一个基于深度学习的新型3D卷积神经网络模型,用于视频分析的时空特征映射:胃肠内镜视频分类的可行性研究
Mrinal Kanti Dhar1, Mou Deb2, Poonguzhali Elangovan3
1Department of Radiology, Mayo Clinic, Rochester, MN 55905, USA.
Journal of imaging
|July 25, 2025
概括
这项研究引入了一个深度学习框架来分析医疗视频,特别是对胃肠道内镜片进行分类. 新模型有效地提取了时空特征,在视频分类任务中实现了高精度.
科学领域:
- 医疗成像医学成像
- 人工智能的人工智能
- 计算机视觉 计算机视觉
背景情况:
- 准确的医疗视频分析是具有挑战性的,因为需要有效的时空特征提取.
- 现有的深度学习模型通常专注于静态图像,忽视视频数据至关重要的时间动态.
- 胃肠道 (GI) 内镜视频分类在随着时间的推移捕捉微妙的视觉线索方面提出了特殊的挑战.
研究的目的:
- 提出一种新的深度学习框架,用于从医疗视频序列中提取时空特征.
- 开发和评估一个3D卷积神经网络 (CNN) 用于分类上下肠道内镜视频.
- 引入和整合一个新的残留与并行注意力 (RPA) 块,以增强特征学习.
主要方法:
- 使用hyperKvasir数据集开发了一个3D CNN架构,用于GI内镜视频分类.
- 通过在多个实验运行中选择匹配的视频对来解决数据不平衡.
- 采用 (2+1) D 卷积来降低计算复杂性,并实现了一种新的 RPA 块,将 P-scSE3D 与剩余连接结合起来.
主要成果:
- 拟议的模型实现了高性能指标,包括0.933的平均精度,0.932的精度,0.944的召回率和0.935.935的F1得分.
- 整合了3D平行空间和通道挤压和激发 (P-scSE3D) 模块,F1得分增加了7%.
- 该模型在分类上部和下部肠道内镜视频方面表现出强的性能.
结论:
- 开发的深度学习框架有效地提取了医疗视频分析的时空特征.
- 新的RPA块和 (2+1) D卷积有助于提高准确性和计算效率.
- 这项工作为先进的胃肠道内镜视频分析和未来的临床应用铺平了道路.
相关概念视频
Endoscopic Procedures III: Video Capsule Endoscopy
332
Capsule endoscopy, or wireless or video capsule endoscopy, is a diagnostic procedure for examining the entire gastrointestinal tract. Patients swallow a capsule about the size of a vitamin tablet. The capsule is equipped with a transmitter, a battery, an LED light source, and a color video camera to capture images throughout the gastrointestinal tract. This procedure is particularly useful for diagnosing conditions such as Crohn's disease, ulcerative colitis, tumors, polyps, ulcers,...
332
Imaging Studies III: Gastrointestinal Motility Studies and Virtual Colonoscopy
144
This lesson explores three gastrointestinal imaging techniques: radionuclide testing, colonic transit studies, and virtual colonoscopy.
Radionuclide Testing
Radionuclide testing is a sophisticated medical technique for assessing gastrointestinal motility. It focuses on gastric emptying and colonic transit time. Radioactive markers track the movement of food through the digestive system, providing insights into gastrointestinal disorders.
In gastric emptying studies, a meal's liquid and...
Radionuclide Testing
Radionuclide testing is a sophisticated medical technique for assessing gastrointestinal motility. It focuses on gastric emptying and colonic transit time. Radioactive markers track the movement of food through the digestive system, providing insights into gastrointestinal disorders.
In gastric emptying studies, a meal's liquid and...
144
Endoscopic Procedures IV: Sigmoidoscopy and Laproscopy
177
Sigmoidoscopy and laparoscopy are distinct medical procedures that enable physicians to internally inspect different parts of the GI tract. Although they serve different purposes, each is essential for diagnosing and, in some cases, treating various medical conditions.
Sigmoidoscopy
Sigmoidoscopy is a diagnostic procedure that uses a flexible sigmoidoscope equipped with a light source and camera to examine the rectum and sigmoid colon. The procedure involves inserting the tube through the anus...
Sigmoidoscopy
Sigmoidoscopy is a diagnostic procedure that uses a flexible sigmoidoscope equipped with a light source and camera to examine the rectum and sigmoid colon. The procedure involves inserting the tube through the anus...
177
Endoscopic Procedures II: Colonoscopy
192
The colon, or large intestine, is the final segment of the digestive system. Its primary functions include absorbing water and vitamins produced by gut bacteria and transforming waste from liquid to solid to form stool. In adults, the large intestine is approximately 5 feet long and consists of four main sections:
192
Endoscopic Procedures I: Esophagogastroduodenoscopy
356
An Esophagogastroduodenoscopy (EGD) is a diagnostic procedure in which an endoscopist uses a flexible, lighted endoscope to visualize the upper gastrointestinal (GI) tract. The procedure includes visualizing the oropharynx, esophagus, stomach, and the first part of the small intestine, the duodenum.
During an EGD, the endoscope can be used to:
During an EGD, the endoscope can be used to:
356


