基于内窥镜的声门识别数据

医务人员在插管时，需要内窥镜的辅助，用于观察到声门位置，可顺利插管。经验不足的实习医生在插管时无法准确判断声门，容易造成患者受伤。本数据通过采集志愿者的声门图像，进行深度学习分析，可以实现摄像自动捕捉判断声门位置，以助于经验不足的医务人员插管。本数据和算法仅针对人体内的声门图像画面显示，不含个人数据、公共数据，无数据标识体现。本算法是在喉镜工作环境下，对声门进行识别和判断。结合非极大值抑制算法对同一个图像中存在不同器官进行过滤，计算出被识别器官占图像的像素比例（即被识别器官所占像素面积除以图像总面积），标签坐标框取为声门器官，将其转换为像素值。识别目标为图像中被识别器官占图像的像素比例与该识别的置信度的乘积求最大值。

Medical personnel rely on endoscopic assistance to visualize the glottis and achieve successful endotracheal intubation. Inexperienced resident physicians often fail to accurately locate the glottis during intubation, potentially causing patient injury. This dataset collects glottal images from volunteers and conducts deep learning analysis, enabling automatic camera capture and glottis position judgment to assist less experienced medical staff in intubation. This dataset and algorithm only focus on the visual display of human intracorporeal glottal images, and do not contain any personal or public data, with no data identifiers included. The algorithm is designed for glottis recognition and judgment under the working environment of laryngoscopes. Combined with the Non-Maximum Suppression (NMS) algorithm to filter out different organs present in the same image, the pixel proportion of the recognized organ in the image is calculated (i.e., the pixel area of the recognized organ divided by the total image area). The bounding box label is set for the glottal organ and converted into pixel values. The recognition target is to maximize the product of the pixel proportion of the recognized organ in the image and the confidence score of the recognition.

搜集汇总

数据集介绍

特点

该数据集包含445条声门图像数据，用于辅助医务人员在插管时识别声门位置。数据来源于浙江优亿医疗器械股份有限公司，通过深度学习算法处理，能够自动捕捉并判断声门位置，帮助经验不足的医务人员进行插管操作。