AI Drives Innovation in Medical Imaging

Artificial intelligence (AI) is being hailed as one of the greatest tools in radiology, with the potential to improve clinician productivity, patient experience, safety and clinical outcomes. This enthusiasm was clearly demonstrated with capital investments in AI-focused imaging startups reaching almost $580 million—more than double the 2017 amount, with over 90 of those solutions cleared for clinical use by at least one global regulatory agency. Health professionals agree that the combination of human clinicians and AI technologies working together delivers significant benefits. Clinicians can provide higher forms of care to patients in a holistic fashion while AI helps streamline and prioritize clinicians’ workflow and assists in timely and accurate diagnosis.

AI was a priority at the Radiological Society of North America (RSNA)’s annual meeting earlier this month, featuring an entire floor densely populated with startups showcasing AI technologies used in chest X-rays, CT scans, and other head and lung imaging solutions. One in particular was Korean company JLK Inspection, which has developed 37 algorithms for inspection of 14 different body parts, many of them deployed on Intel® NUC mini PCs with the Intel® Distribution of OpenVINO™ Toolkit for image recognition. Of note is growing collaboration whereby startups offer their solutions on marketplaces set up by larger, established players. For example, Intel(R) AI Builders program member MaxQ AI’s Accipio intracranial hemorrhage (ICH) and stroke detection platform, optimized with the Intel Distribution of OpenVINO Toolkit to triple the computational power of its platform, is integrating with solutions from other vendors to make it easier for hospitals to adopt and integrate into existing workflows.

Intel AI technologies were on display in a number of other solutions featured at RSNA, including GE Healthcare’s optimized inferencing of an X-ray image within seconds and accelerated magnetic resonance (MR) imaging. Siemens Healthineers speeds up AI for cardiac imaging using 2nd Generation Intel® Xeon® Scalable processors with Intel® Deep Learning Boost and Intel Distribution of OpenVINO toolkit.

Both GE Healthcare and Siemens participated at the Intel SOLVE event at RSNA to present their vision and innovative solutions. As well, a panel of industry experts offered their perspectives on the state of AI adoption in radiology.

Matt DiDonato, GE Healthcare’s Director of Product, shared his vision of elevating radiology through AI technologies. He also noted the importance of AI on the edge in delivering AI solutions where they can directly improve clinical workflows. Among these benefits is speed in set-up, as well as easing the way for MR technologists. GE has calculated that their AIRx technology required 75% fewer clicks, saving time but also reducing the monotony that can come with some of the manual work within imaging.

Figure 1: Matt DiDonato from GE Healthcare at the Intel SOLVE event

Figure 1: Matt DiDonato from GE Healthcare at the Intel SOLVE event.

Puneet Sharma, Senior director for AI at Siemens Healthineers, presented his vision for shaping the future of healthcare, innovations around Digital Twin and their collaboration with Intel using the Intel Distribution of OpenVINO Toolkit for solutions like cardiac MRI segmentation.

Figure 2: Puneet Sharma of Siemens Healthineers at the Intel SOLVE event

Figure 2: Puneet Sharma of Siemens Healthineers at the Intel SOLVE event.

The Intel SOLVE event stage also hosted rich discussion with multiple perspectives from startups, hospitals and radiology centers. Sandeep Akkaraju, Co-Founder and Chief Executive Officer, Exo Imaging; Fabien Beckers, Co-Founder & CEO, Arterys; Wendell Gibby, MD, Director BlueRock Medical & Adjunct Radiology Professor at UCSD; and Howard Berger, MD, RadNet President and CEO graced the stage to share their view of how radiology will continue to evolve.

Figure 3: SOLVE Event Panel

Figure 3: SOLVE Event Panel.

Since current medical imaging scanning measurements are both manual and variable, key benefits of AI-enhanced image acquisition are both a higher image quality and lower dosage of radiation exposure. Howard Berger discussed how shortening scanning time, reducing costs and increasing capacity to address more patients would be key benefits. All of the panelists agreed that clinicians re-engaging with the patient instead of being occupied with manual tasks involved in diagnosis is a key benefit of AI technology used in the field of medical imaging.

Despite the enthusiasm, there are still challenges to address. One of them is the lack of quality structured or annotated data for building AI models. Meanwhile, unstructured data is increasing exponentially, especially as imaging moves from 2D to 3D. Timely transfer of data is another problem, and agencies like the National Institute of Health have assisted by releasing a vast collection of 32,000 CT images. Still, AI developers need access to large representative datasets as AI models developed on local datasets may not work in the US. Federated learning offers the potential to solve the need for large representative datasets, but healthy skepticism exists since there is no standardized way of labeling data based on different protocols. Also, the FDA has not yet developed a way to evaluate a medical technology that continues to learn and improve, such as machine learning and deep learning. Many more issues like the need for human-centered AI, reimbursement methods, and deployment strategies remain, but with the collaboration of clinicians, industry, policy, technology and information, AI will find its way into successful adoption into health and life sciences.

Intel is collaborating with leading health innovators to create new data-driven solutions that will drive transformation in the field of medical imaging enabled by the breadth of the Intel AI portfolio from edge to cloud. Intel envisions a future where insights from all available healthcare data across radiology, pathology, genomics, behavioral and other sources are combined to extract insights and open up new possibilities to predict and prevent disease.