A method for contactless monitoring of tidal volume in radiotherapy based on state awareness

Abstract

The application provides a state-aware radiotherapy tidal volume non-contact monitoring method, which comprises the following steps: collecting a chest and abdomen depth image sequence of a target object, calculating a depth displacement field based on the depth image sequence, and determining a region of interest of the chest and abdomen of the target object; screening effective depth pixels, establishing an effective mask and a static baseline, determining the actual area of the pixels, and constructing a surface volume replacement quantity sequence; obtaining a reference tidal volume signal for individual calibration, and obtaining parameters required for tidal volume estimation; calculating a real-time tidal volume estimation value according to the current surface volume replacement quantity sequence and the parameters obtained through individual calibration, judging the breath-hold stability, and finally outputting the tidal volume estimation value, the target breath-hold interval, the breath-hold stability and the respiratory guidance information. The application realizes non-contact respiratory monitoring, improves the patient comfort, and provides a reliable basis for tidal volume estimation.

Description

Technical Field

[0001] This invention relates to the field of radiotherapy respiratory management and non-contact vital sign monitoring technology, and in particular to a method for non-contact monitoring of radiotherapy respiratory tidal volume based on state perception. Background Technology

[0002] Radiotherapy, a crucial treatment for tumors in the chest, abdomen, and breast, is often hampered by respiratory movements during its precise administration. Respiration causes positional changes in the chest and abdomen, lungs, heart, and tumor target area. Failure to effectively manage respiratory movements during radiotherapy can not only increase target localization errors but also expose surrounding organs to unnecessary radiation doses. To address this, the deep inspiration and breath-hold technique has been introduced clinically. This technique guides patients to hold their breath after a deep inspiration, increasing the spatial distance between the chest wall and the cardiopulmonary tissues, optimizing the geometric relationship between the target area and organs at risk, and thus playing a role in some chest and breast radiotherapy cases. However, this technique requires patients to achieve and maintain a stable breath-hold state during treatment, and relies on real-time monitoring of breath-hold stability by medical staff. Therefore, its practical application demands a high level of patient cooperation and clinical monitoring skills.

[0003] Currently, commonly used respiratory management devices in clinical practice include active breathing control devices, spirometers, breathing belts, optical surface monitoring systems, and camera monitoring devices. While active breathing control devices or spirometers can acquire tidal volume-related signals, they typically require contact or semi-contact components such as nose clips, mouthpieces, and breathing tubing, which can reduce patient comfort and increase the complexity of the treatment process. Optical surface monitoring or ordinary camera monitoring, while reducing contact and primarily used to observe changes in body surface position or displacement, struggles to consistently provide volume information meaningful for tidal volume, thus having certain functional limitations. Furthermore, the deep inspiration and breath-holding process is not a single, stable signal; its free breathing phase exhibits significant periodic fluctuations, while the breath-holding phase shows a low-variance plateau. If only a general regression method is used to convert body surface motion signals into tidal volume, signal drift or fluctuations that do not conform to physiological patterns can easily occur during the breath-holding plateau, affecting the accuracy and reliability of monitoring. Therefore, how to achieve stable and accurate monitoring of the breath-holding process while ensuring patient comfort remains a key issue that needs further resolution in current respiratory management technologies. Summary of the Invention

[0004] To address the above issues, this invention utilizes deep vision to non-contactly acquire chest and abdominal movements and establish individualized surface volume substitutions, converting three-dimensional deformation into quantified volume signals, obtaining personalized tidal volume parameters, and determining respiratory status in real time based on sequences. This enables non-contact respiratory monitoring, improves patient comfort, provides a reliable basis for tidal volume estimation, enhances the individual adaptability of parameter mapping, and can automatically distinguish respiratory stages and assess breath-holding quality, providing precise support for clinical respiratory management.

[0005] According to embodiments of the present invention, a method for non-contact monitoring of tidal volume during radiotherapy based on state awareness is provided. The method includes: Step S01: Acquire a depth image sequence of the chest and abdomen of the target object, calculate the depth displacement field based on the depth image sequence, and determine the region of interest of the chest and abdomen of the target object; Step S02: Select effective depth pixels within the region of interest, establish an effective mask and static baseline, determine the actual area of ​​the pixel based on the static baseline, and construct a surface volume substitution sequence based on the depth displacement field and the actual area; Step S03: Obtain the reference tidal volume signal for individual calibration to obtain the parameters required for tidal volume estimation; Step S04: Construct a historical input window based on the current surface volume substitution sequence, extract the statistical features of the historical input window, and determine whether the target object is in a free breathing state or a breath-holding state based on the statistical features; and calculate the real-time tidal volume estimate based on the parameters obtained from individual calibration. In the free breathing state, the tidal volume estimate that follows the dynamic changes of breathing is output. In the breath-holding state, the tidal volume estimate is made to converge to the steady-state reference value. Finally, the tidal volume estimate, the target breath-holding interval, the breath-holding stability, and the breathing guidance information are output.

[0006] Furthermore, the specific steps for determining the region of interest in the chest and abdomen of the target object in step S01 are as follows: Identify the human body region of the target object in the color image of the initial frame, and align the depth image to this color image coordinate system; The largest bounding box in the human body region is selected as the initial range of the chest and abdomen region of interest. In the processing of subsequent frames, the initial range is fixed, or it is smoothly updated according to the positional changes of its adjacent frames.

[0007] Further, the filtering of effective depth pixels in step S02 includes: calculating the effective proportion per pixel based on several initial frames of depth images, and determining pixels with an effective proportion higher than an adaptive threshold as effective pixels to form the effective mask; wherein, the adaptive threshold is obtained by the Otsu thresholding method or preset according to the depth image quality.

[0008] Furthermore, step S02 further includes: based on the effective mask, forming a dynamic mask according to the depth change amplitude of each effective pixel within the historical time window, which is used to further filter and retain the chest and abdominal surface regions that exhibit significant movement with respiration when constructing the surface volume substitution sequence.

[0009] Furthermore, the static baseline mentioned in step S02 is the median, mean, or robust estimate of the depth value of the same pixel in the initial several frames.

[0010] Furthermore, the formula for calculating the surface area substitution amount Vsurf(t) mentioned in step S02 is: Vsurf(t) = ∑[A i (Di (t) -Di0)], where Di (t) Let Di0 be the depth value of the i-th valid pixel in the current frame, and A be the baseline depth value of that pixel. i A represents the actual area of ​​the pixel in physical space. i =(Di0 / f x )×(Di0 / f y ), where f x f is the horizontal focal length of the depth vision acquisition unit. y This is the vertical focal length of the depth vision acquisition unit.

[0011] Further, the specific steps of step S03 are as follows: time-align the reference tidal volume signal with the synchronously obtained surface volume substitution sequence to form a calibration data pair, and obtain the scale parameters, bias parameters and baseline correction parameters required for tidal volume estimation based on the calibration data pair.

[0012] Further, the step S04, which involves constructing a historical input window based on the current surface volume replacement sequence and extracting statistical features from the historical input window, specifically involves extracting one or more statistical features from the historical input window, including the latest value, mean, standard deviation, local trend, and normalized variance. When the normalized variance is high and the surface volume replacement shows periodic changes, it is determined to be a free breathing state. When the normalized variance decreases and remains within a low fluctuation range for a preset time, it is determined to be a breath-holding state, and the breath-holding stability is judged based on the tidal volume estimate within the low fluctuation range.

[0013] Furthermore, when the breathing state is determined to be free breathing, the estimated tidal volume is output based on the dynamic change trend in the historical input window; When the state is determined to be breath-holding, a steady-state reference value is generated based on the mean, latest value and local trend of the historical input window, and the estimated tidal volume value is made to converge to the steady-state reference value.

[0014] Furthermore, the estimated tidal volume, target breath-hold interval, breath-hold stability, and respiratory guidance information mentioned in step S04 are provided to the radiotherapy operator through a monitoring display screen. At the same time, the respiratory guidance information is sent to a head-mounted display device, mixed reality display device, or visual feedback terminal worn or observed by the target subject to achieve real-time visual feedback.

[0015] This invention achieves contactless respiratory monitoring by using techniques such as depth vision to collect chest and abdominal movements without contact and establish individualized surface volume substitution, converting three-dimensional deformation into quantified volume signals, obtaining personalized tidal volume parameters, and judging respiratory status in real time based on sequences. This improves patient comfort, provides a reliable basis for tidal volume estimation, enhances the individual adaptability of parameter mapping, and can automatically distinguish respiratory stages and assess breath-holding quality, providing precise support for clinical respiratory management.

[0016] It should be understood that the description in the Summary of the Invention is not intended to limit the key or essential features of the embodiments of the present invention, nor is it intended to restrict the scope of the invention. Other features of the invention will become readily apparent from the following description.

[0017] The beneficial effects of this invention are: 1. By using depth vision to collect chest and abdominal movements non-contactly and establishing individualized surface volume substitution, the reliance on traditional contact breathing equipment is reduced, improving patient comfort and hygiene safety. 2. The three-dimensional deformation of the chest and abdomen is transformed into a quantifiable alternative volume signal, providing a reliable physiological and kinematic characterization basis for accurately estimating tidal volume; 3. By obtaining personalized tidal volume estimation parameters, the accuracy and individual adaptability of the mapping relationship from surface motion to tidal volume are significantly improved, and the error due to individual differences is reduced; 4. Based on the surface volume substitution sequence, the system can determine the respiratory status and breath-hold stability in real time, automatically and reliably distinguish between free breathing and breath-holding stages, and assess the quality of breath-holding. This provides key respiratory phase and stability information for treatment, offering precise and intuitive decision support for clinical respiratory management and improving the safety and repeatability of treatment. Attached Figure Description

[0018] The above and other features, advantages, and aspects of the various embodiments of the present invention will become more apparent from the accompanying drawings and the following detailed description. Wherein: Figure 1 A flowchart of a state-aware, contactless monitoring method for radiotherapy respiratory tidal volume according to an embodiment of the present invention is shown. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] According to an embodiment of the present invention, a non-contact monitoring method for tidal volume in radiotherapy based on state perception is proposed. By using techniques such as non-contact acquisition of chest and abdominal movements using depth vision and establishment of individualized surface volume substitution, conversion of three-dimensional deformation into quantified volume signals, acquisition of personalized tidal volume parameters, and real-time judgment of respiratory status based on sequences, non-contact respiratory monitoring is achieved, improving patient comfort, providing a reliable basis for tidal volume estimation, enhancing the individual adaptability of parameter mapping, and automatically distinguishing respiratory stages and assessing breath-holding quality, thus providing precise support for clinical respiratory management.

[0021] The principles and spirit of the present invention will be explained in detail below with reference to several representative embodiments.

[0022] Figure 1 This is a schematic flowchart of a method for non-contact monitoring of tidal volume in radiotherapy based on state awareness, according to an embodiment of the present invention. The method includes: Step S01: Acquire a depth image sequence of the chest and abdomen of the target object, calculate the depth displacement field based on the depth image sequence, and determine the region of interest of the chest and abdomen of the target object; Step S02: Select effective depth pixels within the region of interest, establish an effective mask and static baseline, determine the actual area of ​​the pixel based on the static baseline, and construct a surface volume substitution sequence based on the depth displacement field and the actual area; Step S03: Obtain the reference tidal volume signal for individual calibration to obtain the scale parameters, bias parameters or baseline correction parameters required for tidal volume estimation; Step S04: Construct a historical input window based on the current surface volume substitution sequence, extract the statistical features of the historical input window, and determine whether the target object is in a free breathing state or a breath-holding state based on the statistical features; and calculate the real-time tidal volume estimate based on the parameters obtained from individual calibration. In the free breathing state, the tidal volume estimate that follows the dynamic changes of breathing is output. In the breath-holding state, the tidal volume estimate is made to converge to the steady-state reference value. Finally, the tidal volume estimate, the target breath-holding interval, the breath-holding stability, and the breathing guidance information are output.

[0023] It should be noted that although the operation of the method of the present invention has been described in a specific order in the above embodiments and figures, this does not require or imply that the operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0024] To provide a clearer explanation of the above-described method for non-contact monitoring of tidal volume in radiotherapy based on state awareness, a specific embodiment will be used for illustration below. However, it is worth noting that this embodiment is only for better illustrating the present invention and does not constitute an improper limitation of the present invention.

[0025] The following example will further illustrate the method of non-contact monitoring of tidal volume in radiotherapy based on state awareness.

[0026] Step S01: Acquire a depth image sequence of the chest and abdomen of the target object, calculate the depth displacement field based on the depth image sequence, and determine the region of interest of the chest and abdomen of the target object.

[0027] A depth image sequence of the target's chest and abdomen is acquired using a depth vision acquisition unit positioned above or to the side of the target. This acquisition unit can be an Intel RealSense D435i depth camera, or a 3D vision device capable of outputting depth images, such as a binocular, structured light, or time-of-flight camera. The acquisition unit is fixed by a camera support to ensure its stability near the radiotherapy bed and to avoid interfering with the movement of the treatment bed or the beam output of the radiotherapy equipment.

[0028] Based on the acquired depth image sequence, a depth displacement field reflecting the positional changes of the chest and abdomen surface is calculated. The human body region of the target object is identified in the initial frame's color image, and the depth image is aligned to this color image coordinate system. The largest bounding box within the human body region is selected as the initial extent of the chest and abdomen region of interest (ROI). This extent must cover the target object's chest wall, upper abdomen, or other body surface areas that undulate significantly with respiration. To reduce signal errors caused by boundary jitter in the ROI, in subsequent frame processing, this ROI is either fixed or smoothly updated based on the positional changes in adjacent frames.

[0029] Step S02: Select effective depth pixels within the region of interest, establish an effective mask and static baseline, determine the actual area of ​​the pixel based on the static baseline, and construct a surface volume substitution sequence based on the depth displacement field and the actual area.

[0030] Within a defined region of interest (ROI) in the chest and abdomen, the effective proportion per pixel is calculated based on several initial frames of depth images. For a given pixel, if the percentage of frames with effective depth values ​​in the initial frames exceeds an adaptive threshold, the pixel is retained as an effective pixel; otherwise, it is discarded as an invalid pixel or background pixel. This adaptive threshold can be obtained using the Otsu thresholding method or preset according to the depth image quality. The set of all effective pixels constitutes the effective mask.

[0031] A static baseline is established within the effective pixel range. This baseline is the median, mean, or robust estimate of the depth value of the same pixel in the initial several frames, and is used to represent the reference position of the chest and abdominal surface of the target object when it is in the initial breathing state.

[0032] Based on this static baseline, the actual physical area represented by each valid pixel can be determined. Finally, by combining the depth displacement field calculated from the depth image sequence and the actual area of ​​each pixel, a surface volume substitution sequence characterizing the volume change of the thoracic and abdominal surface is constructed. Specifically, in the calculation, the depth value of the i-th valid pixel in the current frame is denoted as Di. (t) Let Di0 be the static baseline depth value corresponding to the pixel, then the depth displacement of the pixel is ΔDi. (t) =Di (t) -Di0. Based on the horizontal focal length f of the depth vision acquisition unit. x Vertical focal length f y And the baseline depth Di0 of the pixel, to determine the actual area A of the pixel in physical space. i For ΔDi within the effective pixel range (t) With A i Summing the products of , i.e., Vsurf(t) = ∑[A i ×(Di (t) -Di0)],A i =(Di0 / f x )×(Di0 / f y The surface volume substitution Vsurf(t) for the current frame is obtained. This surface volume substitution is used to characterize the volume change trend of the thoracic and abdominal surfaces due to respiration, without requiring the reconstruction of a complete 3D mesh.

[0033] After missing value imputation, low-pass filtering, direction normalization, and unit conversion, the surface volume surrogate sequence is processed further. Direction normalization is used to ensure that the increase or decrease direction of the surface volume surrogate sequence is consistent with the increase or decrease direction of the reference tidal volume signal, and unit conversion is used to ensure that the volume units are consistent with the clinical tidal volume units.

[0034] To further optimize signal quality, the depth variation amplitude of each effective pixel within the historical time window is calculated, and a dynamic mask is formed accordingly. This dynamic mask is used to further filter and retain the chest and abdominal surface areas that undergo significant movement with respiration within the effective pixel range, while reducing the interference of the bed surface, fixed support, clothing edges, or long-term unchanging areas on subsequent tidal volume estimation.

[0035] In subsequent processing, when the target object is determined to have entered the breath-holding stage, if the dynamic pixel ratio is too low, the mask used for signal calculation will be widened to the entire effective mask range to avoid the volume signal degenerating into a near-zero noise sequence at this time, thus ensuring the continuity and reliability of the signal.

[0036] Step S03: Obtain the reference tidal volume signal for individual calibration to obtain the parameters required for tidal volume estimation.

[0037] During the individual calibration phase, such as in respiratory training, CT simulation localization, or initial calibration, a reference tidal volume signal for the target subject is acquired using a reference respiratory monitoring device. This reference respiratory monitoring device can be an active breathing control device, a spirometer, or a respiratory training device.

[0038] Individual calibration is performed using a reference tidal volume signal and a synchronously acquired and constructed surface volume surrogate sequence to establish a correspondence between the two, thereby obtaining the scale parameters, bias parameters, or baseline correction parameters necessary for tidal volume estimation. In one embodiment, the reference tidal volume signal and the surface volume surrogate sequence are time-aligned according to timestamps to form frame-by-frame corresponding calibration data pairs. Individual calibration can be completed before respiratory training, CT simulation localization, or the first treatment; after individual calibration, the online monitoring phase mainly relies on the depth image sequence acquired by the depth vision acquisition unit for tidal volume estimation, without the need for continuous connection to the reference respiratory monitoring device.

[0039] Step S04: Construct a historical input window based on the current surface volume substitution sequence, extract the statistical features of the historical input window, and determine whether the target object is in a free breathing state or a breath-holding state based on the statistical features; and calculate the real-time tidal volume estimate based on the parameters obtained from individual calibration. In the free breathing state, the tidal volume estimate that follows the dynamic changes of breathing is output. In the breath-holding state, the tidal volume estimate is made to converge to the steady-state reference value. Finally, the tidal volume estimate, the target breath-holding interval, the breath-holding stability, and the breathing guidance information are output.

[0040] During the calibration phase, the chest and abdominal depth image sequences are converted into surface volume substitution sequences and time-aligned with the reference tidal volume signal output by a reference respiratory monitoring device to form calibration data pairs. Individual calibration units obtain scale parameters, bias parameters, and baseline correction parameters based on the calibration data pairs. During the online monitoring phase, the surface volume substitution sequence is generated using real-time depth image sequences, and combined with the calibration parameters, tidal volume estimates and breath-hold stability assessment results are output.

[0041] By analyzing the statistical characteristics of the surface volume substitution sequence within a historical time window, the respiratory status of the target object can be determined in real time. Based on the fluctuation pattern and amplitude changes of the sequence, it can be distinguished whether the target object is currently in a state of regular fluctuating free breathing or a relatively stable breath-holding state. This determination provides a state basis for subsequent tidal volume estimation and breath-holding stability analysis.

[0042] Specifically, using the surface volume replacement sequence prior to the current moment as the historical input window, the latest value, mean, standard deviation, local trend, and normalized variance of the historical input window are extracted. The normalized variance is used to determine whether the target object is in a free-breathing or breath-holding state. When the normalized variance is high and the surface volume replacement shows periodic changes, it is determined to be in a free-breathing state; when the normalized variance decreases and remains within a low fluctuation range for a preset time, it is determined to be in a breath-holding state.

[0043] During free breathing, the tidal volume estimation unit outputs an estimated tidal volume value based on the dynamic trend in the historical input window, ensuring that the estimated tidal volume changes with the inhalation and exhalation processes. During breath-holding, the tidal volume estimation unit generates a steady-state reference value based on the mean, latest value, and local trend of the historical input window, and converges the estimated tidal volume value to the steady-state reference value to suppress non-physiological fluctuations caused by depth noise or surface volume substitution drift during the breath-holding plateau phase.

[0044] In subsequent treatment phases, non-contact monitoring primarily relies on depth vision acquisition units. Based on respiratory status, the current surface volume substitution sequence is combined with parameters obtained from individual calibration to calculate a real-time tidal volume estimate. During breath-holding, the system further analyzes signal stability to assess breath-hold quality. Ultimately, the output includes the real-time tidal volume estimate, the preset target breath-hold interval, breath-hold stability indices, and respiratory guidance information. This information is provided to the radiotherapy operator via a monitoring display screen and can also be sent to a head-mounted display, mixed reality display, or visual feedback terminal worn or observed by the target subject for real-time visual feedback. For example, a Microsoft HoloLens 2 can be used as the visual feedback terminal. Simultaneously, all monitoring results are recorded, and alarm information is output when abnormal breath-holding is detected.

[0045] As one embodiment, during the calibration phase, the depth vision acquisition unit acquires a sequence of depth images of the target object's chest and abdomen at a preset frame rate, while a reference respiratory monitoring device synchronously outputs a reference tidal volume signal. Several initial frames are selected to establish the region of interest (ROI), effective mask, and static baseline for the chest and abdomen, and the depth displacement field of each frame is converted into a surface volume substitution quantity (Vsurf(t)). For example, during a breathing training session, as the reference tidal volume signal gradually increases from a low level to the target breath-holding interval, the surface volume substitution quantity sequence synchronously shows an upward trend. When the target object enters a breath-holding state, the reference tidal volume signal enters a plateau interval, and the fluctuation amplitude of the surface volume substitution quantity sequence decreases accordingly. Based on the above synchronous change relationship, scale parameters, bias parameters, and baseline correction parameters are obtained. During the online monitoring phase, without continuous connection to the reference respiratory monitoring device, an estimated tidal volume value can be output based on the real-time surface volume substitution quantity sequence and calibration parameters, and breath-holding stability can be determined based on whether the estimated value is within the target breath-holding interval.

[0046] In one experimental validation, a depth camera was used to acquire chest and abdominal depth image sequences at 30Hz, and a reference tidal volume signal was acquired simultaneously. The experiment included 22 volunteers, obtaining 347,963 frames of valid synchronized data. Each volunteer spent approximately 10 minutes collecting data and performed 6-8 deep breath-holding maneuvers, with each breath-holding lasting approximately 15-30 seconds. Cross-validation with one participant remaining was used for evaluation. The overall Pearson correlation coefficient between the tidal volume estimation results and the reference tidal volume signal was 0.968, the overall mean absolute error was 0.098L, and the coefficient of determination R² was 0.926. Specifically, the mean absolute error during the breath-holding phase was 0.1059L, and the mean absolute error during free breathing was 0.0967L. The GPU latency for surface volume substitution construction and model inference was 12.55ms / frame, which is sufficient for real-time respiratory feedback.

[0047] Furthermore, although the operations are described in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the invention. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0048] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A method for state-aware radiotherapy breath tidal volume contactless monitoring, characterized in that, The method includes: Step S01: Acquire a depth image sequence of the chest and abdomen of the target object, calculate the depth displacement field based on the depth image sequence, and determine the region of interest of the chest and abdomen of the target object; Step S02: Select effective depth pixels within the region of interest, establish an effective mask and static baseline, determine the actual area of ​​the pixel based on the static baseline, and construct a surface volume substitution sequence based on the depth displacement field and the actual area; Step S03: Obtain the reference tidal volume signal for individual calibration to obtain the parameters required for tidal volume estimation; Step S04: Construct a historical input window based on the current surface volume substitution sequence, extract the statistical features of the historical input window, and determine whether the target object is in a free breathing state or a breath-holding state based on the statistical features; and calculate the real-time tidal volume estimate based on the parameters obtained from individual calibration. In the free breathing state, the tidal volume estimate that follows the dynamic changes of breathing is output. In the breath-holding state, the tidal volume estimate is made to converge to the steady-state reference value. Finally, the tidal volume estimate, the target breath-holding interval, the breath-holding stability, and the breathing guidance information are output.

2. The method of claim 1, wherein, The specific steps for determining the region of interest in the chest and abdomen of the target object in step S01 are as follows: Identify the human body region of the target object in the color image of the initial frame, and align the depth image to this color image coordinate system; The largest bounding box in the human body region is selected as the initial range of the chest and abdomen region of interest. In the processing of subsequent frames, the initial range is fixed, or it is smoothly updated according to the positional changes of its adjacent frames.

3. The method for non-contact monitoring of tidal volume in radiotherapy based on state awareness according to claim 1, characterized in that, The filtering of effective depth pixels in step S02 includes: calculating the effective proportion per pixel based on several initial frames of depth images, and determining pixels with an effective proportion higher than an adaptive threshold as effective pixels to form the effective mask; wherein, the adaptive threshold is obtained by the Otsu thresholding method or preset according to the depth image quality.

4. The method for non-contact monitoring of tidal volume in radiotherapy based on state awareness according to claim 1, characterized in that, Step S02 further includes: based on the effective mask, forming a dynamic mask according to the depth change amplitude of each effective pixel within the historical time window, which is used to further filter and retain the chest and abdominal surface regions that exhibit significant movement with respiration when constructing the surface volume substitution sequence.

5. The method for non-contact monitoring of tidal volume in radiotherapy based on state awareness according to claim 1, characterized in that, The static baseline mentioned in step S02 is the median, mean, or robust estimate of the depth value of the same pixel in the initial several frames.

6. The method for non-contact monitoring of tidal volume in radiotherapy based on state awareness according to claim 1, characterized in that, The formula for calculating the surface area substitution amount Vsurf(t) mentioned in step S02 is: Vsurf(t) = ∑[A i (Di (t) -Di0)], where Di (t) Let Di0 be the depth value of the i-th valid pixel in the current frame, and A be the baseline depth value of that pixel. i A represents the actual area of ​​the pixel in physical space. i =(Di0 / f x )×(Di0 / f y ), where f x f is the horizontal focal length of the depth vision acquisition unit. y This is the vertical focal length of the depth vision acquisition unit.

7. The method for non-contact monitoring of tidal volume in radiotherapy based on state awareness according to claim 1, characterized in that, The specific steps of step S03 are as follows: time-align the reference tidal volume signal with the synchronously obtained surface volume substitution sequence to form a calibration data pair, and obtain the scale parameters, bias parameters and baseline correction parameters required for tidal volume estimation based on the calibration data pair.

8. The method for non-contact monitoring of tidal volume in radiotherapy based on state awareness according to claim 1, characterized in that, The step S04, which involves constructing a historical input window based on the current surface volume replacement sequence and extracting statistical features from the historical input window, specifically involves extracting one or more statistical features from the latest value, mean, standard deviation, local trend, and normalized variance. When the normalized variance is high and the surface volume replacement shows periodic changes, it is determined to be a free breathing state. When the normalized variance decreases and remains within a low fluctuation range for a preset time, it is determined to be a breath-holding state, and the breath-holding stability is judged based on the tidal volume estimate within the low fluctuation range.

9. A method for non-contact monitoring of tidal volume in radiotherapy based on state awareness, as described in claim 8, is characterized in that... When the breathing state is determined to be free breathing, the estimated tidal volume is output according to the dynamic change trend in the historical input window; When the breath-holding state is determined, a steady-state reference value is generated based on the mean, latest value and local trend of the historical input window, and the estimated tidal volume value is made to converge to the steady-state reference value.

10. A method for non-contact monitoring of tidal volume in radiotherapy based on state awareness, as described in claim 1, characterized in that, The estimated tidal volume, target breath-hold interval, breath-hold stability, and respiratory guidance information mentioned in step S04 are provided to the radiotherapy operator through a monitoring display screen. At the same time, the respiratory guidance information is sent to a head-mounted display device, mixed reality display device, or visual feedback terminal worn or observed by the target subject to achieve real-time visual feedback.

Images