Medical Technology

AI System Improves Early Gastric Cancer Detection

An artificial intelligence (AI) system called “ENDOANGEL” was effective for real-time monitoring of endoscopic “blind spots” and improved detection of early gastric cancer (EGC) during esophagogastroduodenoscopy (EGD), according to research published recently.

While EGD is widely utilized to look for lesions in the upper gastrointestinal tract, there is significant disparity among endoscopists about the results and results in a high failure rate for EGC. In a study published in the journal Endoscopy researchers suggest a more objective assessment of lesions using AI technology could improve the detection rate in real time and increase the likelihood of establishing a diagnosis earlier and prompt treatment for gastric cancer.

The researchers updated WISENSE the AI system that had previously been created. It was previously capable of monitoring gastric regions that were not covered by EGD (termed “blind spot”). The WISENSE system was upgraded by the investigators to incorporate a real-time EGC detection algorithm.

Researchers from the Renmin Hospital of Wuhan (China) University used deep convolutional neural networks and deep reinforcement learning to create the ENDOANGEL. One hundred fifty patients from five hospitals in China who were receiving EGD were randomly assigned to an ENDOANGEL-assisted protocol (n = 498) or a control group (n = 504) which did not utilize the ENDOANGEL system. Examining included white-light imaging and magnifying image-enhanced Otoscopy observation. The biopsy was used to determine suspicious lesions.

The researchers compared the groups in terms of the amount of blind spots following the intervention. They evaluated the effectiveness of the AI-based ENDOANGEL system in its ability to predict EGC in a clinical environment.

Patients in ENDOANGEL had significantly smaller blind spots than those in the control group (5.38 vs. 982 respectively; P .001). Despite this that patients in the ENDOANGEL group had significantly more inspection time (5.40 minutes vs. 4.38 minutes; P < .001).

There were 819 cases of lesions reported by endoscopists from the ENDOANGEL group, which included 196 gastric lesions that had pathological results. According to the investigators they found that the ENDOANGEL system accurately predicted all three EGCs, which included one mucosal carcinoma and two neoplasias of high grade, and two advanced gastric cancers. The per-lesion accuracy was 84.7 percent, and the sensitivity and specificity rates for detecting cancer of the gastric were 100 percent and 84.3%, respectively.

The authors highlighted the limitations in the analysis as well as the absence of follow-up. They also pointed out the possibility of biases attributed to unblinded statisticians. More research is needed, they wrote.

“In conclusion, ENDOANGEL, a system for improving endoscopy quality built on deep learning has enabled real-time monitoring of endoscopic blind spots, timing, and EGC detection during EGD,” according to the authors. “ENDOANGEL significantly improved the quality of EGD in this multicenter study and also showed the potential for detecting EGC in real clinical settings.”

Spotting the Blind Spots

A system based on AI such as ENDOANGEL could overcome the weaknesses inherent in conventional diagnostic testing, thereby increasing rates of EGC testing. David Hoffman, MD, a gastroenterology doctor who is also the medical director at Cedars-Sinai Cancer Beverly Hills, said this in an interview. “I think that there are ethical issues that we’re going to have to grapple with respect to accessibility and data mining, and what that means in the end,” said Hoffman, who wasn’t involved in the study. “But I believe that the use of AI with machine-learning and deep learning is a huge opportunity for cancer and public health and is optimistic.”

Hoffman said that AI systems could provide additional benefits beyond the detection of cancer early on and, in particular, aiding in the making of personalized medical decisions and real-time surgical interventions. “Using AI with machine learning and deep learning has a huge potential and I think it sort of is an obvious extension of what we’re seeing in … algorithmic approaches to make use of big data, therefore it’s an evolution that is natural in the field of medical applications,” Hoffman explained.

Anuj Patel MD, a medical-oncology specialist at Dana-Farber Cancer Institute in Boston, was not involved in the research. He explained that any strategy that can detect gastric cancer earlier could have a visible global effect. “We have a lower incidence of gastric cancer in the United States, but the training and frequency of early gastric cancers observed by different doctors can differ,” Patel said in an interview. “AI systems could provide an additional layer of analysis during procedures where an endoscopist could overlook the subtle characteristics that are associated with early gastric cancers.”

Patel said that while the ENDOANGEL study’s results are encouraging, the longer-term question is whether the method will yield significant improvements in the patient’s outcomes. The next step is to determine the effectiveness of these techniques utilized in various populations. “AI models such as these must be trained. It is important to determine if they require retrained in countries where gastric cancer is seen differently or with different endoscopic equipment and techniques.

Hoffman and Patel, the study authors, did not have conflicts of interest that they could disclose.

This article was originally published on It is part of the Medscape Professional Network.

Content Source:

The Medical Progress

The Medical Progress is here to provide news for the medical industry on a daily basis which are mainly on the medical cannabis and it’s legalisation.

Related Articles