Lung disease impacts over 500 million people globally, yet access to precise, personalized information for early detection and optimal treatment planning is not readily available. The development of Artificial Intelligence (AI)–driven lung analysis, such as VIDA's LungPrint, helps advance the standard of lung care by uniquely evaluating patients who have or are at risk of lung diseases and several others lung health issues. Using a chest CT scan, functional and other patient data as input, LungPrint provides care teams with a set of precise quantitative insights that can be useful at all points in the care path, from detection through treatment.
The development is done by a team of software engineers, electrical engineers and biomedical engineers. As the fundamental R&D team is made up of engineers by trade and education, it is another innovative way engineering is at the forefront of helping fight this battle against COVID-19. In addition to improving public health and safety in general, lung health is more important that ever during the current world pandemic.
VIDA, a software company within the University of Iowa’s BioVenture Center, introduces a collaborative solution to assist in pulmonary care. An AI-powered CT imaging software and service is developed to aid the early detection, evaluation, and treatment planning of patients with or at risk of lung diseases, including emphysema and airway obstructive diseases (COPD), asthma, interstitial lung disease, and lung cancer.
Engineering.com had the opportunity to speak with Benj Thomas, director of product management at LungPrint, to learn how they developed their AI image analysis software and how they see it being used.
Benj Thomas: We call that our LungPrint analysis, and that LungPrint analysis will uniquely profile a patient with lung disease. So, we are able to quantify and qualify different disease states and to help physicians diagnose and provide the right treatment for patients with lung disease. We are producing software as well as clinical work.
What are the qualities of the images being produced and analysed?
BT: We mainly operate using CT scans, and CT scans vary in image quality. We are able to take a variety of different resolutions of input and produce the same quality of output in terms of the analysis we are doing. So our analysis can adapt to different image qualities, and part of that is leveraging a lot of different technologies from deep learning to kind of traditional image analysis to really be able to extract as much information out of these images as we can.
What were the required breakthroughs to develop this software?
BT: I mean, it’s the same as other softwares: a lot of trial and error and innovation required to come up with the image processing techniques. I will say one of the biggest advantages we had is a curated database of CT images. We have in the order of millions of CT images that have an associated ground truth of diseased state and anatomical boundaries, and so we have really been able to leverage that in breakthroughs in the quality of our services.
How were you able to curate those images of CT scans, how did you have access?
BT: We supported through clinical trials, whether it is pharmaceuticals or device trials. We support through these trials using our LungPrint analysis to show the effectiveness of those new therapies and devices, and so we get access to a lot of those images to work on in our own internal solutions and continue to make our support for those even better.
What will be your next steps to use this technology? Right now, with the COVID-19 coronavirus being so relevant, lung health is at attention. It’s important news, so what would be needed to have this placed in every hospital?
BT: Yes, for sure. We are excited about the current wave of AI being incorporated into clinicals, and we certainly understand it is critical as well, and are looking to make clinicians much more efficient. In the current climate of the coronavirus, and how much of a burden it is on the healthcare system, we really think we are positioned well to help reduce the burden of things like image analysis, and help clinicians be more effective in picking up on the disease more quickly, and focusing their time on the important things on managing patient health.
This software analyses images, and from the analysis a clinician is able to see the lung health and state, and diagnose diseases such as COPD and lung disease, is that correct?
BT: Yes, we take a CTA acquisition which is typically 500 images, and we are able to look at that stack of images as a whole and analyse and uniquely profile the disease state. The structure of the lungs is very complicated, so the airways are branching structures that are hard to navigate and efficiently view. Our technology helps kind of give a global view as well as localize where the disease is and where a radiologist or clinician need to spend more time analysing those images.
What was involved in the engineering of that technology?
BT: We have a lot of great partnerships with doctors and industry professionals. We were able to leverage them to get expertise in what and how we should be analyzing these images. Then we have our database of images that are annotated and we know the disease states, so we are able to apply various different technologies. One of those is obviously supervised machine learning, and so through technology like that we are able to throw in a large number of training images, which enables the neuro-networks to be able to develop high quality algorithms. Then we’re able to accompany more traditional and novel visualization techniques to be able to show the results of these AI algorithms in uniquely compelling ways that really drive efficiency and workflow.
Can you tell us about the LungPrint team?
BT: Absolutely. This development is done by a team of software engineers, so our backgrounds are in electrical engineering, and we have a lot of biomedical engineering degrees, and sort of the fundamental R&D team is made up of engineers by trade and education.
Thomas makes a and significant point: that radiologist and clinicians experience task overload and burnout. Chest CT scans are known to be tedious to traditionally time-consuming to interpret and analyse. However, if an AI-driven automation and visualization is offered as a solution, this may help with task overload, helping to address the issue of physician and healthcare worker overload and burnout.
Visit the LungPrint AI website to learn more.