The COVID-19 crisis has not only exacerbated many of the existing challenges in healthcare delivery, it has also increased the already taxing demands on healthcare providers and medical staff. Limited resources and crowded emergency departments and trauma centers across the globe have highlighted the importance of quickly triaging patients toward the right treatment path. It is here, at the point of care, that imaging plays such a vital role.
How AI tools are enhancing the imaging workflow
Improving the radiology workflow with embedded artificial intelligence (AI) tools in the radiology platform has alleviated some of the increasing time pressures under which technologists and radiologists operate. Nearly 60 percent[i] of today’s cases are sent as urgent or STAT, but delays in diagnosis and treatment can have a significant impact on patient care. While each case still is manually reviewed by a radiologist, new AI solutions are helping draw a finer separation between STAT and critical patients to allow timely reporting and more accurate patient management.
The growing adoption of embedded AI applications within radiology systems is impacting imaging and care delivery in the fight against the pandemic as well as in other clinical areas such as neuroimaging, breast imaging, and cardiothoracic imaging. Using a platform approach, many AI tools can be integrated into the radiology workflow without adding more work and additional training.
Using AI tools helps to automate some of the technologist’s manual tasks and helps provide reading physicians with clinical decision support. To support reading and help prioritize critical cases, AI tools can help to detect subtle or complex patterns within patient images, resulting in improvements across the board in efficiency, quality and clinical accuracy.
AI for cardiothoracic imaging in COVID-19 cases
Cardiothoracic imaging is one example where an embedded AI application has been operating successfully to quickly help identify patients with pneumothorax, or collapsed lung, where air leaks into the space between the lung and the chest wall. The COVID-19 virus is a disease that mainly affects the lungs. Point of care x-ray has played an important role in the initial evaluation and follow up of COVID-19 patients, many of whom may have comorbidities such as Chronic obstructive pulmonary disease (COPD), and can develop pneumonia accompanied by pneumothorax.
Amit Gupta, MD, a cardiothoracic radiologist at University Hospitals in Cleveland, has been using an AI application on his patients for detection of pneumothorax on x-rays, as well as for lung nodule detection on CT scans, and for automatic segmentation of lung cancer patients. He also uses it for the prognostication and prediction of patients’ response to chemotherapy. Dr. Gupta believes that streamlining AI applications into routine practice can only be successful if it is easily implemented into the radiology platforms that the radiologists are already using, facilitating its adoption and use.
Using an embedded AI application that also requires little additional training for technologists helps to improve the complicated and manual imaging process of imaging patients prior to, and during the imaging exam. With no additional delay or processing time, the AI algorithm provides triage notifications to the radiologist that arrive in PACS at the exact same time as the DICOM image.
AI benefits at the point of care for technologists
At the point of care, the AI automatically orients each image properly, so the technologist does not have to do that extra work before sending the image to the PACS, which can save nearly 20 hours of work over a year[ii]. The AI algorithm also quickly reviews the image taken against the imaging protocol that was ordered to ensure the correct image was taken, and calls attention to any image imperfections, such as a clipped lung, which is a common image error. The technologist is able to correct these errors while the patient is still in the room. These solutions, though minor, save a great deal of time for the technologist, and may reduce the need for retakes by getting the right image for each patient at the point of care.
AI technology for neurological imaging and stroke care
Another area benefitting from AI- embedded technology is neuro imaging and stroke care. At Capital Health Hospitals (Capital Health) in Trenton, New Jersey, Ajay Choudhri, MD, Chair and Medical Director of Radiology implemented AI in the neuroradiology workflow to help triage critical stroke patients. With a large neuroscience program and clinical expertise in neuroradiology, Capital Health provides this expertise to its surrounding areas and takes in many transfer patients from other facilities. When patients are brought in with a stroke alert, Capital Health protocol dictates the patient’s CT must be read within 7-minutes.
To help with accurate and fast triage, the AI tool is integrated into the PACS system and accesses inbound patients’ CT scans to help identify patients who may have a hemorrhage and move them to the top of the reading list. This has proven to be very helpful when there are multiple scans to be evaluated at one time, and the most critical can be prioritized. Anticipated future iterations of the AI tool will include the ability to identify the exact location and severity of the hemorrhage as well.
While the radiology team at Capital Health has the clinical expertise to routinely identify hemorrhages in stroke patients, even without the benefit of the AI, Dr. Chaudry believes that this AI application supports the fast and accurate triage of stroke patients at Capital Health and it also could be a significant value to facilities that don’t have a large neuroscience program, or do not have a specialized neuroradiology staff to evaluate these types of critical images.
Using embedded AI tools for breast imaging and ultrasound
Breast imaging is yet another area benefitting from the addition of AI technology. While the widespread adoption of 3D Tomosynthesis has made a significant impact on the detection of breast cancers in women, it has also increased the number of images per mammogram that need to be reviewed by the radiologist. Expert AI technology can assist reading radiologists by looking for subtle or complex patterns within breast images and highlighting possible cancers.
Randy Hicks, MD, CEO of Regional Medical Imaging (RMI), simultaneously upgraded his nine imaging centers located across Michigan to offer 3D Mammography to all the patients they serve. The independent radiology group also utilizes an embedded AI tool that has been trained with millions of images to look for architectural distortions and cancers in the breast images. Dr. Hicks reports they have seen a significant reduction in the need for retakes, partly because of the improved level of confidence the reading radiologists feel they have because of the assistance from the AI tool. They’ve also seen a reduction in the number of biopsies ordered for the patients.
AI applications are also being used in breast ultrasound, where Amy Patel, MD and her radiology team at Liberty Health in the Midwest have implemented an AI tool to optimize workflow efficiency and improve diagnostic confidence. Dr. Patel reports they are seeing a reduction in unnecessary biopsies, as well as a reduction in the number of BI-RADS 3 findings, which require patients to follow up for additional exams more frequently.
Across both specialties and modalities, the availability and adoption of AI applications is quickly growing, and impacts are evident in workflow improvements, as well as improvements in triage, diagnostics and patient management. AI based tools are becoming a necessity to alleviate the time demands on radiology workflows, as well as provide clinical decision support to reading physicians. Taking the time to develop a strategy and choose an approach or platform that will ease the implementation of existing and newly developed AI into a facility is crucial to their success.
Learn how GE Healthcare is working with leading academic organizations and third-party AI application developers to integrate their assistive technologies into the workflow that radiologists are already familiar and comfortable with, enable radiologists to spend less time performing manual tasks, and leave more time interacting with critical members of the care team as well as patients, access this webinar:
How to Implement AI in Imaging Workflows; A platform approach to seamless integration
[i] World Health Organization Report—Communicating Radiation Risks in Pediatric Imaging.
[ii] 1Younis, K., et. al. (2019). Leveraging Deep Learning Artificial Intelligence in Detecting the Orientation of Chest X-ray Images. SIIM Conference on Machine Intelligence in Medical Imaging (C-MIMI), Oral Presentation.
