Scholarlink

Celiac disease is a chronic autoimmune disorder triggered by the ingestion of gluten, leading to an excessive immune response that damages the lining of the small intestine. Should it not be diagnosed, it can cause serious health concerns. Until now, physicians used a relatively slow and uncertain process, in which pathology experts study the samples in a microscope. The objective of this study is to create an AI model using deep learning that can accurately classify intestinal biopsy images for identifying celiac disease. 15,000 high-resolution histopathological images were obtained from Baquba General Hospital. All images taken were processed and given variations to support the model’s successful learning. I developed my model using a modified version of ResNet-50 and added Capsule Layers to focus more on feature relationships. Furthermore, Focal Loss was used to handle the problem of class imbalance in the dataset. It performed very well, reaching an accuracy rate of 99.68%, sensitivity of 99.74% and specificity of 99.67%. The results of these models were more accurate and strong than those of VGG-16 and InceptionV3.Results show that deep learning can greatly contribute to making histopathology diagnoses more accurate, dependable and practical. We plan to broaden the dataset by including various clinical cases, adjust the structure of our model using powerful attention mechanisms and fuse clinical and pathological data to help make the AI more suitable for doctors in the field.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Towards Accurate and Efficient Diagnosis of Celiac Disease Based Deep Learning Approach Using Biopsy Image Analysis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Asra Ali Abas,
  • Jamal Mustafa Al-Tuwaijari

摘要

Celiac disease is a chronic autoimmune disorder triggered by the ingestion of gluten, leading to an excessive immune response that damages the lining of the small intestine. Should it not be diagnosed, it can cause serious health concerns. Until now, physicians used a relatively slow and uncertain process, in which pathology experts study the samples in a microscope. The objective of this study is to create an AI model using deep learning that can accurately classify intestinal biopsy images for identifying celiac disease. 15,000 high-resolution histopathological images were obtained from Baquba General Hospital. All images taken were processed and given variations to support the model’s successful learning. I developed my model using a modified version of ResNet-50 and added Capsule Layers to focus more on feature relationships. Furthermore, Focal Loss was used to handle the problem of class imbalance in the dataset. It performed very well, reaching an accuracy rate of 99.68%, sensitivity of 99.74% and specificity of 99.67%. The results of these models were more accurate and strong than those of VGG-16 and InceptionV3.Results show that deep learning can greatly contribute to making histopathology diagnoses more accurate, dependable and practical. We plan to broaden the dataset by including various clinical cases, adjust the structure of our model using powerful attention mechanisms and fuse clinical and pathological data to help make the AI more suitable for doctors in the field.