Radiotherapy patient state monitoring method based on infrared thermal imaging

Abstract

The invention relates to the technical field of medical monitoring, in particular to a radiotherapy patient state monitoring method based on infrared thermal imaging, which is characterized by comprising the following steps: S1, in radiotherapy, continuously collecting complete infrared thermal images of a patient; s2, identifying a human body temperature area and an environment temperature area in the infrared thermal image; s3, denoising is carried out on the infrared thermal imaging image, and an environment temperature area is eliminated; s4, performing temperature gradient identification and temperature area identification on the human body temperature area in the S3, and distinguishing the head, the trunk and the limbs in the image area; s5, tracking position changes of the head, the trunk and the four limbs in the two continuous infrared thermal images, and when it is judged that the position changes of the head, the trunk and the four limbs in the two infrared thermal images exceed set values, sending out an early warning signal to the outside of the treatment room; the risk of patient privacy exposure in a video monitoring scheme is completely avoided, no physical feature data is stored, and the possibility of patient privacy leakage is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of medical monitoring, and in particular to a method for monitoring the status of radiotherapy patients based on infrared thermal imaging. Background Art

[0002] Radiation therapy is a core method for cancer treatment, and its development has always revolved around the two major goals of "precise irradiation" and "safety control." Modern precision radiotherapy technologies (such as IMRT and IGRT) combined with position fixation devices (thermoplastic film, vacuum negative pressure pads, etc.) can control the target dose error within ±3%. However, clinical practice has revealed that there are two safety risks during treatment: one is millimeter-level position deviation (incidence rate 3.8%-7.2%), which affects efficacy and safety; the other is that patients may suddenly experience significant abnormal conditions (such as coma, convulsions, vomiting, or waving), which urgently requires immediate intervention to interrupt treatment. Existing active monitoring technologies mainly focus on solving high-precision (submillimeter) displacement monitoring, but have significant limitations when applied to abnormal state identification: 1) Computer vision systems (such as OSMS): Although highly accurate, continuous monitoring of sensitive areas (breast, pelvis, etc.) causes serious patient privacy concerns (67% of patients expressed concerns) and psychological pressure on medical staff (32% experienced burnout), and face HIPAA regulatory compliance risks; 2) Wearable sensing systems (such as RT-DoseGuard): There are problems such as equipment interference (affecting dose distribution and image quality), complex operation (requiring long-term calibration and disinfection), and high cost (requiring regular replacement).

[0003] The bottleneck of existing technology is that its core design is to solve the high-precision measurement of "whether the position is accurate", rather than the efficient, low-cost and low-risk detection of "whether the status is abnormal". When the current system identifies key safety events such as coma, convulsions, vomiting, waving, etc., it is difficult to deploy due to privacy ethics restrictions, or difficult to popularize due to cumbersome operations and high costs, or there are side effects that interfere with treatment. Therefore, there is an urgent need for a complementary monitoring solution in the clinic. Its core should focus on high-sensitivity identification of patients with significant abnormal conditions (with low requirements for position accuracy), avoiding privacy risks, avoiding treatment interference, and being easy to operate and cost-controlled, so as to fill the gaps in the existing high-precision monitoring system in macro-safety event warning and jointly build a more comprehensive radiotherapy safety guarantee. Summary of the Invention

[0004] To address the above issues, it is necessary to develop a method for monitoring the status of radiotherapy patients based on infrared thermal imaging, which includes the following steps:

[0005] S1: During radiotherapy, continuous acquisition of complete infrared thermal images of the patient;

[0006] S2: Identify the human body temperature zone and the ambient temperature zone in the infrared thermal imaging image;

[0007] S3: De-noise the infrared thermal image and remove the ambient temperature area;

[0008] S4: performing temperature gradient recognition and temperature zone area recognition on the human body temperature zone in S3, and distinguishing the head, torso and limbs in the image area;

[0009] S5: Track the position changes of the head, torso and limbs in two consecutive infrared thermal images. When it is determined that the position changes of the head, torso and limbs in the two infrared thermal images exceed the set value, an early warning signal is sent to the outside of the treatment room.

[0010] S1 continuously collects complete infrared thermal images of the patient, including front-angle images and side-angle images of the patient in a supine position. The position tracking of the head, torso and limbs in the front-angle image and the side-angle image is independent of each other, and the paths for transmitting warning signals based on the judgment results of the front-angle image and the side-angle image are in a serial relationship.

[0011] An image is captured at the top position of the torso in the frontal angle image, and the same color area peak image within a set period is selected as the first typical value. During the treatment, the same position of the frontal angle image is captured frame by frame and compared with the first typical value for similarity. The image with the highest similarity is extracted as the first comparison image, and the time interval between the first comparison images is obtained as the first judgment of the respiratory frequency, thereby obtaining the first respiratory signal, and sending the first respiratory signal to the outside of the treatment room.

[0012] An image is captured at the top position of the torso in the side angle image, and the area valley value image of the same color within a set period is selected as the second typical value. During the treatment, the same position of the side angle image is captured frame by frame and compared with the second typical value for similarity. The image with the highest similarity is extracted as the second comparison image, and the time interval between the second comparison images is obtained as the second judgment of the respiratory frequency. When the deviation between the second judgment and the first judgment exceeds the set value, an early warning of an abnormal first respiratory signal is issued to the outside of the treatment room.

[0013] In step S4, the millimeter-wave radar point cloud data and infrared thermal imaging are introduced for multimodal fusion, including:

[0014] A millimeter-wave radar array is installed diagonally above the treatment bed, and the radar point cloud data and infrared thermal imaging timestamps are aligned through a time synchronization module;

[0015] A weighted fusion algorithm is used to fuse the thermal imaging temperature gradient characteristics with the radar micro-motion characteristics;

[0016] When constructing the 3D motion vector model, the axial displacement parameters of the radar data are preferentially used in combination with the tangential displacement parameters of the thermal imaging.

[0017] The dynamic ambient temperature compensation algorithm is used in step S3, specifically including:

[0018] Deploy reference temperature sensors at the four corners of the treatment room to collect ambient temperature gradient distribution in real time

[0019] When establishing a three-dimensional temperature field model, the grid units are divided with the treatment bed as the center.

[0020] The influence of ambient temperature fluctuation is eliminated by adaptive Kalman filter, and its state equation parameters are set as:

[0021]

[0022] Where: T_{corr} is the corrected body surface temperature (℃ / S); T_{raw} is; is the ambient temperature change rate (°C / s); α is the infrared data confidence weight; the value range is 0.82±0.03; β is the environmental change sensitivity coefficient, the value range is 0.18±0.03 (clinically verified optimization coefficient) to eliminate false positive alarms caused by temperature fluctuations in the treatment room (such as air conditioning start and stop, door opening and closing).

[0023] For local temperature changes lasting more than 5 seconds, ΔT>0.5℃, the artifact recognition module is activated and automatically switches to the millimeter-wave radar-dominated monitoring mode.

[0024] The respiratory rate determination introduces a deep learning compensation mechanism, and the specific implementation includes:

[0025] Build an LSTM neural network model with training data consisting of 1,200 clinical respiratory waveforms, including abnormal patterns such as coughing and breath holding;

[0026] The input layer receives: thermal imaging breathing signals and millimeter wave chest rise and fall signals;

[0027] The hidden layer adopts a bidirectional LSTM structure, and the output layer provides the respiratory phase prediction value;

[0028] When the time domain difference between the infrared and radar signals exceeds 200ms, the confidence-weighted decision is initiated:

[0029] F_{final}=w1·F_{thermal}+w2·F_{radar};

[0030] Where: F_{final} is the fused respiratory rate (times / minute); F_{thermal} is the thermal imaging respiratory signal (based on chest rise and fall); F_{radar} is the radar micro-Doppler frequency shift signal; w1 and w2 are dynamic weights, with the initial value of w1 being 0.6 and the initial value of w2 being 0.4. w1 and w2 are automatically adjusted according to the signal-to-noise ratio.

[0031] The early warning signal triggering mechanism includes a multi-level response strategy:

[0032] Level 1 warning (displacement 5-10mm): The treatment bed will automatically pause and a voice prompt "Please maintain your position" will be activated.

[0033] Level 2 warning (displacement 11-20mm): triggers the treatment gantry to pause rotation and simultaneously activates the positioning laser indication system;

[0034] Level 3 warning (displacement > 20mm): Immediately stop the radiation output and unlock the emergency lighting system in the treatment room.

[0035] All warning events are recorded with timestamps, displacement parameters and disposal measures through the blockchain evidence storage module.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] Breakthrough innovation in privacy protection

[0038] Completely avoid the risk of patient privacy exposure in video surveillance solutions. Clinical tests have shown improved patient acceptance. No physical feature data is stored, which reduces the possibility of patient privacy leakage compared to pure visual monitoring solutions. Compared with non-contact sensors, it improves detection efficiency and reduces false alarm rates while reducing equipment layout and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art; in all drawings, similar elements or parts are generally identified by similar figure marks; in the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0040] Figure 1 is a flow chart of the method of the present invention;

[0041] Figure 2 This is a reference diagram of the hardware layout of the present invention;

[0042] Figure 3 This is a reference image for the front angle view;

[0043] Figure 4 This is a reference image for the side angle view. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, but not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0045] The following is combined with Figure 1-4 , introduces a method for monitoring the status of radiotherapy patients based on infrared thermal imaging proposed by the present invention, comprising the following steps:

[0046] S1: During radiotherapy, continuous acquisition of complete infrared thermal images of the patient is required. In specific implementation, a short-focus thermal imager can be considered. The principle of acquisition and framing is to increase the proportion of the human body in the field of view as much as possible and reduce the proportion of the environment. The specific method of acquiring complete infrared thermal images of the patient can be referred to Figure 1 .

[0047] S2: Identify the human body temperature zone and the ambient temperature zone in the infrared thermal imaging image; it is understandable that, under the premise of ensuring the normal and stable operation of the treatment equipment, the room temperature of the treatment room is generally set at around 25 degrees Celsius, and the average temperature of various parts of the normal adult body surface is between 32.6℃-35.9℃. The ambient temperature obviously has a large gradient difference from the body surface temperature, so the color block of room temperature (ambient temperature) can be well distinguished and eliminated.

[0048] S3: De-noise the infrared thermal image and remove the ambient temperature area; the purpose is to reduce the difficulty of subsequent data processing, reduce the amount of graphic data processing, shorten the data processing cycle, and only process the color blocks within the human body surface temperature range.

[0049] S4: Perform temperature gradient and area identification on the human body temperature zones in S3, distinguishing the head, torso, and limbs within the image region. The average temperature of the human body surface temperature is highest at the navel and nape, followed by the anterior neck and popliteal fossa. The average temperature on the back of the hand is lowest, and the palm and anterior pubic region are slightly higher, respectively. In infrared thermal images, the general trend is that the torso and head are hot, with their temperature patterns forming a blocky pattern; the limbs are cold, with their temperature patterns forming a stripe pattern. During algorithm recognition, further optimization can be performed based on shape, using a shape model to further optimize the recognition of the pattern and reduce data processing. It is understood that in a static state, the temperature zones of a human body infrared image remain stable over time. Even if there are amplitude changes, they are gradual, continuous, and slow. For example, as treatment progresses, the patient may become tense, causing the surface temperature to rise; or, if the patient feels well and gradually relaxes, the surface temperature may rise, without sudden changes. The core innovation of this method lies in identifying temperature zones within captured infrared thermal images. For example, in a side-view image, the surface temperature of the torso is 35.2°C, representing 20% ​​of the 32.6°C-35.9°C range. If the patient then struggles or turns sideways, the 35.2°C area in the next captured infrared thermal image drops to 15%. This temperature variation allows for the identification of many significant limb changes.

[0050] S5: Track the position changes of the head, torso and limbs in two consecutive infrared thermal images. When it is determined that the position changes of the head, torso and limbs in the two infrared thermal images exceed the set value, an early warning signal is sent to the outside of the treatment room.

[0051] Non-contact infrared thermal imaging continuously monitors the patient's overall thermal distribution, avoiding the privacy risks associated with video surveillance (no anatomical details are imaged in sensitive areas) while also preventing treatment interference from wearable devices. Macroscopic displacement monitoring (not submillimeter) of the head, torso, and limbs can highly sensitively identify significant abnormalities such as coma, convulsions, and vomiting, triggering immediate intervention and filling a gap in existing high-precision monitoring systems for early warning of safety incidents.

[0052] S1 continuously captures complete infrared thermal images of the patient, including frontal and side-angle images of the patient in a prone position. The position tracking of the head, torso, and limbs in the frontal and side-angle images is independent of each other, and the paths for transmitting warning signals based on the judgment results of the frontal and side-angle images are in series. The dual-angle independent monitoring and series warning mechanism significantly improves the reliability of abnormal state identification: the frontal angle image optimizes the capture of limb twitching / waving movements, and the side-angle image enhances the detection of torso rolling / vomiting movements. Dual-path verification can reduce the false alarm rate caused by occlusion from a single perspective (such as an arm blocking the torso).

[0053] The top position of the torso in the side angle image is captured, and the peak image of the same color area within a set period is selected as the first typical value. During the treatment process, the same position of the side angle image is captured frame by frame and compared with the first typical value for similarity. The image with the highest similarity is extracted as the first comparison image. The time interval between the first comparison images is obtained as the first determination of the respiratory frequency, and the first respiratory signal is obtained. The first respiratory signal is then sent to the outside of the treatment room. Since human breathing is regular, the regular graphic changes of the top position of the torso (corresponding to the chest and abdomen) can be analyzed. By analyzing the respiratory frequency, the patient's vital signs can also be monitored. Combined with the changes in body surface temperature during the treatment process, the changes in the patient's vital signs during radiotherapy can be further studied.

[0054] An image is captured at the top position of the torso in the side angle image, and the area valley value image of the same color within a set period is selected as the second typical value. During the treatment, the same position of the side angle image is captured frame by frame and compared with the second typical value for similarity. The image with the highest similarity is extracted as the second comparison image, and the time interval between the second comparison images is obtained as the second judgment of the respiratory frequency. When the second judgment and the first judgment deviate from the set value, an early warning of the abnormality of the first respiratory signal is issued to the outside of the treatment room. The second comparison image is an error correction judgment of the first comparison image.

[0055] Thermal imaging of the top of the torso extracts respiratory signals for contactless vital sign monitoring. A typical value comparison algorithm eliminates complex sensor calibration and provides early warning clues for coma or respiratory arrest. A dual-angle respiratory signal cross-validation mechanism effectively identifies abnormal breathing patterns (such as respiratory disturbances caused by convulsions). A first / second judgment deviation warning can detect single-angle monitoring failures (such as a patient rolling over, causing distortion of the frontal signal), improving system robustness.

[0056] In step S4, the millimeter-wave radar point cloud data and infrared thermal imaging are introduced for multimodal fusion, including:

[0057] A millimeter-wave radar array is installed diagonally above the treatment bed, and the radar point cloud data and infrared thermal imaging timestamps are aligned through a time synchronization module;

[0058] A weighted fusion algorithm is used to fuse the thermal imaging temperature gradient characteristics with the radar micro-motion characteristics;

[0059] When constructing a 3D motion vector model, the axial displacement parameters of radar data are preferentially used, combined with the tangential displacement parameters of thermal imaging. Millimeter-wave radar compensates for the defects of thermal imaging that are affected by sudden changes in ambient temperature (such as sudden sweating causing distortion of heat distribution).

[0060] Weighted fusion algorithm improves displacement tracking confidence (especially for weak twitching movements);

[0061] The radar axial displacement data and the thermal imaging tangential data complement each other to construct a three-dimensional motion vector model to improve the recognition accuracy of body collapse after coma. The purpose is to add a new error correction mechanism, which can be used as a preferred solution, but it will increase the layout cost of the equipment.

[0062] The dynamic ambient temperature compensation algorithm is used in step S3, specifically including:

[0063] Deploy reference temperature sensors at the four corners of the treatment room to collect the ambient temperature gradient distribution in real time;

[0064] When establishing a three-dimensional temperature field model, the grid cells are divided with the treatment bed as the center;

[0065] The influence of ambient temperature fluctuation is eliminated by adaptive Kalman filter, and its state equation parameters are set as:

[0066]

[0067] Where: T_{corr} is the corrected body surface temperature (℃ / S); T_{raw} is; is the rate of change of ambient temperature (°C / s); α is the confidence weight of infrared data; the value range is 0.82±0.03; β is the sensitivity coefficient of environmental change; the value range is 0.18±0.03 (clinically verified optimization coefficient) to eliminate false positive alarms caused by temperature fluctuations in the treatment room (such as air conditioning start and stop, door opening and closing);

[0068] For local temperature changes lasting more than 5 seconds (ΔT>0.5°C), the artifact recognition module is activated and automatically switches to the millimeter-wave radar-dominated monitoring mode. Based on the three-dimensional temperature field modeling of the reference sensor, the artifacts caused by the air flow of the treatment room air conditioning are eliminated (the false alarm rate is reduced by 62% when ΔT>0.5°C)

[0069] Adaptive Kalman filter parameter optimization (α / β coefficients) ensures monitoring stability of hypothermic body surface areas (such as limbs).

[0070] Automatically switching to millimeter-wave dominant mode when the temperature suddenly changes to ensure the continuous operation of the system is intended to add a new error correction mechanism. This can be used as a preferred solution, but it will increase the layout cost of the equipment.

[0071] The respiratory rate determination introduces a deep learning compensation mechanism, and the specific implementation includes:

[0072] Construct an LSTM neural network model with training data containing 1,200 clinical respiratory waveforms, including abnormal patterns such as coughing and breath holding.

[0073] The input layer receives: thermal imaging breathing signals and millimeter wave chest rise and fall signals;

[0074] The hidden layer adopts a bidirectional LSTM structure, and the output layer provides the respiratory phase prediction value;

[0075] When the time domain difference between the infrared and radar signals exceeds 200ms, the confidence-weighted decision is initiated:

[0076] F_{final}=w_1·F_{thermal}+w_2·F_{radar}.

[0077] It can identify complex breathing patterns such as coughing and breath holding, and is trained with data from over 1,200 cases to reduce false shutdowns caused by brief movements.

[0078] The early warning signal triggering mechanism includes a multi-level response strategy:

[0079] Level 1 warning (displacement 5-10mm): The treatment bed will automatically pause and a voice prompt "Please maintain your position" will be activated.

[0080] Level 2 warning (displacement 11-20mm): triggers the treatment gantry to pause rotation and simultaneously activates the positioning laser indication system;

[0081] Level 3 warning (displacement > 20mm): Immediately stop the radiation output and unlock the emergency lighting system in the treatment room. At this time, medical staff will enter the treatment room urgently after receiving the signal to conduct manual investigation and treatment.

[0082] All warning events are recorded with timestamps, displacement parameters and disposal measures through the blockchain evidence storage module.

[0083] It will be understood that when used in this specification and the appended claims, the terms “comprises” and “comprising” indicate the presence of stated features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof;

[0084] It should also be understood that the terms used in the present description are for the purpose of describing particular embodiments only and are not intended to limit the present invention; as used in the present description and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates otherwise;

[0085] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

Claims

1. A method for monitoring the status of a radiotherapy patient based on infrared thermal imaging, characterized in that: The following steps are involved: S1: During radiotherapy, continuous acquisition of complete infrared thermal images of the patient; S2: Identify the human body temperature zone and the ambient temperature zone in the infrared thermal imaging image; S3: De-noise the infrared thermal image and remove the ambient temperature area; S4: performing temperature gradient recognition and temperature zone area recognition on the human body temperature zone in S3, and distinguishing the head, torso and limbs in the image area; S5: Track the position changes of the head, torso and limbs in two consecutive infrared thermal images. When it is determined that the position changes of the head, torso and limbs in the two infrared thermal images exceed the set value, an early warning signal is sent to the outside of the treatment room.

2. The method for monitoring the status of a radiotherapy patient based on infrared thermal imaging according to claim 1, characterized in that: S1 continuously collects complete infrared thermal images of the patient, including front-angle images and side-angle images of the patient in a supine position. The position tracking of the head, torso and limbs in the front-angle image and the side-angle image is independent of each other, and the paths for transmitting warning signals based on the judgment results of the front-angle image and the side-angle image are in a serial relationship.

3. The method for monitoring the status of a radiotherapy patient based on infrared thermal imaging according to claim 2, characterized in that: An image is captured at the top position of the torso in the frontal angle image, and the same color area peak image within a set period is selected as the first typical value. During the treatment, the same position of the frontal angle image is captured frame by frame and compared with the first typical value for similarity. The image with the highest similarity is extracted as the first comparison image, and the time interval between the first comparison images is obtained as the first judgment of the respiratory frequency, thereby obtaining the first respiratory signal, and sending the first respiratory signal to the outside of the treatment room.

4. The method for monitoring the status of a radiotherapy patient based on infrared thermal imaging according to claim 3, characterized in that: An image is captured at the top position of the torso in the side angle image, and the area valley value image of the same color within a set period is selected as the second typical value. During the treatment, the same position of the side angle image is captured frame by frame and compared with the second typical value for similarity. The image with the highest similarity is extracted as the second comparison image, and the time interval between the second comparison images is obtained as the second judgment of the respiratory frequency. When the deviation between the second judgment and the first judgment exceeds the set value, an early warning of an abnormal first respiratory signal is issued to the outside of the treatment room.

5. The method for monitoring the status of a radiotherapy patient based on infrared thermal imaging according to claim 4, characterized in that: In step S4, the millimeter-wave radar point cloud data and infrared thermal imaging are introduced for multimodal fusion, including: A millimeter-wave radar array is installed diagonally above the treatment bed, and the radar point cloud data and infrared thermal imaging timestamps are aligned through a time synchronization module; A weighted fusion algorithm is used to fuse the thermal imaging temperature gradient characteristics with the radar micro-motion characteristics; When constructing the 3D motion vector model, the axial displacement parameters of the radar data are preferentially used in combination with the tangential displacement parameters of the thermal imaging.

6. The method for monitoring the status of a radiotherapy patient based on infrared thermal imaging according to claim 1, characterized in that: The dynamic ambient temperature compensation algorithm is used in step S3, specifically including: Deploy reference temperature sensors at the four corners of the treatment room to collect ambient temperature gradient distribution in real time When establishing a three-dimensional temperature field model, the grid units are divided with the treatment bed as the center. The influence of ambient temperature fluctuation is eliminated by adaptive Kalman filter, and its state equation parameters are set as: Where: T_{corr} is the corrected body surface temperature (℃ / S); T_{raw} is; is the ambient temperature change rate (℃ / S); α is the infrared data confidence weight, with a value range of 0.82±0.03; β is the environmental change sensitivity coefficient, with a value range of 0.18±0.03; For local temperature changes lasting more than 5 seconds, ΔT>0.5℃, the artifact recognition module is activated and automatically switches to the millimeter-wave radar-dominated monitoring mode.

7. The method for monitoring the status of a radiotherapy patient based on infrared thermal imaging according to claim 5, characterized in that: The respiratory rate determination introduces a deep learning compensation mechanism, including: Build an LSTM neural network model. The training data includes 1,200 clinical respiratory waveforms, including abnormal cough and breath-hold patterns. The input layer receives: thermal imaging respiratory signals and millimeter-wave chest rise and fall signals. The hidden layer adopts a bidirectional LSTM structure, and the output layer provides the respiratory phase prediction value; When the time domain difference between the infrared and radar signals exceeds 200ms, the confidence-weighted decision is initiated: F_{final}=w1·F_{thermal}+w2·F_{radar}; Among them; F_{final} is the fused respiratory rate (times / minute); F_{thermal} is the thermal imaging respiratory signal (based on chest rise and fall); F_{radar} is the radar micro-Doppler frequency shift signal; w1 and w2 are dynamic weights, with the initial value of w1 being 0.6 and the initial value of w2 being 0.

4. w1 and w2 are automatically adjusted according to the signal-to-noise ratio.

8. The method for monitoring the status of a radiotherapy patient based on infrared thermal imaging according to claim 1, characterized in that: The early warning signal triggering mechanism includes a multi-level response strategy: Level 1 warning (displacement 5-10mm): The treatment bed will automatically pause and a voice prompt "Please maintain your position" will be activated. Level 2 warning (displacement 11-20mm): triggers the treatment gantry to pause rotation and simultaneously activates the positioning laser indication system; Level 3 warning (displacement > 20mm): Immediately stop the radiation output and unlock the emergency lighting system in the treatment room; All warning events are recorded with timestamps, displacement parameters and disposal measures through the blockchain evidence storage module.

Images