FDA Clears AI Applications to Assist Radiologists

Artificial intelligence (AI) applications in radiology have been and are being developed by the established imaging companies and a host of new companies. Many of the recent FDA cleared applications have focused on the radiologist, aiding image analysis or improving workflow.

The major imaging companies both develop AI applications, often to improve scanner image acquisition and incorporate 3rd party solutions. Some companies offer a platform to host AI applications from companies that have been cleared by the FDA, either via the 510 (k) process or the De Novo process.

Applications to assist radiologists fall into three categories: workflow improvements, clinical decision support, or image interpretation. Workflow improvements include providing relevant patient summaries extracted from the EHR, ordering worklists to place the most critical preliminary findings first and volume segmentation and automatic labeling. Image interpretation provides identification of suspect lesions and nodules.

The table on the right (click to enlarge) presents some AI Algorithms cleared by the FDA as of October 2018, many of which received clearance in 2018. Please notify us of other FDA cleared AI applications at info@medtechcon.com

Some companies have developed AI platforms to host AI algorithms from 3rd party developers. Envoy AI hosts a platform that hosts AI algorithms from developers that can be used to test and refine their applications. Once cleared by the FDA, the algorithms are available to providers to try out the algorithms and incorporate them into their practices.

Nuance has introduced its AI Marketplace to host machine machine learning and deep learning algorithms for image analysis, workflow optimization, clinical decision support, and other radiology applications and use cases. Vendors can host their applications and when subscribed to by providers, receive feedback to improve the applications.

Philips has developed the HealthSuite Insights platform which is available to healthcare providers to develop AI algorithms. The platform will also host 3rd party and Philips AI algorithms for test and further development.

The Siemens Healthineers Digital Ecosystem presents a variety of digital solutions, including AI algorithms in a store where users can purchase and deploy these offerings either installed locally or cloud-based. The Ecosystem includes clinical, operational, and financial solutions in healthcare delivery.

Carestream Vue Clinical Collaboration platform includes Carestream AI developments. One example is the use use of AI to triage exams identifying studies with urgent findings moving these to the top of the worklist. In addition, the platform integrates algorithms from other venders.

Agfa Healthcare has integrated AI into its Enterprise Imaging solution embedding algorithms into clinical workflows. Algorithms include Agfa developments and those from other industry vendors and care providers.

Change Healthcare has been incorporating AI Solutions from other vendors. GE incorporates other vendors algorithms in addition to developing their own. Many of GE’s AI developments have focused on improving the performance of its scanners, e.g., CT, ultrasound, and MRI.

Enterprise Imaging: Not Business as Usual

A common complaint is the difficulty in getting traction for Enterprise Imaging (EI) Systems in health care systems, especially in IT departments. Three issues arise when the topic of EI comes up, usually in the following order. First is that EI is just another project. Estimate the cost and the timeline, and it may get on the approved list. Next would be what is the ROI or why should we do this? The last hurdle may be a stipulation to use the existing imaging resources for EI.

Image from: A White Paper Foundation for Enterprise Imaging: HIMSS-SIIM Collaborative Roth, C.J., Lannum, L.M. & Persons, K.R. J Digit Imaging (2016) 29: 530. https://doi.org/10.1007/s10278-016-9882-0

Enterprise imaging is not another project but an enterprise strategy. EI puts all of the images generated in the health system into an indexed, digital archive. Images can be immediately accessed by all clinicians with relevant information on how, when, and where the image was acquired.

An assessment of image generating departments can illustrate the magnitude of an EI program. Finding 60 plus departments acquiring images employing hundreds of devices is common. EI is not just another radiology or cardiology PACS but is an enterprise strategy and a commitment to 60 plus or more imaging projects over a period of years.

Enterprise imaging is a value proposition to both the image generating departments and to the enterprise as a whole. Storing images with appropriate documentation allows these exams to be billed recovering lost revenue – a benefit to the department and the enterprise.

Having the exams indexed in a central archive reduces cost by eliminating manual searches by departmental staff to retrieve images. These images are also viewable in the EMR by appropriate clinical staff which benefits the enterprise.

Moving images from departmental silos increases security avoiding potential breaches and fines. Cost avoidance benefits the enterprise and the department. The increased revenue, decreased cost, and cost avoidance can be quantified for each department that moves to the EI system.

The HIMSS Analytics Electronic Medical Record Adoption Model (EMRAM) was modified to include patient centric storage of non-DICOM images in Stage 1 on January 1, 2018. The HIMSS Analytics Digital Imaging Adoption Model (DIAM) addresses Enterprise Imaging with an 8 stage model. It is on track to be finalized by the end of 2018.

Most US Hospitals participate in the EMRAM program.  All will need to address EI to continue in the future.

Enterprise imaging is not another PACS project or a series of PACS projects. Nor is EI a radiology project. The work flows and the types of images do not follow the PACS paradigm. Often images are taken at the point of care with no orders with visible light devices that are not DICOM compatible. A separate Enterprise Imaging team and governance structure are needed to manage and implement these projects. Successful EI strategies usually have strong C-Suite support.

The governance council will determine the EI roadmap, set priorities, and criteria for selecting candidates to move to the EI system. The council should include members from IT, operations, cardiology, radiology, and representatives from the other o’logis.

The governance council membership will evolve as the EI strategy progresses and other departments are brought onto the EI system. Initial members from the other o’logies are often departments with the greatest interest and/or greatest needs. Some health systems require a clinician champion from the department prior to starting a project to include it on the EI system.

Understanding the breadth of an EI strategy, the overall value to both the departments and the enterprise, and the importance of governance are crucial to the success of Enterprise Imaging.

Radiology: The FDA and Artificial Intelligence

Out of the dozens of AI companies in radiology, eight have apps cleared to market by the FDA (as of March 18, 2018). Almost all of these with FDA clearance have successfully applied via the FDA 510(k) premarket notification process by identifying a similar device already cleared to market by the FDA, the “predicate device”.

One company has been FDA cleared to market by using the De Novo premarket review pathway. The De Novo process provides a pathway to classify novel medical devices for which controls proved reasonable assurance of safety and effectiveness for the intended use but for which there is no marketed predicate device. De Novo is a risk-based classification process. Those devices that are classified into class 1 or II may be marketed and used as predicates for future premarket notification (FDA 510(k)) submissions. The De Novo classification process was created in 1997 and expanded in 2012.

The FDA has been struggling to figure how to handle AI apps. AI apps, especially those employing deep learning neuro networks, have two major issues for the agency. One problem is the lack of transparency since it cannot be determined how the AI app reached its conclusion. The other issue is that the AI app is dynamic since it keeps learning. The app that was shipped on day 1 to hospital A will not be the same app on day 5. In addition, that app will also be different that the one shipped to hospital B since both are learning on different data.

On July 27, 2017 the FDA published its “Digital Health Innovation Action Plan”. In this plan, the FDA recognized that its traditional approach to moderate hardware-based medical devices is not applicable for the faster iterative design, development, and validation methods employed for software-based technologies. The FDA also thought that traditional implementation of premarket requirements might impede or delay access to evolution of software products.

In the “Digital Health Innovation Action Plan”, the FDA laid out a trial program to “pre-certify” digital health developers who demostrate operational evidence that they excel in software design, development, and testing. Pre-certified developers could qualify to market their lower-risk devices without additional FDA review or with a more streamlined premarket review.

A Pre-cert pilot program was also launched on July 27, 2017 with the Digital Health Innovation Action Plan. On September 26, 2017 the nine companies selected to participate in the Pre-cert pilot program were announced as: Apple, Fitbit, Johnson & Johnson, Pear Therapeutics, Phosphorus, Roche, Samsung, Tidepool, and Verily. Over 100 companies expressed interest in the program. The FDA provides status updates to the pre-cert pilot and has published the Software Precertification Program Model.

The FDA also announced three new guidances, two draft and one final, addressing provisions of the 21st Century Cures Act about where the FDA doesn’t need to be involved and its role where there is a need for FDA involvement. The draft guidance on “Clinical and Patient Decision Support Software” outlines the FDA’s approach to Clinical Decision Support (CDS) which is germane to many of the AI applications in radiology.

The CDS draft guidance is intended to make clear what types of CDS would no longer be defined as a medical device, and thus would not be regulated by the agency. For example, generally, CDS that allows for the provider to independently review the basis for the recommendations are excluded from the FDA’s regulation. This type of CDS can include software that suggests a provider order liver function tests before starting statin medication, consistent with clinical guidelines and approved drug labeling.

However, the FDA will continue to enforce oversight of software programs that are intended to process or analyze medical images, signals from in vitro diagnostic devices or patterns acquired from a processor like an electrocardiogram that use analytical functionalities to make treatment recommendations, as these remain medical devices under the Cures Act. These are areas in which the information provided in the clinical decision software, if not accurate, has the potential for significant patient harm, and the FDA has an important role in ensuring the safety and effectiveness of these products.

Interestingly these FDA documents make no mention of artificial intelligence. Nor do any of the FDA 501(k) summaries of AI applications that have been cleared to market. However, the FDA does address AI in the announcement of the Viz.AI Contact application that was cleared to market under the De Novo process. The Contact application is a CDS package that analyzes CT scans and other information to notify providers of a potential stroke patient.

Another company has entered the De Novo process for an AI application. IDx filed its De Nono application in early February, 2018 for its AI-based system for the autonomous detection of diabetic retinopathy.

The Viz.AI announcement stated that “The FDA is currently creating a regulatory framework for these products that encourages developers to create, adapt and expand the functionalities of their software to aid providers in diagnosing and treating diseases and conditions.” This may be a combination of the CDS Draft Guidance and the Pre-cert program. The FDA has included AI in its definition of Software as a Medical Device (SaMD) which is being addressed by the pre-cert program model. Specifically addressing AI by the FDA is currently a “works in progress”. In the meantime, the De Novo process may become the path for clearance of new AI apps by the FDA.

Artificial Intelligence at RSNA 17: Some Observations

Artificial Intelligence (AI) continued to be a hot topic at the Radiology
Society of North America 2017 annual meeting. It was reported that RSNA 2017 had four times the number of AI sessions as there were at RSNA 2016. Two sessions that I attended were standing room only with attendees turned away prior to the start of the papers.

One of the new introductions at this meeting was the AI imaging distribution platform. EnvoyAI (a TeraRecon company) announced its “EnvoyAI Exchange” where end users can buy access to FDA 510 (k) cleared AI algorithms and where developers can test and refine their products. At RSNA 2017, 3 algorithms on the Exchange were FDA 510(k) cleared and available for purchase. A total of 35 algorithms from 14 developers were on the exchange in various stages of development.

Nuance had a Work In Progress demonstrating its “AI Marketplace” that integrated AI applications from multiple vendors into its PowerShare image sharing platform. The marketplace will host multiple AI applications that can be selected by the user to run on specific exams in PowerShare and have the results auto populate the PowerScribe reports. Commercial availability is yet to be determined.

Siemens is adding AI applications to its “Digital Ecosystem” platform and announced Arterys as a partner in February 2017. Arterys has received FDA 510K clearance for its web-based imaging interpretation platform, MICA that supports interactive AI imaging applications. The Arterys MICA platform currently offers an AI assistant for cardiac MR image analysis that is FDA 510K cleared and has lung and liver analysis solutions pending FDA clearance.

Most PACS vendors were either using 3rd parties for AI application in addition to or in lieu of, their own development. Partnerships announced as of the end of RSNA 2017 were as follows.

As of 12/28/2017, companies in the above list with FDA 510(k) clearance to market for some of their AI applications, are: Arterys, DiA Imaging Analysis(formerly DiACardio), Imbio, and RadLogics.

Presenters cautioned that deep learning neural networks are very large models that are difficult to train. Large amounts of annotated and diverse data sets are required. The models are not transparent meaning that one cannot see how the results are obtained. Validation of these models is nontrivial and will need to be done at multiple sites by clinicians. Overcoming these challenges will take time.

Interventional Cardiology in Transition

Aging baby boomers, new clinical therapies, and evolving regulations are increasing the work load of interventional cardiologists. Do these three factors represent a Perfect Storm for cardiology?

If so, let’s hope that cardiology weathers their storm better than the crew of the Andrea Gail fishing trawler did in the “perfect storm” of 1991 as depicted in the 2000 movie. Instead of a combination of meteorological conditions, cardiology’s brewing storm results from changing demographics, added clinical applications, and evolving regulatory requirements.

As the baby boomers age into retirement they are also entering the peak of cardiovascular disease. Cardiologists are also entering retirement resulting in a shortage of specialists to deal with the increased influx of patients.

The last ten years have seen major advances of less invasive treatments for major types of cardiac disease. Most notably heart valve replacements and ablation therapy for atrial fibrillation are now routinely treated by interventional cardiologists. The new therapies are further increasing the demand on interventional cardiologists and on cardiac catheterization laboratories.

As if the increase in demand driven by the aging population and new interventional therapies was not enough, the healthcare system is in a major sea change. The payment system is starting to shift from fee-per-service to value-based care in 2017. This change will occur over the next few years.

The rules for the new payment systems are evolving resulting in increased demands on providers for additional documentation of clinical decisions and procedures. Providers performing better than average will receive increased Medicare reimbursements and those providers performing below average will see decreased Medicare reimbursements.

Population Increase
AHA CV Disease

Interventional cardiologists are just starting to deal with a larger patient base, more of whom will be treated in the catheterization laboratory, and will be working with an evolving payment system requiring more documentation.

Changing Population Demographics
The baby boomers are not only moving into their retirement years but are also moving into their peak cardiac disease years. The US Census Bureau projects that the 65 and older population is projected to grow from 43.1 million in 2012 to 72.8 million in 2030 – a growth of 70%.

The 2015 American Heart Association’s “Heart Disease and Stroke Disease” statistics shows a remarkable increase in cardiovascular disease in the over 60 population in the US.

The changing demographics of the population is reflected in the interventional cardiologist population as well. The 2014 MedAxiom Survey showed 34% of the interventional cardiology workforce to be over 59 years old with a median age of 54.

Estimates of the shortages of interventional cardiologists vary but the rate of retiring cardiologists has not been balanced by a growth in fellowship positions. Reasons for the projected shortages vary as well. In addition to the changing demographics, other factors listed include the increased demand and the lack of growth in fellowship positions.

All estimates agree that there will be insufficient interventional cardiologists to meet the meet in ten years.

Cardiovascular disease is increasingly being treated in cardiac catheterization laboratories. Minimally invasive techniques have been developed and proven for disease once treated surgically if at all.

Clinical Therapy
For various types of heart disease, the preferred treatment has moved from open heart surgery to minimally invasive techniques, either minimally invasive surgery or delivered via a catheter (transcatheter) similar to angioplasty or stent placements. In come cases, a combined minimally invasive surgical approach and transcatheter therapy provide the best results.

For example for multi-vessel coronary, a minimally invasive surgically technique is most effective on a particular artery (left anterior descending coronary) while stents in other vessels inserted with a catheter by an interventional cardiologist are more effective in the other vessels. Both procedures are performed in a “hybrid” lab, a combination surgical and cardiac catheterization suite.

Similarly ablation therapy used to treat atrial fibrillation may be a combined surgical and catheter based procedure. Each procedure is more effective to access different areas of the heart.

More recently, heart valve replacement surgery is being replaced with a transcatheter procedure by an interventional cardiologist. This started with pulmonary valve replacement for pediatric patients, followed by aortic valve replacement, and now mitral valve replacements for adults. All of these new therapies creat an increased demand on interventional cardiologists.

As the types of interventional procedures have increased so have the reporting requirements. For example the FDA all transcatheter valve procedures be documented and submitted to a CMS (Centers for Medicare & Medicaid Services) approved registry which tracks procedures and outcomes.

The Transcatheter Valve Therapy(TVT) registry is joint collaboration between the American College of Cardiology (STS) and the Society of Thoracic Surgeons. It is the only registry approved by CMS for transcather valve replacment reporting.

In addition to the TVT registry, the ACC maintains nine additional registries for various types of cardiovascular transcatheter interventions. These registries include the CMS mandated ICD Registry for implantable cardioverter defibrillator(ICD) patients and the CMS mandated LAAO Registry for left atria appendage occlusion procedures.

Most facilities elect to participate in the non-mandated transcather registries as well as those mandated by CMS. The other registries are employed for quality control and provide outcomes data to insurance companies.

The healthcare system is moving from the fee-for-service to value-based payment models. Much of this change is driven by the Medicare Access and CHIP Reauthorization Act (MACRA) going into effect starting in 2017.

MACRA’s Quality Payment Program implements two payment options: the MIPS and the APM. The Merit-Based Incentive Payment Systems (MIPS) is a complex pay-for-performance system combining previous programs: the PQRS (Physician Quality Reporting Program), the VBPM (Value Based Payment Modifier, and the MU (Meaningful Use EHR Incentive Program). MIPS also adds an additional measure: Clinical Practice Improvement Activities. Medicare reimbursement gets adjusted based a weighted average of these four components.

The Alternative Payment Models will encompass a variety of shared risk programs such as Accountable Care Organizations (ACOs) and Medical Homes. These models are few to start with and more models are under development.

Most physicians will fall under the MIPS payment plan initially. However, the program plan is to move everyone to a APM in the future.

Payments under MACRA start in 2019 based on data submitted in 2017 and 2018. Supposedly there is more flexibility in the program in 2017 when data submission starts than in the following year. All of the measures under both payment plans must be carefully documented and reported.

Some of the documentation is taken care of by participation in the Clinical Registries, most notably the ACC’s NCDR (National Clinical Data Registry) suite of cardiovascular registries. However, some of the measures require additional documention and reporting.

One of these measures is the Appropriate Use Criteria (AUC) for imaging exams and associated therapies. ACU includes criteria for transcatheter procedures by interventional cardiologists. A deliberate assessment is difficult to make when a patient is coming from the emergency room with a serious cardiac event where “time is muscle” and every minute countsm The ACC AUC definitions are careful to state that a score of “rarely appropriate care” for a angioplasty and stent does not mean that it should not be undertaken in specific cases. However, “exceptions should have documentation of the clinical reasons” for proceeding.

Additional complicating factors are bundled payments where the hospital is paid for an episode of care which includes not just the inpatient stay and associated interventional procedures for a cardiac event but also any related services for 90 days after discharge. The rationale being that higher quality of care and results in fewer post procedure complications.

As most cardiologists are now hospital employees, hospital administration will be watching these events very closely. All of these changes result in increased responsibilities for the interventional cardiologist and a measure of uncertainty as the policies evolve. This environment may also lead to earlier retirement of older cardiologists.