

Introduction<\/em><\/strong><\/p>\n

Will your next x-ray be read by an algorithm? Will artificial intelligence replace radiologists? Merge Healthcare seems to think so. Technological advances in deep learning combined with the digitization of medical images has created one of the largest opportunities to disrupt the practice of radiology.<\/p>\n

What is Radiology<\/em><\/strong><\/p>\n

Radiologists diagnose and treat diseases and injuries through the use of medical imaging techniques such as x-rays, CT scans, and MRIs. The role of a radiologist is to function as a consultant to the referring physician by helping choose the appropriate diagnostic tests, interpreting the medical images, and recommending further treatment as needed (1).<\/p>\n

The Digitization Revolution<\/em><\/strong><\/p>\n

Over the last 10 years the field of radiology has transitioned away from physical film towards digitized medical images. Much like Kodak and the advent of digital photos, many companies that manufacture x-ray equipment or physical chemicals\/film have gone bankrupt due to their inability to adapt to an increasingly digital world. Digital records disrupted the specialty by enabling remote access to patient records, removing physical barriers to sharing and storing information, and delivered better quality \/ higher resolution images that improved diagnosis. From a process perspective, the digitization of medical images improved workflows, increased productivity in terms of number of patients seen per day, eliminated the costs of chemicals and film, and reduced error rates (2).<\/p>\n

However, the downside of digitization is that the amount of information a radiologist has to manage has grown exponentially. The number of images radiologists have to interpret has increased 100 fold in the last 20 years with some radiologists reviewing 20,000 studies per year (3). Doctors now have to process increasingly more detailed and complex images while trying to simultaneously cross-reference a patient\u2019s medical history, lab data, and the latest medical research to make the most accurate diagnosis.<\/p>\n

Overview of Merge <\/em><\/strong><\/p>\n

Merge Healthcare sits at the center of the technological revolution in radiology. Merge is one of the largest providers of imaging processing software with the largest database of digital medical images having processed 30 billion images to date across 7,500 healthcare facilities in the US (3). The company\u2019s technology delivers value to its users, who are largely physicians and other healthcare providers, by creating a common platform that enables radiologist to view, share, and interpret clinical images in a more productive, accurate manner. \u00a0As a result of its unique access to a large share of US radiologists and its vast library of historical images, Merge has the opportunity to be at the forefront of the next major disruptive technological innovation in the industry: machine learning.<\/p>\n

Machine Learning: The Opportunity<\/em><\/strong><\/p>\n

Machine learning has the potential to meaningfully improve the speed and accuracy of medical diagnosis by processing vast amounts of imaging data more quickly and accurately than a physician. In a world where a typical trauma patient\u2019s \u201cpan scan\u201d results in over 4,000 images it\u2019s no wonder artificially intelligence is better equipped to avoid missed or inaccurate diagnosis resulting from visual fatigue (4). Deep learning algorithms can deliver value by improving both image processing and image interpretation. On the processing front, machine learning algorithms can help cut through the noise and extract the most relevant features from medical images as well as cross references images with a database of scans and the latest medical research. On the interpretation front, deep learning can not only improve the identification, classification, and quantification of disease patterns from images, but also generate predictive insights into the most relevant care pathway (5).<\/p>\n

The Next Evolution of Merge<\/em><\/strong><\/p>\n

In 2015, Merge was acquired by IBM for $1bn in a move that cemented the shift in its business model away from purely serving as an imaging workflow platform towards the development of a sophisticated smart diagnosis IT service for radiologists (3). By leveraging Merge\u2019s platform and database with IBM Watson\u2019s machine earning algorithms, the combined business hopes to disrupt the field of radiology by offering IT solutions that can dramatically improve a doctor\u2019s ability to accurately diagnose and predict diseases. IBM Watson\/Merge\u2019s foray into medical imaging AI has the potential to disintermediate the profession of radiology, raising important ethical and regulatory questions about machines ability to make medical decisions. From a business model perspective, Watson\/Merge should carefully consider the scope of the initial product and its positioning to clinicians. I would focus on selling a more \u201cassistive\u201d version of the software that enables clinicians to make better decisions rather than one that generates independent diagnosis. For example, the software should pre-highlight key areas of concern in an image and synthesize relevant takeaways from the image database or literature \u2013 leaving the ultimate medical decision to the doctor. A major barrier to the commercial success of the product will be the physician adoption rate so designing a solution that physicians can trust without worrying about disintermediation will be critical.<\/p>\n

 <\/p>\n