How an algorithm is taking the guesswork out of lung biopsies

Lung biopsies are quite common. So are mistakes. Can algorithms help?
Written by Greg Nichols, Contributing Writer

When a doctor suspects a patient may have lung lesions or pulmonary nodules, the next step is usually a CT scan. If lesions show up, doctors often recommend biopsies to obtain an accurate diagnosis and determine if the lesions are benign or malignant.

The biopsy procedure is common, albeit intricate and error prone. Imaging can produce false positives, for one thing, resulting in unnecessary intervention. When surgeons perform biopsies, they can accidentally damage border areas of the lungs. It's also common for lesions to be obscured in a scan, resulting in a missed diagnosis.

Better scanning hardware is one solution, albeit an expensive one. An alternative solution is to rely better interpretation of scans using AI-enhanced image processing.

That's the technique used by RSIP Vision, an Israeli AI, computer vision, and image processing technology firm. It's also part of a move toward greater precision in lung surgery using technologies like computer vision and robotics

The company's newly released lung segmentation module uses what are known as segmentation algorithms to divide scanned images into clusters of pixels according to their characteristics. Precise segmentation makes it easier to pinpoint specific points and boundaries in images, which in turn leads to greater accuracy during surgeries. Such precise mapping of small airways enables surgeons to approach lesions through the trachea and perform biopsies with minimal intervention. 

"From the Radiology point of view, this new AI module, is becoming an increasingly helpful tool, especially in detecting small peripheral or very proximal lung lesions that could be of clinical importance and might be missed without it," says Dr Rabeeh Fares, Senior resident in Tel Aviv Medical Center.

The lung segmentation module can be integrated into the software environments of existing scanning technology. Other modules from RSIP are useful orthopedics and cardiology. The AI modules add value to existing medical devices and imaging machines by ensuring greater precision. Not surprisingly, medical device developers are key to RSIP's customer base. 

"We're offering this technology to the medical device industry as an absolutely necessary solution for navigation," RSIP Vision founder and CEO Ron Soferman says. "Today we're seeing  new players in the market using advanced technologies, sometimes with robotic assistance and infrastructure. For them, as for everyone in the industry, having the best possible airway segmentation is a vital element in the success of any procedure." 

In the future, AI segmentation technology could potentially assist with navigation through other interventional procedures.

Editorial standards