Best Lung Nodule Detecting AI Read more
Back to List

Lunit INSIGHT: Q&A with Brandon B. Suh, Chief Medical Officer at Lunit

Nov 28, 2017 - Lunit Media

Lunit INSIGHT: Q&A with Brandon B. Suh, Chief Medical Officer at Lunit


Brandon B. Suh, Chief Medical Officer of Lunit answers the most frequently asked questions about Lunit INSIGHT.






·      What is Lunit INSIGHT?

Lunit INSIGHT is a new and advanced cloud-based artificial intelligence solution for real-time medical image analysis. During RSNA 2017, we will be live-demonstrating the software which was developed with our own cutting-edge deep learning technology. Lunit INSIGHT is currently available to the public at http://insight.lunit.io, free of charge.





·      How is it operated?

Users can upload their medical images and receive AI analysis results in just a few seconds. The analysis results include not only the level of abnormality shown in terms of a “confidence score”, but also the visualization of the AI’s attention map. In addition, other functions commonly found in PACS viewers, such as contrast level adjustment, zoom in/out, etc, are all available as well so that users can take a good look at the images uploaded.





·      How accurate is Lunit INSIGHT?

Lunit’s chest x-ray solution detects major chest abnormalities, lung nodule/mass, consolidation, and pneumothorax, with an unprecedented high level of accuracy 一 97% standalone accuracy in nodule detection, 99% for consolidation and pneumothorax. According to the National Lung Screening Trial (NLST), 26.5% of lung cancer was shown to be missed by chest x-ray. Noting the common use of chest x-ray throughout general medical practice with more than 1 billion chest x-ray exams performed every year worldwide, decreasing the proportion of missed cases even by 10% would translate into a significant clinical benefit.


What’s foremost significant is how our solutions have been proven to significantly increase the diagnostic performance of its users up to 20%, from non-radiology physicians to thoracic radiology experts. Such increase in performance level has been, interestingly, shown to be dependent on the level of trust a user has on the results of our solutions, especially for equivocal lesions or lesions that are hidden and difficult to be visualized by extrapulmonary structures, such as the clavicle, hilar structure, heart, or the diaphragm.





·      Is it only available for x-ray solutions?

Other than the chest x-ray solution, Lunit’s mammography solution to detect suspicious breast cancer lesions is in its final stages of development. Lunit INSIGHT for Mammography is expected to be publically released by the first quarter of 2018. Lunit is also doing research in developing solutions for digital breast tomosynthesis, chest CT, and coronary CT angiography.





·      How do you train Lunit’s Artificial Intelligence?

Lunit’s AIs are trained by a huge collection of de-identified clinical data from Lunit’s partner hospitals, 18 in a total number of partnerships. The total number of images that have been directly used in our research has reached over 1 million case images, of which around 200k are chest x-ray images and another 200k are mammography images. Moreover, most of the images collected are curated with proper ground truth labels including follow-up data, clinical data, as well as pathology data, which significantly help Lunit’s AI technology to be translated into high performing algorithms. Based on the given image data, the AI algorithms are then specifically trained to detect target diseases or radiologic findings, including lung cancer, tuberculosis, pneumonia, pneumothorax, and breast cancer for chest x-ray and mammograms, for example.






·      How do you conduct proper clinical validation?

Proper clinical validation is an important part of presenting meaningful AI solutions that have a high clinical impact. We put a lot of efforts to conduct large-scale clinical studies, not just in Korea where Lunit is based, but also in the United States and elsewhere. Large-scale multi-center reader studies are set to be conducted in early 2018 with multiple leading hospitals in Korea and the US for Lunit’s chest x-ray and mammography solutions, for which publication of the results are targeted for late 2018. FDA approval for Lunit’s chest x-ray and mammography solutions are expected to be achieved by end of 2018.





·      What do you expect from presenting Lunit INSIGHT at RSNA 2017?

We would like to connect with more people who are interested in integrating our solutions into their systems to get valuable feedback and promote our solutions. Testing our solutions for pre-market evaluation is an important endeavor for Lunit, because testing in actual routine clinical settings would be an important step to double-check for any unexpected problems and eventually earn the trust of end users, not only in terms of the accuracy of the solutions but also in terms of usability. While early adoption will be based on a small number of reference sites, we developed multiple ways that can make Lunit INSIGHT easily accommodated, including both cloud-based systems and library packaging systems that can be run locally without the need for images to leave the hospital premises.


The public release of Lunit INSIGHT is also expected to help us establish partnerships with vendors and platform companies that will eventually help distribute Lunit’s products, especially once the products are regulatory approved. We are hoping that after users witness the high level of our solutions themselves, it would drive the demand for adoption of our solutions in their affiliated institutions.










Brandon B. Suh, Chief Medical Officer, Lunit


Lunit INSIGHT




Lunit members at RSNA 2017



https://insight.lunit.io/

ConferenceCorporateLunit INSIGHTProductRadiology

More from News

No Data