In the Journals

Analysis of microendoscopic images helps diagnose esophageal squamous neoplasia

High-resolution microendoscopy with quantitative image analysis was useful for identifying esophageal squamous neoplasia with high sensitivity and specificity, according to recent study data.

Rebecca R. Richards-Kortum

Using high-resolution microendoscopy, researchers performed an image analysis of 177 patients who underwent standard high-definition white light upper endoscopy with Lugol’s chromoendoscopy (LCE) for screening or surveillance of esophageal squamous neoplasia at two hospitals in China and one in the United States from May 2010 to October 2012. Two expert gastrointestinal pathologists provided a consensus diagnosis to be used as the standard for the 375 biopsy specimens collected from imaged sites.

The researchers used quantitative information from the images to develop an algorithm for identifying high-grade squamous dysplasia or invasive squamous cell cancer. Data from the U.S. centers were divided randomly into a training set to develop and optimize the algorithm and an independent test set to estimate its performance, whereas data obtained from the Chinese center were assigned to an independent validation set.

The mean nuclear area was the best performing feature used for the basis of classification (P < .001). The algorithm had 93% sensitivity and 92% specificity in the training set, 87% sensitivity and 97% specificity in the test set, and 84% sensitivity and 95% specificity in the independent validation set. The area under the receiver operating characteristic curve was 0.92 in the training set, 0.95 in the test set and 0.93 in the independent validation set.

“For comparison, the sensitivity and specificity of LCE were 100% and 51% in the training set, 93% and 55% in the test set, and 100% and 54% in the validation set, respectively; [high-resolution microendoscopic] imaging is associated with substantially improved specificity relative to LCE,” Richards-Kortum said. “This in vivo study demonstrates that such a system can be used to identify neoplasia with high accuracy and may be a valuable tool in low-resource settings where access to trained endoscopists and access to pathologists is limited.” 

“The ability of endoscopists to interpret the [high-resolution microendoscopic] images in real time with the assistance of the image analysis for [esophageal squamous cell neoplasia] is being studied in a prospective, international, multicenter trial,” the researchers wrote. – by Adam Leitenberger

Disclosure: Richards-Kortum reports serving as an unpaid scientific adviser to Remicalm LLC, holds patents related to optical diagnostic technologies that have been licensed to Remicalm, and holds a minority ownership in Remicalm.  

High-resolution microendoscopy with quantitative image analysis was useful for identifying esophageal squamous neoplasia with high sensitivity and specificity, according to recent study data.

“Early detection has the potential to improve survival and quality of life for cancer patients,” Rebecca R. Richards-Kortum, PhD, from the department of bioengineering at Rice University in Houston, told Healio Gastroenterology. “We used a novel, low-cost approach for quantitative image analysis of microendoscopic images to identify neoplastic lesions in patients being screened or surveyed for esophageal squamous cell dysplasia or cancer.” 

Rebecca R. Richards-Kortum

Using high-resolution microendoscopy, researchers performed an image analysis of 177 patients who underwent standard high-definition white light upper endoscopy with Lugol’s chromoendoscopy (LCE) for screening or surveillance of esophageal squamous neoplasia at two hospitals in China and one in the United States from May 2010 to October 2012. Two expert gastrointestinal pathologists provided a consensus diagnosis to be used as the standard for the 375 biopsy specimens collected from imaged sites.

The researchers used quantitative information from the images to develop an algorithm for identifying high-grade squamous dysplasia or invasive squamous cell cancer. Data from the U.S. centers were divided randomly into a training set to develop and optimize the algorithm and an independent test set to estimate its performance, whereas data obtained from the Chinese center were assigned to an independent validation set.

The mean nuclear area was the best performing feature used for the basis of classification (P < .001). The algorithm had 93% sensitivity and 92% specificity in the training set, 87% sensitivity and 97% specificity in the test set, and 84% sensitivity and 95% specificity in the independent validation set. The area under the receiver operating characteristic curve was 0.92 in the training set, 0.95 in the test set and 0.93 in the independent validation set.

“For comparison, the sensitivity and specificity of LCE were 100% and 51% in the training set, 93% and 55% in the test set, and 100% and 54% in the validation set, respectively; [high-resolution microendoscopic] imaging is associated with substantially improved specificity relative to LCE,” Richards-Kortum said. “This in vivo study demonstrates that such a system can be used to identify neoplasia with high accuracy and may be a valuable tool in low-resource settings where access to trained endoscopists and access to pathologists is limited.” 

“The ability of endoscopists to interpret the [high-resolution microendoscopic] images in real time with the assistance of the image analysis for [esophageal squamous cell neoplasia] is being studied in a prospective, international, multicenter trial,” the researchers wrote. – by Adam Leitenberger

Disclosure: Richards-Kortum reports serving as an unpaid scientific adviser to Remicalm LLC, holds patents related to optical diagnostic technologies that have been licensed to Remicalm, and holds a minority ownership in Remicalm.