Gastric cancer remains one of the main threats to global public health, currently ranking as the fifth most common cancer and the third leading cause of cancer death. Late diagnosis is the biggest obstacle, as identification in advanced stages is associated with an overall survival rate of only 24%. Despite this outlook, evidence shows that early detection can significantly alter the natural history of the disease. The European Society of Gastrointestinal Endoscopy estimates that mortality could be reduced by around 40% through more effective screening and the use of esophagogastroduodenoscopy, a minimally invasive examination capable of identifying precancerous and malignant lesions in the in the esophagus, stomach, and proximal small intestine (duodenum) at early stages with high sensitivity.

Home / Publications / Publication

Home / Publications / Publication

Gastrenterologia

NEURAL NETWORKS THAT SEE BEYOND THE HUMAN EYE: DETECTION OF INTESTINAL METAPLASIA

September 29, 25

Publication type: Article Summary
Original title: Diagnostic Performance of Deep Learning Models for Gastric Intestinal Metaplasia Detection in Narrow-band Images
Article publication date: May 2023
Source: TechRxiv
Authors: Miguel Martins, Maria Pedroso, Diogo Libânio, Mário Dinis-Ribeiro Miguel Coimbra & Francesco Renna

What is the goal, target audience, and areas of digital health it addresses?
     The study aims to evaluate the performance of different deep neural network models in detecting gastric intestinal metaplasia in endoscopy images. It is intended for healthcare professionals, particularly endoscopists, as well as researchers and programmers of artificial intelligence applied to healthcare and decision-makers involved in cancer screening and management policies. The study falls within the areas of digital health, namely artificial intelligence applied to medical diagnosis, endoscopic image processing and analysis, clinical decision support systems, early detection of precancerous conditions and optimization of endoscopic procedures.

What is the context?
     Gastric cancer remains one of the main threats to global public health, currently ranking as the fifth most common cancer and the third leading cause of cancer death. Late diagnosis is the biggest obstacle, as identification in advanced stages is associated with an overall survival rate of only 24%. Despite this outlook, evidence shows that early detection can significantly alter the natural history of the disease. The European Society of Gastrointestinal Endoscopy estimates that mortality could be reduced by around 40% through more effective screening and the use of esophagogastroduodenoscopy, a minimally invasive examination capable of identifying precancerous and malignant lesions in the in the esophagus, stomach, and proximal small intestine (duodenum) at early stages with high sensitivity.

     In this context, gastric intestinal metaplasia is particularly important. This is a transformation of the gastric lining, in which the normal epithelium of the stomach is replaced by intestinal-type epithelium. The presence of gastric intestinal metaplasia increases the risk of developing gastric cancer tenfold, making it a condition of high clinical relevance for surveillance and early intervention. However, its detection remains a challenge, as endoscopic signs are subtle and subject to variability in interpretation, even among experienced endoscopists.

What are the current approaches?
     Currently, the detection of gastric intestinal metaplasia depends on esophagogastroduodenoscopy. With technological advances, techniques such as narrow-band imaging and optical amplification have expanded diagnostic possibilities, allowing mucosal patterns to be highlighted and subtle changes suggestive of metaplasia to be identified. When there is suspicion, confirmation is made through biopsy with histopathological examination, considered the gold standard method. Despite its reliability, this process has significant limitations, as it is invasive, time-consuming, and dependent on the representativeness of the sample collected. Furthermore, even with the support of the most modern imaging techniques, endoscopic interpretation continues to depend heavily on the clinician’s experience, leading to significant variability between observers and compromising diagnostic consistency.

     In recent years, complementary approaches based on artificial intelligence have begun to emerge, particularly through the application of deep neural networks to the analysis of endoscopic images. Models such as EfficientNet have already shown capabilities to distinguish stomach images with and without signs of metaplasia, while VGG-16 has been used to support the diagnosis of metaplasia and gastric atrophy (loss of stomach glands). The ResNet model, when applied with the narrow-band imaging technique, enables a more detailed analysis of the gastric mucosa. Architectures such as DenseNet and Inception have also shown potential in classifying gastric mucosa patterns. These advances point to a future in which clinical practice can benefit from more consistent and objective tools, although important limitations to their implementation remain, such as the need for large volumes of annotated data, the heterogeneity of image sets, and the difficulty in ensuring uniform performance in different clinical contexts.

What does innovation consist of? How is the impact of this study assessed?
     The innovation of this study lay in the direct comparison of five deep neural network architectures — VGG-16, ResNet-50, Inception-v3, DenseNet-121, and EfficientNet-b4 — for the detection of gastric intestinal metaplasia in narrow-band imaging endoscopy images. Unlike previous studies, which analyzed models in isolation or in controlled experimental contexts, this research used data from real clinical practice, collected at the department of gastroenterology of the Instituto Português de Oncologia do Porto. The images were reviewed by an experienced endoscopist and subjected to strict exclusion criteria — such as incorrect diagnoses or low resolution — resulting in a final set of 125 high-quality images, 65 classified as normal and 60 with evidence of gastric intestinal metaplasia. This process ensured that the networks were tested on material representative of clinical reality, increasing the practical relevance of the results.

     The impact assessment was based on a robust experimental methodology, using 5-fold cross-validation: in each cycle, 4 subsets were used to train the model and the remaining one to test, repeating the process until all subsets had been used for testing. The performance of each architecture was measured using metrics: sensitivity (ability to correctly identify images with metaplasia), specificity (ability to correctly recognize images of normal mucosa), positive and negative predictive values (probability that cases identified as positive actually correspond to metaplasia, and that negatives truly correspond to normality), accuracy (overall percentage of correct predictions), and area under the ROC curve, an indicator of overall discriminatory power (ranging from 0 to 1, with 1 representing perfect discrimination and 0.5 representing no discrimination).To reinforce confidence in the results, detailed statistical analyses were performed to compare the models and identify significant errors.

What are the main results? What is the future of this approach?
     The results of this study demonstrated that, despite the small dataset, deep neural networks achieved high performance, with accuracy between 72% (EfficientNet-b4) and 82% (VGG-16). Among the architectures evaluated, VGG-16 and ResNet-50 stood out, presenting areas under the ROC curve greater than 0.80 and greater diagnostic robustness. VGG-16 proved to be more sensitive to the presence of gastric intestinal metaplasia, while ResNet-50 proved to be more specific in identifying normal mucosa, a particularly relevant balance in a clinical context, as higher sensitivity reduces false negatives and higher specificity reduces false positives. In contrast, EfficientNet-b4 and DenseNet-121 showed less consistency and less promising results, confirming limitations in their applicability in different clinical settings. Statistical analysis confirmed significant differences between the best and worst performing models, while error analysis revealed inter-fold variability and difficulties in adapting to different image scales: the models were able to interpret situations with endoscopic artifacts, such as reflections or lighting variations, but failed in seemingly simple cases, such as amplified images. In clinical terms, VGG-16 appears to be the most promising architecture, as its higher sensitivity is particularly relevant for  patient follow-up, while simultaneously ensuring high accuracy.

     The future of this approach involves validation in broader and more diverse data sets, including multicentre studies, and integration into real-time diagnostic support systems capable of complementing clinical decision-making during examinations. In addition, the development of more adaptive models could overcome current technical limitations, enhancing the reliability and consistency of these tools. In terms of public health, this line of research could transform cancer screening programs, making the detection of gastric intestinal metaplasia earlier, contributing to reducing mortality associated with gastric cancer.

Would you like to know all the details?

Know more here

Do you have an innovative idea in healthcare field?

Share it with us and see it come to life.
We will help bring your projects to life!

Share my idea

Newsletter

Receive the latest updates from the InovarSaúde portal.

Subscribe

Support

FAQs

Contacts

User Manual

República Portuguesa logo
logotipo SNS
SPMS logotipo

Follow Us

YouTube
LinkedIn

Co-funded by

PRR Logotipo
república Portuguesa logo
União Europeia Logo

Newsletter

Receive the latest updates from the InovarSaúde portal.

Subscribe

Support

FAQs

Contacts

User Manual

República Portuguesa logo
SNS Logo
SPMS Logo

Follow Us

Co-funded by

PRR Logotipo
República Portuguesa logo
União Europeia Logo

Home / Publications / Publication

Gastrenterologia

NEURAL NETWORKS THAT SEE BEYOND THE HUMAN EYE: DETECTION OF INTESTINAL METAPLASIA

September 29, 25

Publication type: Article Summary
Original title: Diagnostic Performance of Deep Learning Models for Gastric Intestinal Metaplasia Detection in Narrow-band Images
Article publication date: May 2023
Source: TechRxiv
Authors: Miguel Martins, Maria Pedroso, Diogo Libânio, Mário Dinis-Ribeiro Miguel Coimbra & Francesco Renna

What is the goal, target audience, and areas of digital health it addresses?
     The study aims to evaluate the performance of different deep neural network models in detecting gastric intestinal metaplasia in endoscopy images. It is intended for healthcare professionals, particularly endoscopists, as well as researchers and programmers of artificial intelligence applied to healthcare and decision-makers involved in cancer screening and management policies. The study falls within the areas of digital health, namely artificial intelligence applied to medical diagnosis, endoscopic image processing and analysis, clinical decision support systems, early detection of precancerous conditions and optimization of endoscopic procedures.

What is the context?
     Gastric cancer remains one of the main threats to global public health, currently ranking as the fifth most common cancer and the third leading cause of cancer death. Late diagnosis is the biggest obstacle, as identification in advanced stages is associated with an overall survival rate of only 24%. Despite this outlook, evidence shows that early detection can significantly alter the natural history of the disease. The European Society of Gastrointestinal Endoscopy estimates that mortality could be reduced by around 40% through more effective screening and the use of esophagogastroduodenoscopy, a minimally invasive examination capable of identifying precancerous and malignant lesions in the in the esophagus, stomach, and proximal small intestine (duodenum) at early stages with high sensitivity.

     In this context, gastric intestinal metaplasia is particularly important. This is a transformation of the gastric lining, in which the normal epithelium of the stomach is replaced by intestinal-type epithelium. The presence of gastric intestinal metaplasia increases the risk of developing gastric cancer tenfold, making it a condition of high clinical relevance for surveillance and early intervention. However, its detection remains a challenge, as endoscopic signs are subtle and subject to variability in interpretation, even among experienced endoscopists.

What are the current approaches?
     Currently, the detection of gastric intestinal metaplasia depends on esophagogastroduodenoscopy. With technological advances, techniques such as narrow-band imaging and optical amplification have expanded diagnostic possibilities, allowing mucosal patterns to be highlighted and subtle changes suggestive of metaplasia to be identified. When there is suspicion, confirmation is made through biopsy with histopathological examination, considered the gold standard method. Despite its reliability, this process has significant limitations, as it is invasive, time-consuming, and dependent on the representativeness of the sample collected. Furthermore, even with the support of the most modern imaging techniques, endoscopic interpretation continues to depend heavily on the clinician’s experience, leading to significant variability between observers and compromising diagnostic consistency.

     In recent years, complementary approaches based on artificial intelligence have begun to emerge, particularly through the application of deep neural networks to the analysis of endoscopic images. Models such as EfficientNet have already shown capabilities to distinguish stomach images with and without signs of metaplasia, while VGG-16 has been used to support the diagnosis of metaplasia and gastric atrophy (loss of stomach glands). The ResNet model, when applied with the narrow-band imaging technique, enables a more detailed analysis of the gastric mucosa. Architectures such as DenseNet and Inception have also shown potential in classifying gastric mucosa patterns. These advances point to a future in which clinical practice can benefit from more consistent and objective tools, although important limitations to their implementation remain, such as the need for large volumes of annotated data, the heterogeneity of image sets, and the difficulty in ensuring uniform performance in different clinical contexts.

What does innovation consist of? How is the impact of this study assessed?
     The innovation of this study lay in the direct comparison of five deep neural network architectures — VGG-16, ResNet-50, Inception-v3, DenseNet-121, and EfficientNet-b4 — for the detection of gastric intestinal metaplasia in narrow-band imaging endoscopy images. Unlike previous studies, which analyzed models in isolation or in controlled experimental contexts, this research used data from real clinical practice, collected at the department of gastroenterology of the Instituto Português de Oncologia do Porto. The images were reviewed by an experienced endoscopist and subjected to strict exclusion criteria — such as incorrect diagnoses or low resolution — resulting in a final set of 125 high-quality images, 65 classified as normal and 60 with evidence of gastric intestinal metaplasia. This process ensured that the networks were tested on material representative of clinical reality, increasing the practical relevance of the results.

     The impact assessment was based on a robust experimental methodology, using 5-fold cross-validation: in each cycle, 4 subsets were used to train the model and the remaining one to test, repeating the process until all subsets had been used for testing. The performance of each architecture was measured using metrics: sensitivity (ability to correctly identify images with metaplasia), specificity (ability to correctly recognize images of normal mucosa), positive and negative predictive values (probability that cases identified as positive actually correspond to metaplasia, and that negatives truly correspond to normality), accuracy (overall percentage of correct predictions), and area under the ROC curve, an indicator of overall discriminatory power (ranging from 0 to 1, with 1 representing perfect discrimination and 0.5 representing no discrimination).To reinforce confidence in the results, detailed statistical analyses were performed to compare the models and identify significant errors.

What are the main results? What is the future of this approach?
     The results of this study demonstrated that, despite the small dataset, deep neural networks achieved high performance, with accuracy between 72% (EfficientNet-b4) and 82% (VGG-16). Among the architectures evaluated, VGG-16 and ResNet-50 stood out, presenting areas under the ROC curve greater than 0.80 and greater diagnostic robustness. VGG-16 proved to be more sensitive to the presence of gastric intestinal metaplasia, while ResNet-50 proved to be more specific in identifying normal mucosa, a particularly relevant balance in a clinical context, as higher sensitivity reduces false negatives and higher specificity reduces false positives. In contrast, EfficientNet-b4 and DenseNet-121 showed less consistency and less promising results, confirming limitations in their applicability in different clinical settings. Statistical analysis confirmed significant differences between the best and worst performing models, while error analysis revealed inter-fold variability and difficulties in adapting to different image scales: the models were able to interpret situations with endoscopic artifacts, such as reflections or lighting variations, but failed in seemingly simple cases, such as amplified images. In clinical terms, VGG-16 appears to be the most promising architecture, as its higher sensitivity is particularly relevant for  patient follow-up, while simultaneously ensuring high accuracy.

     The future of this approach involves validation in broader and more diverse data sets, including multicentre studies, and integration into real-time diagnostic support systems capable of complementing clinical decision-making during examinations. In addition, the development of more adaptive models could overcome current technical limitations, enhancing the reliability and consistency of these tools. In terms of public health, this line of research could transform cancer screening programs, making the detection of gastric intestinal metaplasia earlier, contributing to reducing mortality associated with gastric cancer.

Would you like to know all the details?

Know more here
Cadeira de rodas

Autonomous Patient Mobility in a Hospital Environment

The internal transport of patients in healthcare institutions, although at first glance it may seem like a simple task, represents a complex, continuous, demanding and time-consuming logistical operation that cuts across all levels of the…

Read more
Voz em IA

The Future of Diagnostics: Speech and AI

Speech is a biomarker that reflects, in a sensitive way, the integrated functioning of several physiological systems, namely the nervous, respiratory, and muscular systems. This complexity makes it a promising resource for detecting changes associated…

Read more
Literatura sobre os cuidados de saúde no futuro

What Literature Reveals About Healthcare in the Future

The healthcare sector is undergoing rapid transformation driven by population aging, increasing complexity of care, and digital advancements, in a context that requires greater integration, sustainability and adaptation to new realities such as the European…

Read more
Perturbação do sono

A Digital Intervention for Insomnia in Oncology

Insomnia is a sleep disorder characterised by persistent difficulties in initiating sleep, maintaining sleep during the night, or achieving restful sleep. These difficulties arise even in the presence of adequate sleeping conditions and are often…

Read more
Sistema de telemonitorização remota

Digital Technology Revolutionising Post-cardiac Surgery

According to the World Health Organisation, cardiovascular disease remains the leading cause of death worldwide, responsible for around 17.9 million deaths a year. Its high prevalence is associated with unhealthy lifestyles characterised by poor diet,…

Read more
Sistema robótico autónomo INSIDE

Autonomous Robotics System for Autism Therapy

Autism spectrum disorder is a neurodevelopmental condition with significant clinical, social and economic repercussions throughout life. According to the World Health Organization, it is estimated to affect approximately 1 in 160 children worldwide. Its origin…

Read more
Enfermeira com um telefone

Mobile Application to Improve Workflows in Nursing Homes

Portugal has one of the highest aging populations in the world, placing increasing pressure on elderly care services, especially in nursing homes. Healthcare professionals in these facilities are often overwhelmed due to the increasing number…

Read more
troca de informações de saúde e interoperabilidade

New Era of Interoperability in Healthcare Systems

The growing use of electronic health records, digital diagnostic systems and remote monitoring technologies has led to a significant increase in the volume and complexity of health data. This increase intensifies the need for continuous,…

Read more
robótica colaborativa

Collaborative Robotics Improves Working Conditions

Workers face growing challenges in the industrial environment. Among the most critical are fatigue and inappropriate postures, often associated with repetitive tasks and working conditions that lack ergonomic suitability. These factors represent significant risks for…

Read more
Benefícios da Eletrônica Médica

Detection of Anxiety and Panic Attacks in Real Time

The growing number of people with anxiety disorders, along with increased awareness of mental health, drives the need for new technological tools that provide remote and continuous monitoring of anxiety and panic disorders. Thus, the…

Read more
tele-ecografia

A Novel Approach for Robotic-assisted Tele-echography

Currently, robotic systems for ultrasound diagnostic procedures fall into two main categories: portable robots that require manual positioning and fully autonomous robotic systems that independently control the ultrasound probe’s orientation and positioning. Portable robots rely…

Read more
Personalização e tecnologia na gestão da Diabetes

Personalization and Technology in Diabetes Management

IPDM has significant potential to improve diabetes management and drive health system reforms to become high-performing, effective, equitable, accessible, and sustainable. Evidence and good practices inspire health system transformation. Adopting person-centred approaches like co-creation and…

Read more
TEF-HEALTH Logo

SPMS Integrates the TEF-Health Initiative

SPMS participates in the TEF-Health initiative as a partner in a consortium composed of 51 entities from 9 European Union countries. This action is co-financed by the European Commission and has a duration of five…

Read more
Global Digital Health Partnership Logo

SPMS Represents Portugal as Vice-president of GDHP

The GDHP is an intergovernmental organization in the digital health sector that facilitates cooperation and collaboration between government representatives and the World Health Organization (WHO). Its purpose is to foster policymaking that promote the digitalization…

Read more
Portugal INCoDe.2030

Digital Transformation of Health at INCoDe.2030 in Tomar

The “National Digital Skills Initiative e.2030, Portugal” (INCoDe.2030) is an initiative that aims to improve the Portuguese population’s level of digital skills, placing Portugal at the level of the most advanced European countries in this…

Read more
HealthData@PT Logo

HealthData@PT: New SPMS Initiative for Health Data

Action HealthData@PT is launched in the context of the implementation of the European Health Data Space, and is an initiative approved by the European Commission under the EU4Health 2021-2027 programme. This initiative contributes to the…

Read more

Do you have an innovative idea in healthcare field?

Share it with us and see it come to life.
We will help bring your projects to life!

Share my idea

Newsletter

Receive the latest updates from the Inovarsaúde portal.

Subscribe

Support

FAQs

Contacts

User Manual

República Portuguesa logo
SNS Logo
SPMS Logo

Follow Us

YouTube
LinkedIn

Co-funded by

PRR Logotipo
República Portuguesa logo
União Europeia Logo
Scroll to Top