Robin Joffe

The dust is now beginning to settle on the idea of AI in imaging-an idea that once took the industry by storm. The questions being asked by radiologists have now changed from “will it replace me?” to “how can it help me?”-and rightly so. AI continues to make significant progress in the field of diagnostic imaging, as can be gauged using the recently concluded Radiological Society of North America’s (RSNA) Annual Meeting in Chicago as a barometer.

Last year, there were 49 exhibitors at RSNA tagged as machine learning companies, and 22 of those were first time exhibitors. This year, the same number more than doubled to 104, 25 of which were first-time exhibitors. More importantly, the incumbents of medical imaging equipment have also made notable AI efforts, with each one launching new or enhanced AI capabilities. Clearly, AI was one of the key themes at RSNA.

Based on recent analysis, Frost & Sullivan notes that of the 114 startups active in the AI for medical imaging space, a significant majority target the image analysis aspect of medical imaging. Identifying and analyzing specific features in an image form the crux of a radiologists’ job, and since they base their findings on this analysis, it forms the most important clinical step in the imaging workflow. The startup disruption rampant in overall healthcare has focused on image analysis in the case of medical imaging AI.

While computed tomography and magnetic resonance imaging are the top modalities of focus, these are followed closely by digital x-ray, mammography (including 3D tomosynthesis), fundus imaging (of the eye), ultrasound, and echocardiography. While most imaging equipment vendors have not touched upon this area, Siemens Healthineers debuted the AI-Rad Companion Chest CT at RSNA this year, leveraging datasets that add up to 325 million curated and annotated images available for training AI algorithms. The intelligent software assistant can take vendor-agnostic CT images and highlight various thoracic structures, as well as mark any potential abnormalities and include their measurements in structured reports-targeting the heart, aorta, lungs, and coronary arteries. GE Healthcare, on the other hand, launched the Edison platform that brings together all AI algorithms from its existing partnerships.

Beyond image analysis, there are several other steps in the imaging workflow that can also benefit from AI. Right from the ordering stage, all the way down to reporting, AI solutions either exist or can be developed.

Naturally, image analysis is the most advanced application, followed by assignments-AI applications for triaging, worklist assignment, and workflow orchestration are available from Aidoc, Zebra Medical Vision, vRad, and GE Healthcare, for example.

The next most advanced area is of decision support for helping decide the next steps in the clinical pathway-representative examples include Imaging Related Clinical Context from GE healthcare, IntelliSpace Oncology by Philips, and PowerScribe 360 Clinical Guidance from Nuance.

Three Reasons AI is ‘Wildly Different’ than Other Healthcare Technologies

Without going into the details of all other steps, the two workflow steps that could still benefit from the attention of AI solution developers are the ordering and scheduling steps. Companies providing decision support to physicians are best suited to take advantage of such solutions for the ordering stage, while scheduling solutions would be useful for optimizing utilization via scheduling improvements around cancellations and no-shows, for example. There is ample opportunity for growth in this area, using AI to make the workflow more efficient and support the radiologists in ways not previously possible.

Equipment manufacturers are gearing up to make their imaging machines intelligent by infusing AI in their scanners, or at the edge. Canon, for example, debuted the Advanced Intelligent Clear-IQ Engine (AICE) image reconstruction solution that uses AI to deliver high-quality images from lower radiation-dose, low-quality images. Siemens Healthineers already has the FAST 3D camera, which helps with patient isocentering, eliminating the need for a higher dose, improving image quality, and preventing repeat scans which may not be reimbursed.

Beyond these use cases, there are other areas such as self-awareness that allow equipment to “self-diagnose” health of their components, self-driving capabilities to make suggestions or perform additional steps when warranted based on patient condition and so on. These functionalities could be the future of AI with imaging equipment, making the machines truly intelligent.

While on the technology front we continue to make several strides, the value proposition of AI needs to be studied and well-understood for a solution to be adopted in the market. Different stakeholders in the radiology workflow-technicians, radiologists, radiology managers, hospital managers, and even payers-have different requirements that can be addressed by AI. For example, automation of redundant time-intensive tasks can help with increasing productivity, AI-supported image analysis can assist with improving accuracy, and predictability and personalization of patient care pathways can help improve institutional outcomes, and so on.

The development, validation, and deployment of these solutions on a commercial scale requires an ecosystem approach, and several kinds of partnerships and agreements make this a complex ecosystem. From the development of online marketplaces that provide hospitals access to several vendors’ AI solutions on a single platform, to equipment manufacturers partnering with vendors to integrate the solutions with their own are examples that are closest to the end-users. But there are other developments happening in the background as well-technology giants partnering with vendors for cloud-based approaches, computing hardware manufacturers providing solutions for improved computing prowess to equipment manufacturers, and academic institutions continuing to build new AI algorithms and spinning out startups based on those solutions, for example.

The investment horizon for the technology has been promising-Frost & Sullivan estimates that $3.7 billion has been invested globally in the development of this technology so far (including $1.9 billion in startup funding alone, as of September 2018), and we continue to see more AI startups being funded.

However, the promise of AI in the imaging field will have to deliver savings for the end-users: time savings, resource optimization, accuracy gains, and perception gains (bordering on precision health approaches). While the first two refer to the productivity aspect, the latter two cater to quality. It seems that the initial promised savings will be delivered on the productivity front, whereas the quality savings will probably take a long time to catch up-this is also attributed to the lack of appropriate metrics to measure outcomes and resultant quality savings, which is relatively easier to measure from a productivity standpoint.

AI in imaging is not just here to stay-it’s already helping build new equations in the industry. The next frontier will be improving upon the lives of patients and helping radiologists do this in a more efficient way-from image analysis, workflow applications, and later with intelligent medical imaging machines.

Robin Joffe and Siddharth Shah work in transformational health at Frost & Sullivan

Articles

No longer in its infancy, AI is beginning to make real waves in radiology. Here's how. 

Robin Joffe

Articles

How AI is Evolving in Diagnostic Imaging