基于喉镜的会厌识别数据

气管插管过程：喉镜插入部从患者的右口角轻轻插入，当推进喉镜片至会厌谷处，上提喉镜，间接挑起会厌，即显露声门，随即将右手的导管按弧线经口送人咽部，通过声门插入气管。从喉镜进入咽喉腔挑起会厌至气管导管插入气管内，动作越轻柔，应激反应越低。本数据通过采集志愿者的咽喉部图像，进行深度学习分析，可以实现对于会厌的识别，防止气管插管过程中误触或伤害会厌产生后遗症。本数据和算法仅针对人体内的咽喉部图像画面显示，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，对会厌进行识别和判断。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），标签坐标框取为会厌器官，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。

Tracheal intubation procedure: The insertion part of the laryngoscope is gently inserted into the right corner of the patient's mouth. When the laryngoscope blade is advanced to the epiglottic vallecula, the laryngoscope is lifted to indirectly elevate the epiglottis, thereby exposing the glottis. Subsequently, the endotracheal tube held in the right hand is sent into the pharynx along an arc path through the mouth, and inserted into the trachea through the glottis. The gentler the movements from inserting the laryngoscope into the pharyngeal cavity and elevating the epiglottis to inserting the endotracheal tube into the trachea, the lower the patient's stress response will be. This dataset collects pharyngeal images from volunteers and conducts deep learning analysis, enabling the recognition of the epiglottis to prevent accidental contact or injury to the epiglottis during tracheal intubation and avoid subsequent complications. This dataset and the corresponding algorithm only target the display of human pharyngeal images, and do not contain personal data, public data, or any data identifiers. The proposed algorithm performs recognition and judgment on the epiglottis under the working environment of a laryngoscope. Combined with the non-maximum suppression (NMS) algorithm, different organs in the same image are filtered out. The pixel proportion of the recognized organ in the image is calculated, that is, the pixel area occupied by the recognized organ divided by the total image area. The bounding box of the label is set for the epiglottis 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 this recognition.

搜集汇总

数据集介绍

特点

该数据集包含337条喉镜图像数据，用于气管插管过程中的会厌识别，采用深度学习算法分析图像中的会厌位置和比例。