Teleradiology company Nines Inc. won 510(k) clearance from the FDA for Ninesmeasure, an artificial intelligence (AI)-powered tool for measuring lung nodules. The Palo Alto, Calif.-based company says the new tool can help speed the diagnosis of certain respiratory diseases, resulting in better patient outcomes.
“To our knowledge, Ninesmeasure is the only lung nodule measurement tool cleared by the FDA that was developed by a combined team of radiologists and engineers collaborating every day,” said Michael Kelleher, president of Nines Radiology. “This advanced tool can significantly reduce the amount of time our radiologists spend measuring pulmonary nodules, improving to diagnosis for patients without rushing our radiologists.”
Machine learning algorithms
Ninesmeasure is a semi-automatic tool for use by radiologists in the analysis and review of adult thoracic CT images. The device uses a machine learning algorithm to compute segmentations of nodules and calculate long and short axis measurements. According to the 510(k), It provides quantitative information about pulmonary nodule size on a single study or a series of studies by charting long and short axis diameter measurements in the axial plane.
“Lung nodule measurement can be tedious and time-consuming as each nodule has to be measured carefully to determine changes in size over time,” Joel Kronander, Nines’ head of engineering and machine learning, told BioWorld. “Ninesmeasure enables radiologists to quickly measure the long and short axes of selected nodules with a high level of accuracy.”
He added that the tool can be useful in addressing interstudy consistency over the course of a patient’s treatment program.
Nines validated the algorithm’s performance with a retrospective, multicenter image comparison study. The primary endpoint for all nodules was normalized error on long and short axis diameter of 95% confidence level. The results were 0.113 (upper bound 0.124) and 0.131 (upper bound 0.143), respectively. The primary endpoint was also stratified by nodule size. Kronander said the test dataset was diverse and included three different scanner manufacturers, seven scanner models and 11 clinical sites.
The tool operates over a standard network interface and is designed to be used with standard picture archiving and communication systems. It is intended for use only on solid pulmonary nodules.
Second FDA-cleared product
The lung nodule measurement tool is Nine’s second FDA-cleared product. In April 2020, the company got the green light for an AI tool that helps to automate radiological review of CT head images for evidence of intracranial hemorrhage and mass affect. The device, called Ninesai, is intended to assist radiologists in triaging cases.
In addition to FDA clearance of Ninesmeasure, the company said its radiologists are seeing a 40% increase in efficiency over three months due to clinical workflow enhancements, including one-click communications with emergency room doctors and an “always ready” worklist of studies.
“In general, radiology is tech-forward in its use of digital imaging. But innovation can make it better,” said David Stavens, co-founder and CEO of Nines. “Nines has been leading the way by pairing two seemingly disparate groups – skilled radiologists and brilliant engineers – to transform the practice of radiology to be more accessible and more efficient, delivering faster results for quality patient care.”
Stavens co-founded Nines in 2017 with Alexander Kagen, Nines’ chief medical officer ant site chair of diagnostic, molecular and interventional radiology at Mount Sinai West and Mount Sinai Morningside hospitals in New York City. In November 2019, the company raised $16.5 million in a series A financing led Accel and 8VC.
Nines isn’t the only company developing AI products to aid in lung diagnosis. Last week, Oxford, U.K.-based Optellum Ltd. landed FDA clearance for its Virtual Nodule Clinic. The AI-aided image analysis tool helps clinicians evaluate small, potentially malignant lung lesions and nodules.
Also this month, Paris-based Median Technologies SA announced plans to develop its Ibiopsy platform for the early diagnosis of lung cancer in at-risk populations using low-dose CT scans. The company hopes to demonstrate the potential of deep learning to identify lung lesions and characterize them as benign or malignant.
And last fall, New York-based Aidoc Inc. got the go-ahead from the FDA to market the first AI-powered software solution for flagging and triaging incidental pulmonary embolism. The tool, which includes triaging and notification algorithms, is an “always on” technology that analyzes chest CTs in real time and alerts the radiologist of any potentially abnormal findings.