Search

CN-121997843-A - Non-contact type respiratory mechanics reverse reconstruction system based on 4D holographic field perception

CN121997843ACN 121997843 ACN121997843 ACN 121997843ACN-121997843-A

Abstract

The invention belongs to the technical field of holographic respiration analysis, and discloses a non-contact type respiratory mechanics reverse reconstruction system based on 4D holographic field perception; the method comprises the steps of receiving an echo sequence of a millimeter wave sparse aperture array on a chest and abdomen area, generating a 4D point cloud sequence of a chest and abdomen surface time-varying morphology according to frames, constructing a respiratory observation sequence of the chest and abdomen surface, generating respiratory drive characteristics by analyzing the relevance of a geodesic length change attribute and a curvature tensor evolution sequence energy change attribute, obtaining observation confidence according to the time change of the chest and abdomen area in the respiratory drive characteristics, inputting the respiratory drive characteristics into a reverse reconstruction network containing nonlinear viscoelastic respiratory dynamics constraint, and outputting a mechanical parameter set of airway resistance and pleural cavity pressure change. According to the invention, 4D morphological observation is converted into structural boundary excitation, and the inversion process is weighted and constrained by the observation confidence, so that the expression integrity of the inhalation driving is enhanced, and the stability and the reliability of mechanical parameter output are improved.

Inventors

  • WU JINGHAN
  • GAO XIAOWEI
  • WANG E
  • DING RUIYANG
  • Zhuang Wuping

Assignees

  • 湖南夏天医疗器械有限公司
  • 中南大学湘雅医院

Dates

Publication Date
20260508
Application Date
20260409

Claims (10)

  1. 1. Non-contact respiratory mechanics reverse reconstruction system based on 4D holographic field perception, which is characterized by comprising: The 4D holographic sensing and respiration observation module receives an echo sequence of the millimeter wave sparse aperture array on the chest and abdomen area, generates a 4D point cloud sequence of a time-varying form of the chest and abdomen surface according to frames, and constructs a respiration observation sequence of the chest and abdomen surface, wherein the respiration observation sequence comprises a geodesic line length sequence, an area expansion rate sequence and a chest and abdomen area velocity field sequence; the geodesic and curvature energy correlation module is used for generating breathing drive characteristics by analyzing the correlation between the geodesic length change attribute and the curvature tensor evolution sequence energy change attribute; the breath drive confidence evaluation module is used for obtaining an observation confidence according to the time change of the chest and abdomen area in the breath drive characteristic; And the viscoelastic dynamic reverse reconstruction module inputs the breathing driving characteristics into a reverse reconstruction network containing nonlinear viscoelastic breathing dynamic constraints and outputs a mechanical parameter set of airway resistance and pleural cavity pressure change.
  2. 2. The non-contact respiratory mechanics reverse reconstruction system based on 4D holographic field perception of claim 1, wherein the method of constructing a respiratory observation sequence of a thoracoabdominal surface comprises: constructing a triangle mesh set of the chest and abdomen surface by analyzing the adjacency relation between different points in each sampling frame; calculating the local principal curvature of each point and converging to generate a curvature tensor evolution sequence; Constructing a monitoring anchor point track set of the thoracic and abdominal surface areas, and calculating a geodesic length sequence, an area expansion rate sequence and a thoracic and abdominal area speed field sequence; And combining the ground wire length sequence, the area expansion rate sequence and the chest and abdomen area velocity field sequence to generate a respiration observation sequence.
  3. 3. The 4D holographic field aware based non-contact respiratory mechanics reverse reconstruction system according to claim 2, wherein the method of generating curvature tensor evolution sequence comprises: traversing the point set under each frame, and marking the points obtained by traversing as target points; Calculating normal vector of the target point, constructing a normal plane coordinate system perpendicular to the normal vector, constructing edge vectors of each point in the target point and adjacent point list, and marking the edge vectors as adjacent edge vectors; Calculating the normal difference between the adjacent point and the target point, and respectively projecting the adjacent side projection vector and the normal difference into a normal plane coordinate system to obtain an adjacent side projection vector and a normal difference projection vector; respectively calculating the outer product matrixes of the projection vectors of different adjacent edges and accumulating to obtain the projection matrixes of the adjacent edges; calculating the product matrix of adjacent edge projection vectors and normal difference projection vectors formed by the same adjacent points, and accumulating the product matrix of different adjacent points to obtain a normal response matrix; Calculating a curvature tensor matrix of the target point according to the adjacent edge projection matrix and the normal response matrix, carrying out symmetrical processing and feature decomposition on the curvature tensor matrix, and taking two feature values as local principal curvatures of the target point; and combining local principal curvatures of each point in the point set under different adjacent frames to obtain curvature tensors, and splicing the curvature tensors according to the sequence of frame numbers to obtain a curvature tensor evolution sequence.
  4. 4. The 4D holographic field aware-based non-contact respiratory mechanics reverse reconstruction system of claim 3, wherein the method of generating a respiratory observation sequence comprises: Selecting triangular grids respectively positioned in a thoracic region and an abdominal region from a point set of each frame, screening the triangular grids respectively positioned in three target regions, and marking the triangular grids as target grids; marking vertexes of different target grids in the target area as anchor points of the target area; respectively enumerating anchor point pairs positioned in a chest region and an abdomen region for each frame, calculating all side lengths of triangular meshes on the chest and abdomen surface of each frame for the anchor point pairs in each region, randomly selecting one anchor point between the anchor point pairs, starting from the edge of the anchor point to search for another anchor point, and obtaining a search path; enumerating the search path to obtain an anchor point pair path, accumulating all side lengths in the anchor point pair path to obtain path length; Selecting an anchor point pair path corresponding to the minimum path length as a geodesic line for the same anchor point pair, and combining the sequences of passing points in the process of generating the geodesic line to obtain a geodesic line path; And repeatedly solving the geodesic path on the continuous frames for each anchor point pair, and accumulating the path length to obtain a geodesic length sequence.
  5. 5. The non-contact respiratory mechanics reverse reconstruction system based on 4D holographic field perception of claim 4, wherein the method for calculating the sequence of area expansion ratios comprises: Taking different geodesic wires of each frame of chest and abdomen surface area as the center, and expanding the geodesic wires to two sides according to fixed widths to obtain geodesic wire strip-shaped areas; counting the area of the geodesic banded region, and calculating the difference of geodesic banded regions in adjacent frames for geodesic banded regions under different frames of the same anchor point pair; and calculating the ratio of the differential result to the area of the geodesic banded region of the last frame in the adjacent frames to obtain the area expansion rate, and combining the area expansion rates according to the development sequence of the frame sequence numbers to obtain an area expansion rate sequence.
  6. 6. The non-contact respiratory mechanics reverse reconstruction system based on 4D holographic field perception of claim 5, wherein the method for calculating the chest-abdomen region velocity field sequence comprises: Counting the displacement vector of each point under the adjacent frames, calculating the dot product of the unit normal vector and the displacement vector, and marking the dot product as normal displacement; Integrating each point in the point set to participate in the constructed triangular grids, and marking the integration result as a participation grid list of the corresponding point; Calculating area accumulated values of all triangular grids in the grid list, halving the area accumulated values and carrying out normalization processing to obtain area weights of corresponding points; The normal displacements of different points in the surface areas of the thorax and the abdomen are respectively weighted and fused according to the areas, and the weighted fusion result and the frame interval duration of the adjacent frames are calculated to obtain the thorax displacement speed and the abdomen displacement speed; And combining the chest displacement speed and the abdomen displacement speed according to the frame sequence number development sequence to obtain a chest-abdomen area speed field.
  7. 7. The 4D holographic field aware based non-contact respiratory mechanics reverse reconstruction system according to claim 6, wherein the method of generating respiratory drive features comprises: Identifying a monotonic section of the geodesic length sequence, and respectively incorporating the monotonic section with ascending geodesic length and the monotonic section with descending geodesic length into an inhalation segment candidate set and an exhalation segment candidate set; respectively calculating the rising slope and the falling slope of curvature tensor evolution sequence energy in the inspiration segment candidate set and the expiration segment candidate set to generate an expansion rate indication sequence; constructing a set of key moments including inspiration onset, dwell time and expiration onset by analyzing consistency of slope change attributes in the expansion rate indication sequence and growth in the area expansion rate sequence; Extracting inhalation duration and exhalation duration from the set of key moments; Extracting the maximum displacement speed of the chest and abdomen area in each respiratory cycle in the chest and abdomen area speed field to obtain the peak expansion speed of the chest and abdomen area; counting the area expansion rate sequences in each respiratory cycle, respectively acquiring and accumulating the area expansion amounts of the thoracic and abdominal regions, and marking the accumulated result as a body surface expansion agent; And integrating the inspiration duration, expiration duration, peak expansion rate, inspiration-to-expiration conversion steepness and the body surface expansion agent quantity to obtain the breathing drive characteristics of the chest-abdomen body surface area.
  8. 8. The 4D holographic field aware based non-contact respiratory mechanics reverse reconstruction system according to claim 7, wherein the method of generating the dilation rate indication sequence comprises: The method comprises the steps of converging thoracic and abdominal regions, squaring and summing local principal curvatures of each point in a curvature tensor evolution sequence to obtain curvature energy; combining curvature energy of all points in the same area according to frame numbers; the curvature energy sequence is subjected to interval slicing according to the inhalation segment candidate set and the exhalation segment candidate set to respectively obtain an inhalation segment energy sub-sequence set and an exhalation segment energy sub-sequence set; Respectively carrying out sliding fitting on each inhalation segment energy sub-sequence and exhalation segment energy sub-sequence in the segment to respectively obtain an ascending slope sequence and a descending slope sequence; and splicing the ascending slope sequence and the descending slope sequence, and marking the splicing result as an expansion rate indication sequence.
  9. 9. The 4D holographic field aware based non-contact respiratory mechanics reverse reconstruction system of claim 8, wherein the constructing comprises a method of a set of key moments of inspiration onset, dwell moments and expiration onset, comprising: Identifying a continuous non-negative number segment and a continuous non-positive number segment in the area expansion rate sequence, and matching a timestamp corresponding to the continuous non-negative number segment with a timestamp of an ascending slope; Marking the earliest time obtained by matching as the inspiration starting time, and marking the latest time obtained by matching as the inspiration peak time; Matching the time stamp corresponding to the continuous non-positive number segment with the time stamp of the descending slope, and marking the earliest time obtained by matching as the expiration starting time; And integrating key moments of inspiration starting, inspiration peak value and expiration starting to obtain a key moment set.
  10. 10. The non-contact respiratory mechanics reverse reconstruction system based on 4D holographic field perception of claim 9, wherein the method of outputting the set of mechanical parameters of airway resistance, pleural cavity pressure variation, comprises: Defining the pleural cavity pressure change and airway resistance as state variables output by a network, and taking the breathing driving characteristics as boundary excitation observance; inputting the breathing driving characteristics and the observation confidence as boundary excitation into a reverse reconstruction network, and drawing the observation confidence as the sample weight of training loss to generate weighted joint loss; and determining network convergence and outputting a state variable by analyzing the relevance of the descending rate change attribute of the weighted joint loss in the iterative process and the smoothness change attribute of the state variable time sequence, so as to obtain a mechanical parameter set.

Description

Non-contact type respiratory mechanics reverse reconstruction system based on 4D holographic field perception Technical Field The invention relates to the technical field of holographic respiration analysis, in particular to a non-contact type respiratory mechanics reverse reconstruction system based on 4D holographic field perception. Background Respiratory mechanics parameters (such as airway resistance, compliance, and pleural cavity pressure changes) are important grounds for assessing ventilation status, identifying respiratory depression risk, and guiding perioperative respiratory management. In recent years, non-contact sensing means such as millimeter wave radar can acquire motion information of the chest and abdomen surface changing along with time without wearing a sensor, and a new technical path is provided for continuous respiration evaluation in clinical monitoring, postoperative recovery and home monitoring scenes. The existing non-contact respiration monitoring and respiration parameter deducing method takes single-point displacement, local amplitude or single time sequence curve as observation input, and when the chest and abdomen surface deformation presents local fluctuation, uneven expansion or different area response differences, more complete boundary description on procedural information such as respiration conversion, expansion rate and the like is difficult to form, so that conditional expression for further developing reverse reconstruction of respiratory mechanics is limited. Meanwhile, in continuous monitoring, the observation stability and the effective information amount of different respiratory cycles may be different, if quantification of the observation reliability and explicit constraint in the inversion process are absent, the problem that the stability of mechanical parameter output is insufficient or the confidence is difficult to evaluate under the influence of a fluctuation segment is easy to occur, so that it is necessary to develop a non-contact type respiratory mechanics reverse reconstruction method capable of extracting structural driving characteristics from the time-varying morphology of the chest and abdomen surface and performing constraint in combination with the observation reliability. Disclosure of Invention In order to overcome the defects in the prior art and achieve the purposes, the invention provides a non-contact type respiratory mechanics reverse reconstruction system based on 4D holographic field perception, which comprises the following technical scheme: The 4D holographic sensing and respiration observation module receives an echo sequence of the millimeter wave sparse aperture array on the chest and abdomen area, generates a 4D point cloud sequence of a time-varying form of the chest and abdomen surface according to frames, and constructs a respiration observation sequence of the chest and abdomen surface, wherein the respiration observation sequence comprises a geodesic line length sequence, an area expansion rate sequence and a chest and abdomen area velocity field sequence; the geodesic and curvature energy correlation module is used for generating breathing drive characteristics by analyzing the correlation between the geodesic length change attribute and the curvature tensor evolution sequence energy change attribute; the breath drive confidence evaluation module is used for obtaining an observation confidence according to the time change of the chest and abdomen area in the breath drive characteristic; And the viscoelastic dynamic reverse reconstruction module inputs the breathing driving characteristics into a reverse reconstruction network containing nonlinear viscoelastic breathing dynamic constraints and outputs a mechanical parameter set of airway resistance and pleural cavity pressure change. Preferably, the method for constructing a respiratory observation sequence of the thoracoabdominal surface comprises the following steps: constructing a triangle mesh set of the chest and abdomen surface by analyzing the adjacency relation between different points in each sampling frame; calculating the local principal curvature of each point and converging to generate a curvature tensor evolution sequence; Constructing a monitoring anchor point track set of the thoracic and abdominal surface areas, and calculating a geodesic length sequence, an area expansion rate sequence and a thoracic and abdominal area speed field sequence; And combining the ground wire length sequence, the area expansion rate sequence and the chest and abdomen area velocity field sequence to generate a respiration observation sequence. Preferably, the method for generating a curvature tensor evolution sequence comprises the following steps: traversing the point set under each frame, and marking the points obtained by traversing as target points; Calculating normal vector of the target point, constructing a normal plane coordinate system perpendicular to the normal vector,