A decision-making method for endoscopic intelligent intubation based on target point detection

Abstract

The invention discloses an endoscopic intelligent intubation decision-making method based on target point detection, comprising the following steps: marking training data; training target point classification network; training target point coordinate regression network; splicing to realize the overall target point regression network; and Endoscopic Intubation Decision Module. The invention explicitly separates perception and decision-making, makes the network more transparent, has interpretability, and is convenient for modification, transplantation and expansion. As far as the algorithm is concerned, the regression target point position provides a basis for continuous decision-making, which is conducive to achieving higher-precision intubation decisions.

Description

technical field [0001] The invention relates to the fields of computer vision and medical image processing, in particular to an endoscope intelligent intubation decision method based on target point detection. Background technique [0002] Endoscope is a detection instrument that integrates traditional optics, ergonomics, precision machinery, modern electronics, mathematics, and software. It can enter the stomach, trachea or other natural orifices through the mouth to see lesions that cannot be seen on X-rays. At present, endoscopic intubation is performed by professional doctors, which greatly increases the workload of doctors. In addition, the endoscopic intubation scenes are mostly narrow lumens of the human body, dark and lack of texture, which are difficult to identify with the naked eye, resulting in high personnel training costs, lack of universality in manual operations, and difficulty in handling many complicated intubations question. The use of computer-assisted...

AI Technical Summary

Problems solved by technology

Although this method is effective and convenient, and can achieve preliminary decision-making, it can only give rough direction information, but cannot give continuous specific rotation and bending angles.

For the continuous change process of endoscopic intubation, discrete and rough direction instructions are difficult to control the endoscopic intubation machine to make precise and intelligent operations

Moreover, the end-to-end black-box model lacks intermediate steps, and the overall lack of interpretability means that further tuning will become difficult. It is impossible to refine the overall problem, and it is also difficult to improve the overall performance from a single step, which is not convenient for further improvement. Transplantation, Need for a More Flexible, Transparent Decision-Making Scheme for Endoscopic Cannulation

Images