Pituitary tumor information acquisition method and device

Abstract

The present application provides a method and device for obtaining pituitary tumor information. In each step of obtaining pituitary tumor information through the nasal cavity, the method automatically identifies and calibrates the parts and regions based on the trained model, so the accuracy of part identification is high, and the accuracy, rationality and safety of regional calibration are also improved. In addition, each step automatically determines the next operation task based on the identification result and calibration content, and generates the next operation prompt information based on the operation task, so that the dependence on the physician's experience is reduced and the efficiency is improved. Therefore, the present application can realize real-time intelligent navigation throughout the process, reduce the difficulty of information acquisition, and facilitate the rapid and accurate acquisition of pituitary tumor information, as well as the determination of subsequent plans based on the information.

Description

Technical Field

[0001] The present application relates to the field of medical assistance technology, and in particular to a method and device for acquiring pituitary tumor information. Background Art

[0002] Pituitary tumors are a common intracranial tumor. The symptoms vary depending on the location of the tumor compression. Gigantism, dwarfism, and acromegaly are all caused by pituitary tumors, which are very harmful and need to be removed surgically. Traditional surgery requires opening the skull to remove pituitary tumors. The operation is long, risky, and has many postoperative complications. However, with the continuous improvement of nasal endoscopy technology, a nasal endoscope is used to enter from one nasal cavity, open the sphenoid sinus, and then pass through the sphenoid sinus to reach the pituitary fossa, obtain information about the pituitary tumor, and determine the subsequent removal plan based on this information.

[0003] However, currently, obtaining information about pituitary tumors through nasal endoscopy mostly relies on the physician's experience to control the endoscope to collect images in the nasal cavity, identify various tissue structures, and determine the location and shape of the area that needs to be removed. The accuracy and precision of manual identification of parts and determination of areas are low, and there are certain risks when removing parts, making it difficult to obtain pituitary tumor information and the entire process inefficient.

[0004] Therefore, there are currently technical problems in obtaining pituitary tumor information through the nasal cavity, such as difficulty in acquisition and low efficiency, which need to be improved. Summary of the Invention

[0005] The embodiments of the present application provide a method, device, electronic device and storage medium for obtaining pituitary tumor information, which are used to alleviate the technical problems of difficulty and low efficiency in obtaining pituitary tumor information through the nasal cavity.

[0006] To solve the above technical problems, the embodiments of the present application provide the following technical solutions:

[0007] The present application provides a method for obtaining pituitary tumor information, comprising:

[0008] Obtaining a first white-light image captured by the endoscope within the nasal cavity of the target case based on first operation prompt information, the first operation prompt information being generated based on a superior turbinate search task, calling a trained superior turbinate recognition model to identify the superior turbinate in the first white-light image to obtain first recognition information, determining a sphenoid sinus opening search task based on the first recognition information, and generating second operation prompt information;

[0009] obtaining a second white-light image acquired by the endoscope based on the second operation prompt information, calling the trained sphenoid sinus opening recognition model to recognize the sphenoid sinus opening in the second white-light image to obtain second recognition information, determining a sphenoid sinus target opening area calibration task based on the second recognition information, and generating third operation prompt information;

[0010] acquiring a third white light image acquired by the endoscope based on the third operation prompt information, and calling the trained sphenoid sinus target opening region segmentation model to segment and calibrate the sphenoid sinus target opening region in the third white light image to obtain first calibration data;

[0011] In response to the sphenoid sinus removal signal in the sphenoid sinus target opening area, a sellar floor bone target opening area calibration task is determined, and fourth operation prompt information is generated; a fourth white light image acquired by the endoscope based on the fourth operation prompt information is acquired, a trained sellar floor target detection model is called to identify and calibrate the sellar floor in the fourth white light image to obtain second calibration data, and pre-stored third calibration data of the sellar floor and fourth calibration data of the sellar floor posterior structure group are called, the third calibration data and the fourth labeling data are obtained based on the MRI coronal scan image of the target case through the pituitary gland; fifth calibration data of the sellar floor bone target opening area is obtained according to the second calibration data, the third calibration data and the fourth calibration data;

[0012] In response to the sellar floor bone removal signal in the sellar floor bone target opening area, a sellar floor dura mater target opening area calibration task is determined, and fifth operation prompt information is generated; a fifth white light image acquired by the endoscope based on the fifth operation prompt information is acquired, a trained sellar floor dura mater target detection model is called to identify and calibrate the sellar floor dura mater in the fifth white light image to obtain sixth calibration data, a trained sellar floor posterior structure group segmentation model is called to segment and calibrate the sellar floor posterior structure group in the fifth white light image to obtain seventh calibration data, and eighth calibration data of the sellar floor dura mater target opening area is obtained based on the sixth calibration data and the seventh calibration data;

[0013] In response to the removal signal of the sellar dura mater in the sellar dura mater target opening area, the pituitary tumor information acquisition task is determined and the sixth operation prompt information is generated; the sixth white light image collected by the endoscope based on the sixth operation prompt information is obtained, the trained tumor segmentation model is called to segment the pituitary tumor in the sixth white light image, and the pituitary tumor information is obtained according to the segmentation result.

[0014] At the same time, the embodiment of the present application also provides a pituitary tumor information acquisition device, including:

[0015] a first acquisition module, configured to acquire a first white-light image captured by the endoscope within the nasal cavity of a target case based on first operation prompt information, the first operation prompt information being generated based on a superior turbinate search task, calling a trained superior turbinate recognition model to identify the superior turbinate in the first white-light image to obtain first recognition information, determining a sphenoid sinus opening search task based on the first recognition information, and generating second operation prompt information;

[0016] a second acquisition module, configured to acquire a second white light image acquired by the endoscope based on the second operation prompt information, call the trained sphenoid sinus opening recognition model to recognize the sphenoid sinus opening in the second white light image to obtain second recognition information, determine a sphenoid sinus target opening area calibration task based on the second recognition information, and generate third operation prompt information;

[0017] a third acquisition module, configured to acquire a third white light image acquired by the endoscope based on the third operation prompt information, and call the trained sphenoid sinus target opening area segmentation model to segment and calibrate the sphenoid sinus target opening area in the third white light image to obtain first calibration data;

[0018] A first obtaining module is configured to determine a sellar floor bone target opening area calibration task in response to a sphenoid sinus removal signal in the sphenoid sinus target opening area, and generate fourth operation prompt information; obtain a fourth white light image acquired by the endoscope based on the fourth operation prompt information, call a trained sellar floor target detection model to identify and calibrate the sellar floor in the fourth white light image to obtain second calibration data, and call pre-stored third calibration data of the sellar floor and fourth calibration data of the sellar floor posterior structure group, the third calibration data and the fourth labeling data being obtained based on a trans-pituitarism MRI coronal scan image of the target case; obtain fifth calibration data of the sellar floor bone target opening area based on the second calibration data, the third calibration data and the fourth calibration data;

[0019] a second obtaining module, configured to determine a sellar floor dura target opening area calibration task in response to a sellar floor bone removal signal in the sellar floor bone target opening area, and generate fifth operation prompt information; obtain a fifth white light image acquired by the endoscope based on the fifth operation prompt information, call a trained sellar floor dura target detection model to identify and calibrate the sellar floor dura in the fifth white light image to obtain sixth calibration data, call a trained sellar floor posterior structure group segmentation model to segment and calibrate the sellar floor posterior structure group in the fifth white light image to obtain seventh calibration data, and obtain eighth calibration data of the sellar floor dura target opening area based on the sixth calibration data and the seventh calibration data;

[0020] The third obtaining module is used to determine the pituitary tumor information acquisition task in response to the removal signal of the sellar dura mater in the sellar dura mater target opening area, and generate sixth operation prompt information; obtain the sixth white light image collected by the endoscope based on the sixth operation prompt information, call the trained tumor segmentation model to segment the pituitary tumor in the sixth white light image, and obtain the pituitary tumor information according to the segmentation result.

[0021] Beneficial effects: The present application provides a method and device for obtaining pituitary tumor information. In each step of obtaining pituitary tumor information through the nasal cavity, the method automatically identifies and calibrates the parts and regions based on the trained model, so the accuracy of part identification is high, and the accuracy, rationality and safety of regional calibration are also improved. In addition, each step automatically determines the next operation task based on the identification result and calibration content, and generates the next operation prompt information based on the operation task, so that the degree of dependence on the physician's experience is reduced and the efficiency is improved. Therefore, the present application can realize real-time intelligent navigation throughout the process, reduce the difficulty of information acquisition, and facilitate the rapid and accurate acquisition of pituitary tumor information, as well as the determination of subsequent plans based on the information. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The following detailed description of the specific embodiments of the present application in conjunction with the accompanying drawings will make the technical solutions and other beneficial effects of the present application apparent.

[0023] Figure 1 This is a schematic diagram of an application scenario of the method for obtaining pituitary tumor information provided in an embodiment of the present application.

[0024] Figure 2 A flowchart of the method for obtaining pituitary tumor information provided in an embodiment of the present application.

[0025] Figure 3 This is a schematic diagram of various parts of the pituitary tumor information acquisition path in the embodiment of this application.

[0026] Figure 4 Schematic diagram of the sphenoid sinus opening and candidate sphenoid sinus opening areas in an embodiment of the present application.

[0027] Figure 5 This is a schematic diagram of the calibration of the candidate sphenoid sinus opening area and the removal of the sphenoid sinus target removal site in an embodiment of the present application.

[0028] Figure 6 This is a schematic diagram of the saddle bottom and the saddle bottom rear structure group in the embodiment of the present application.

[0029] Figure 7 This is a schematic diagram of the fifth calibration data in an embodiment of the present application.

[0030] Figure 8This is a schematic diagram of the target bone removal site at the sellar floor after removal in an embodiment of the present application.

[0031] Figure 9 This is a schematic diagram of the saddle dura mater and the eighth calibration data in an embodiment of the present application.

[0032] Figure 10 Schematic diagram of the removal of a pituitary tumor in an embodiment of the present application.

[0033] Figure 11 Schematic diagram of the removal of an invading tumor in an embodiment of the present application.

[0034] Figure 12 This is a schematic diagram of the structure of the pituitary tumor information acquisition device provided in an embodiment of the present application.

[0035] Figure 13 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the embodiments described are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0037] The embodiments of the present application provide a method, device, electronic device and computer-readable storage medium for obtaining pituitary tumor information, wherein the pituitary tumor information obtaining device can be integrated into an electronic device, which can be a server, a terminal or other device.

[0038] See also Figure 1 , Figure 1 A schematic diagram of a scenario for the application of the method for obtaining pituitary tumor information provided in an embodiment of the present application, wherein the scenario may include terminals and servers, and the terminals, servers, and terminals and servers are connected and communicated through the Internet composed of various gateways, and the application scenario includes a neuroendoscope and an intelligent navigation system. During the entire process from the beginning of the endoscope entering the nasal cavity to the in-depth acquisition of pituitary tumor information, the intelligent navigation system calls various trained models in real time to process the images according to the information processing status of the images collected by the endoscope, obtains useful information from the images, determines the next operation task based on the useful information, and generates various recognition results, calibration data, and prompt information for the physician, so that the physician can perform each step of the operation based on this information, quickly, accurately, and efficiently obtain pituitary tumor information, and efficiently execute subsequent plans after obtaining the pituitary tumor information.

[0039] It should be noted that Figure 1The system scenario diagram shown is only an example. The server and scenario described in the embodiment of this application are intended to more clearly illustrate the technical solution of the embodiment of this application and do not constitute a limitation on the technical solution provided by the embodiment of this application. A person skilled in the art will know that with the evolution of the system and the emergence of new business scenarios, the technical solution provided by the embodiment of this application is also applicable to similar technical problems. The following are detailed descriptions. It should be noted that the order of description of the following embodiments is not necessarily a limitation on the preferred order of the embodiments.

[0040] See also Figure 2 , Figure 2 This is a first flow chart of a method for obtaining pituitary tumor information provided in an embodiment of the present application, which specifically includes:

[0041] S1: Obtain a first white light image captured by the endoscope in the nasal cavity of the target case based on the first operation prompt information, the first operation prompt information is generated based on the superior turbinate search task, call the trained superior turbinate recognition model to identify the superior turbinate in the first white light image, obtain first recognition information, determine the sphenoid sinus opening search task based on the first recognition information, and generate second operation prompt information.

[0042] like Figure 3 As shown, the pituitary tumor is distributed in the pituitary gland of the target case. When the pituitary tumor is surgically removed through the nasal cavity, it is necessary to obtain the pituitary tumor information first. In this process, the doctor needs to first insert the neuroendoscope into the nasal cavity, and identify the superior turbinate by moving the endoscope forward, and then identify the sphenoid sinus opening, and then open the sphenoid sinus, identify the sellar floor, and then open the sellar floor bone and sellar floor dura mater in turn, and finally obtain the pituitary tumor information in the pituitary gland. In this process, it is necessary to identify various parts such as the superior turbinate, sphenoid sinus opening, sellar floor, sellar floor posterior structure group, etc., and it is also necessary to identify and calibrate various areas such as the sphenoid sinus target opening area, sellar floor bone target opening area, sellar floor dura mater target opening area. Through the above-mentioned intelligent navigation system, this application can make the accuracy of part identification higher, and the accuracy, rationality and safety of regional calibration are also improved. The following is a detailed description of each step in conjunction with a specific embodiment.

[0043] In the initial state, the intelligent navigation system first provides the physician with the first operation prompt information. The first operation prompt information is generated based on the superior turbinate search task. The first operation prompt information is used to prompt the physician for the next operation based on the current position of the endoscope and the superior turbinate search task to be performed, such as prompting "go forward to the upper left or forward to the lower left to find the superior turbinate", etc. The first operation prompt information will be calculated and updated in real time as the current position of the endoscope changes, so as to assist the physician in efficiently moving the endoscope to the appropriate position.

[0044] During the camera movement process, the intelligent navigation system receives in real time the first white light image collected by the endoscope based on the first operation prompt information, and calls the trained superior turbinate recognition model to identify the superior turbinate in the first white light image to obtain the first recognition information. During the execution of the superior turbinate search task, the images collected by the endoscope are all first white light images, that is, there can be more than one first white light image. When the camera is not moved to the appropriate position, the first recognition information of the first white light image is that the superior turbinate is not recognized. When the camera is moved to the appropriate position, the first recognition information of the first white light image is that the superior turbinate is recognized. The superior turbinate recognition model preferably selects the VGG16 model. Before calling, a large number of historical neuroendoscopic white light images are first obtained, and the labels of "superior turbinate" and "other" are added. Then, the model is trained as a training data set. After the training is completed, the model has a high accuracy in identifying the superior turbinate.

[0045] When the first identification information is that the superior turbinate is identified, the intelligent navigation system automatically determines a new operation task, namely the sphenoid sinus opening search task. The second operation prompt information is used to prompt the physician for the next operation based on the current position of the endoscope and the sphenoid sinus opening search task to be performed, such as prompting "go forward to the upper left or forward to the lower left to find the sphenoid sinus opening". The second operation prompt information will be calculated and updated in real time according to the current position of the endoscope to assist the physician in efficiently moving the endoscope to the appropriate position.

[0046] S2: Obtain a second white light image collected by the endoscope based on the second operation prompt information, call the trained sphenoid sinus opening recognition model to identify the sphenoid sinus opening in the second white light image, obtain second recognition information, determine the sphenoid sinus target opening area calibration task according to the second recognition information, and generate third operation prompt information.

[0047] During the mirror movement process, the intelligent navigation system receives in real time the second white light image collected by the endoscope based on the second operation prompt information, and calls the trained sphenoid sinus opening recognition model to recognize the sphenoid sinus opening in the second white light image to obtain second recognition information, such as Figure 4 As shown in the figure, the area pointed by the arrow in the dotted box is the sphenoid sinus opening. During the sphenoid sinus opening search task, the images collected by the endoscope are all second white-light images, that is, there can be more than one second white-light image. When the mirror is not moved to the appropriate position, the second identification information of the second white-light image is that the sphenoid sinus opening is not identified. When the mirror is moved to the appropriate position, the second identification information of the second white-light image is that the sphenoid sinus opening is identified. The sphenoid sinus opening recognition model preferably uses the VGG16 model. Before calling it, a large number of historical neuroendoscopic white-light images are first obtained and labeled with "sphenoid sinus opening" and "others". Then, the model is trained as a training data set. After the training is completed, the model has a high accuracy in identifying the sphenoid sinus opening.

[0048] When the second identification information is that the sphenoid sinus opening is recognized, the intelligent navigation system automatically determines a new operation task, namely the sphenoid sinus target opening area calibration task. The third operation prompt information is used to prompt the physician for the next operation based on the current position of the endoscope and the sphenoid sinus target opening area calibration task to be performed, such as prompting "push the lens forward or move the lens backward or keep the lens still", etc. The third operation prompt information will be calculated and updated in real time according to the current position of the endoscope to assist the physician in efficiently moving the lens to the appropriate position.

[0049] S3: Acquire a third white light image collected by the endoscope based on the third operation prompt information, call the trained sphenoid sinus target opening area segmentation model to segment and calibrate the sphenoid sinus target opening area in the third white light image, and obtain first calibration data.

[0050] During the mirror movement process, the intelligent navigation system receives the third white light image collected by the endoscope based on the third operation prompt information in real time, calls the trained sphenoid sinus target opening area segmentation model to segment and calibrate the sphenoid sinus target opening area in the third white light image, obtains the first calibration data and displays it to the physician, such as Figure 4 and Figure 5 As shown, the dotted box is the first calibration data for the sphenoid sinus target opening area. The area within this area is the sphenoid sinus target removal site, which includes part of the superior turbinate, the natural ostium of the sphenoid sinus, the middle turbinate, and the nasal septum. Performing the site removal operation in this area can maximize safety while taking into account the requirements for pituitary tumor removal. The sphenoid sinus target opening area model preferably uses the Unet++ model. Before calling it, a large number of historical neuroendoscopic white light images are first obtained, and the boundaries of the sphenoid sinus target opening area are outlined by professional physicians. This is then used as a training dataset to train the model. After training, the model has high accuracy in segmenting and calibrating the sphenoid sinus target opening area.

[0051] S4: In response to the removal signal of the sphenoid sinus in the sphenoid sinus target opening area, the sellar floor bone target opening area calibration task is determined, and a fourth operation prompt information is generated; a fourth white light image collected by the endoscope based on the fourth operation prompt information is acquired, and the trained sellar floor target detection model is called to identify and calibrate the sellar floor in the fourth white light image to obtain the second calibration data, and the pre-stored third calibration data of the sellar floor and the fourth calibration data of the sellar floor posterior structure group are called, and the third calibration data and the fourth marking data are obtained based on the MRI coronal scanning image of the target case through the pituitary gland; according to the second calibration data, the third calibration data and the fourth calibration data, the fifth calibration data of the sellar floor bone target opening area is obtained.

[0052] After the physician removes the sphenoid sinus target removal portion within the sphenoid sinus target opening area based on the first calibration data, Figure 5 and Figure 6The sellar floor will be exposed as shown. The physician can provide feedback to the intelligent navigation system on the removal results of the sphenoid sinus target removal site and send a removal signal to the intelligent navigation system. In response to the removal signal, the intelligent navigation system automatically determines a new operation task, namely, the sellar floor bone target opening area calibration task. The fourth operation prompt information is used to prompt the physician for the next operation based on the current position of the endoscope and the sellar floor bone target opening area calibration task to be performed, such as prompting "forward to the upper left or forward to the lower left". The fourth operation prompt information will be calculated and updated in real time according to the current position of the endoscope to assist the physician in efficiently moving the endoscope to the appropriate position.

[0053] During the mirror movement process, the intelligent navigation system receives in real time the fourth white light image collected by the endoscope based on the fourth operation prompt information, calls the trained saddle bottom target detection model to recognize the saddle bottom in the fourth white light image, and calibrates the recognition result in the form of a saddle bottom rectangular frame to obtain second calibration data. Figure 7 As shown, the outermost rectangular frame is the saddle bottom rectangular frame.

[0054] The sellar floor includes the sellar floor bone and sellar floor dura mater. Under the sellar floor dura mater, there is a sellar floor posterior structure group, such as Figure 6 and Figure 7 As shown in the figure, the posterior sellar floor structure group includes multiple structures such as the pituitary gland, internal carotid artery, basilar artery, corpus cavernosum, optic nerve, and other blood vessels. Parts of the sellar floor bone and sellar floor dura mater need to be partially removed during the acquisition of pituitary tumor information. During this removal process, the target opening areas of the sellar floor bone and sellar floor dura mater need to be determined. To improve the safety of the posterior sellar floor structure group, these opening areas need to avoid high-risk structures in the posterior sellar floor structure group. Therefore, when calibrating these target opening areas, it is necessary to refer to the calibration data after the segmentation of the posterior sellar floor structure group.

[0055] This step is used to calibrate the target opening area of the sellar floor bone. Due to the poor transparency of the sellar floor bone, the endoscope cannot directly penetrate the sellar floor bone to obtain the segmentation of the sellar floor posterior structure group. Therefore, before the operation, this application first obtains the MRI coronal scan image of the target case through the pituitary gland, that is, Figure 6In the rightmost image, the sellar floor is identified and calibrated to obtain the third calibration data. At the same time, the structures behind the sellar floor are segmented and calibrated in the image to obtain the fourth calibration data. Since the second calibration data is the calibration of the sellar floor in the white light image, the third calibration data is the calibration of the sellar floor in the MRI coronal scan image through the pituitary gland, and the fourth calibration data is the calibration of the structures behind the sellar floor in the MRI coronal scan image through the pituitary gland, the third calibration data can be used as a transfer to transfer the segmentation and calibration of the structures behind the sellar floor in the fourth calibration data to the fourth white light image, so that the fourth white light image simultaneously has the second calibration data of the sellar floor and the fourth calibration data of the structure group behind the sellar floor. At this time, an initial area can be determined based on the second calibration data, and then the special areas that need to be avoided within the initial area can be determined based on the fourth calibration data. The two can be combined to obtain the sellar floor bone target opening area. The site removal operation is performed in this area, which can maximize safety and take into account the removal requirements of the pituitary tumor. The area is represented by a wireframe to obtain fifth calibration data, and the portion located in the sellar floor bone target opening area is the sellar floor bone target removal portion.

[0056] In one embodiment, before S1, it also includes: obtaining an MRI coronal scan image of the target case through the pituitary gland; calling a trained sellar floor target detection model to identify and calibrate the sellar floor in the MRI coronal scan image through the pituitary gland, and obtaining third calibration data, the third calibration data including a first sellar floor rectangular frame, a first centroid position of the first sellar floor rectangular frame, and a first area of the first sellar floor rectangular frame; calling a trained sellar floor posterior structure group segmentation model to segment and calibrate the sellar floor posterior structure group in the MRI coronal scan image through the pituitary gland, and obtaining fourth calibration data, the fourth calibration data including a segmentation boundary line and a centroid position of each structure in the sellar floor posterior structure group.

[0057] Before surgery, a transpituitary MRI coronal scan image of the target case is obtained. The trained sellar floor target detection model is then used to identify and calibrate the sellar floor in the image, obtaining the first sellar floor rectangular frame P1P2P3P4, the first centroid O1 position of the first sellar floor rectangular frame P1P2P3P4, and the first area s1 of the first sellar floor rectangular frame P1P2P3P4 as the third calibration data. Then, the trained sellar floor posterior structure group segmentation model is used to segment and calibrate the various structures posterior to the sellar floor in the image, obtaining the segmentation boundary lines and centroid positions of each structure as the fourth calibration data. Finally, these data are pre-stored in the intelligent navigation system. When the sellar floor bone target opening area calibration task needs to be performed, the intelligent navigation system can automatically load these calibration data and then automatically perform subsequent calculations and calibration work. In this way, efficiency can be further improved.

[0058] The sellar floor target detection model prefers the Yolov3 model. Before calling it, it is first trained based on the first training data set and the second training data set respectively. Specifically, a large number of historical neuroendoscopic white-light images are obtained, and the sellar floor rectangular frame range is calibrated on this type of image. Then, the model is trained as the first training data set. After the training is completed, the model has a high accuracy when calibrating the sellar floor rectangular frame in the white-light image. At the same time, a large number of historical transpituitary MRI coronal scan images are obtained, and the sellar floor rectangular frame range is calibrated on this type of image. Then, the model is trained as the second training data set. After the training is completed, the model has a high accuracy when calibrating the sellar floor rectangular frame in the MRI coronal scan image.

[0059] The Unet++ model is preferred for the posterior sellar floor structure group segmentation model. Before calling it, it is first trained based on the third training data set and the fourth training data set. Specifically, a large number of historical neuroendoscopic white-light images are obtained, and the boundaries of various posterior sellar floor structures are outlined in these images. Then, the model is trained as the third training data set. After the training is completed, the model has a high accuracy in calibrating the segmentation boundaries of various posterior sellar floor structures in white-light images. At the same time, a large number of historical transpituitary MRI coronal scan images are obtained, and the boundaries of various posterior sellar floor structures are outlined in these images. Then, the model is trained as the fourth training data set. After the training is completed, the model has a high accuracy in calibrating the segmentation boundaries of various posterior sellar floor structures in MRI coronal scan images.

[0060] In one embodiment, the second calibration data includes a second saddle bottom rectangular frame, a second centroid position of the second saddle bottom rectangular frame, and a second area of the second saddle bottom rectangular frame. The step of obtaining fifth calibration data of the saddle bottom bone target opening area based on the second calibration data, the third calibration data, and the fourth calibration data includes: determining the relative translation parameters of the first saddle bottom rectangular frame and the second saddle bottom rectangular frame based on the position deviation of the first centroid and the second centroid, and determining the first position parameters of the centroid of each structure in the posterior saddle bottom structure group in the fourth white light image based on the relative translation parameters; determining the relative scaling parameters of the first saddle bottom rectangular frame and the second saddle bottom rectangular frame based on the ratio of the first area to the second area, and determining the second position parameters of the segmentation boundary line of each structure in the posterior saddle bottom structure group in the fourth white light image based on the relative scaling parameters; and migrating the segmentation boundary line and centroid position of the posterior saddle bottom structure group to the fourth white light image based on the first position parameter and the second position parameter.

[0061] like Figure 7As shown, the second calibration data includes the second saddle bottom rectangular frame P5P6P7P8, the second centroid O2 position of the second saddle bottom rectangular frame P5P6P7P8 (not shown in the figure), and the second area s2 of the second saddle bottom rectangular frame P5P6P7P8 (not shown in the figure). The third calibration data stored in advance and called includes the first saddle bottom rectangular frame P1P2P3P4, the first centroid O1 position of the first saddle bottom rectangular frame P1P2P3P4, and the first area s1 of the first saddle bottom rectangular frame P1P2P3P4. The fourth calibration data includes the segmentation boundary line and centroid position of each structure behind the saddle bottom. Since the three calibration data are obtained based on two different types of images, when the calibration data are fused, the coordinate difference (△x, △y) between the first centroid O1 and the second centroid O2, that is, the relative translation parameter, is first calculated. Then, after the two centroids are aligned by translation, the new coordinates (x, y) of each structure in the MRI coronal scan image in the fourth white light image are calculated. i +△x,y i +△y) to obtain the first position parameter. Next, s1 / s2, also known as the relative scaling parameter, is calculated. The segmentation boundary lines of each structure in the MRI coronal scan image are scaled along their respective centroids by the ratio s1 / s2 to obtain the contours of each segmentation boundary in the fourth white light image, which is also known as the second position parameter. Finally, based on the first and second position parameters, the segmentation boundary lines and centroid positions of each structure in the MRI coronal scan image are pre-displayed in the fourth white light image. Through the above process, the automatic fusion of pre-stored calibration data and real-time calibration data is achieved.

[0062] In one embodiment, after the step of migrating the segmentation boundary line and centroid position of the sellar floor posterior structure group to the fourth white light image, the step further includes: acquiring from the fourth white light image the second sellar floor rectangular frame, the centroid position and lower segmentation boundary line of the optic nerve, the centroid position and segmentation boundary line of the pituitary gland, the centroid position and segmentation boundary line of the left carotid artery, and the centroid position and segmentation boundary line of the right carotid artery; obtaining the first outer expansion line of the lower segmentation boundary line of the optic nerve according to the centroid position of the optic nerve and the first outer expansion parameter; obtaining the first outer expansion arc of the segmentation boundary line of the left carotid artery according to the centroid position and the second outer expansion parameter, the first outer expansion arc intersecting with the first outer expansion line at a first intersection point and intersecting with the second sellar floor rectangular frame at a second intersection point; obtaining the first outer expansion arc of the segmentation boundary line of the right carotid artery according to the centroid position and the second outer expansion parameter. Two outward expansion arcs are obtained, and the second outward expansion arc intersects with the first outward expansion line at a third intersection point and intersects with the second sellar floor rectangular frame at a fourth intersection point; according to the centroid position of the pituitary gland and the third outward expansion parameter, the second outward expansion line of the lower segmentation boundary line of the pituitary gland is obtained, and the fifth and sixth intersection points of the horizontal line passing through the lowest point of the second outward expansion line and the second sellar floor rectangular frame are obtained; according to the first outward expansion arc, the second outward expansion arc, the first outward expansion line, the first intersection point and the third intersection point, the first upper calibration line is obtained; according to the second sellar floor rectangular frame, the second intersection point and the fifth intersection point, the left calibration line is obtained; according to the second sellar floor rectangular frame, the fourth intersection point and the sixth intersection point, the right calibration line is obtained; according to the horizontal line, the fifth intersection point and the sixth intersection point, the first lower calibration line is obtained; according to the first upper calibration line, the first lower calibration line, the left calibration line and the right calibration line, the fifth calibration data of the sellar floor bone target opening area is obtained.

[0063] like Figure 7 As shown, after the second calibration data and the fourth calibration data are displayed in the fourth white light image, the segmentation boundary line and centroid position of each structure can be obtained from the fourth white light image. The lower segmentation boundary of the optic nerve is shifted downward by △h based on the first expansion parameter to obtain a first expansion line. Based on the centroid O3 and the second expansion parameter, the segmentation boundary of the left internal carotid artery is expanded by △d1 to obtain a first expansion arc AB. The first expansion arc AB intersects with the first expansion line at a first intersection A and intersects with the second sellar floor rectangular frame P5P6P7P8 at a second intersection B. Based on the centroid O4 and the second expansion parameter, the segmentation boundary of the right internal carotid artery is expanded by △d1 to obtain a second expansion arc EF. The second expansion arc EF intersects with the first expansion line at a third intersection F and intersects with the second sellar floor rectangular frame P5P6P7P8 at a fourth intersection E. Based on the centroid position of the pituitary gland and the third expansion parameter, the lower segmentation boundary line of the pituitary gland is expanded by △d2 to obtain the second expansion line. A horizontal line passing through the lowest point of the second expansion line is drawn, which intersects with the second saddle bottom rectangular frame P5P6P7P8 at the fifth intersection C and the sixth intersection D.

[0064] The part of the first outward expansion line located between A and F forms the first upper calibration line AF, the part of the second saddle bottom rectangular frame P5P6P7P8 located between B and C forms the left calibration line BC, the part of the second saddle bottom rectangular frame P5P6P7P8 located between E and D forms the right calibration line ED, and the part of the horizontal line located between C and D forms the first lower calibration line CD. The first upper calibration line AF, the first outward expansion arc AB, the left calibration line BC, the first lower calibration line CD, the right calibration line ED, and the second outward expansion arc EF are connected in sequence to form the wireframe, which is the fifth calibration data of the sellar floor bone target opening area.

[0065] The target opening area of the sellar floor bone obtained by the above method can completely avoid the optic nerve and the carotid artery during opening, leaving a safe distance, while completely covering the pituitary gland, thus taking into account both safety and surgical operability.

[0066] In one embodiment, after the step of obtaining the fifth calibration data of the sellar floor bone target opening area according to the first upper calibration line, the first lower calibration line, the left calibration line and the right calibration line, it also includes: obtaining the center line of the pituitary stalk from the fourth white light image, and extending the center line to obtain a center extension line located below the first lower calibration line; according to the center extension line and the fourth expansion parameter, obtaining a third expansion line and a fourth expansion line located on both sides of the center extension line, and according to the third expansion line and the fourth expansion line, obtaining the calibration data of the arterial risk area.

[0067] like Figure 7 As shown, in the fourth white light image, the center line of the pituitary stalk is found by the Zhang-Suen thinning algorithm and extended to below the first lower calibration line CD to form the center extension line of the pituitary stalk. Based on the fourth outward expansion parameter, the center extension line is expanded in both directions by △w. The area within this range is the high-risk area for the artery, that is, the area indicated by the arrow of the basilar artery in the figure. Calibration can serve as a warning to physicians and reduce the probability of accidental injury to the artery.

[0068] S5: In response to the removal signal of the sellar floor bone in the sellar floor bone target opening area, determine the sellar floor dura mater target opening area calibration task and generate the fifth operation prompt information; obtain the fifth white light image collected by the endoscope based on the fifth operation prompt information, call the trained sellar floor dura mater target detection model to identify and calibrate the sellar floor dura mater in the fifth white light image, and obtain the sixth calibration data, call the trained sellar floor posterior structure group segmentation model to segment and calibrate the sellar floor posterior structure group in the fifth white light image, and obtain the seventh calibration data; according to the sixth calibration data and the seventh calibration data, obtain the eighth calibration data of the sellar floor dura mater target opening area.

[0069] When the doctor removes the sellar floor bone target removal part in the sellar floor bone target opening area based on the fifth calibration data, Figure 8 As shown, the sellar dura mater will be exposed. The physician can feedback the removal results of the sellar dura mater target removal site to the intelligent navigation system and send a removal signal to the intelligent navigation system. In response to the removal signal, the intelligent navigation system automatically determines a new operation task, namely, the sellar dura mater target opening area calibration task. The fifth operation prompt information is used to prompt the physician for the next operation based on the current position of the endoscope and the sellar dura mater target opening area calibration task to be performed, such as prompting "forward to the upper left or forward to the lower left". The fifth operation prompt information will be calculated and updated in real time according to the current position of the endoscope to assist the physician in efficiently moving the endoscope to the appropriate position.

[0070] During the mirror movement process, the intelligent navigation system receives the fifth white light image collected by the endoscope based on the fifth operation prompt information in real time, calls the trained sellar dura mater target detection model to identify the sellar dura mater target open area in the fifth white light image, and calibrates the recognition result in the form of a sellar dura mater rectangular frame to obtain the sixth calibration data. Figure 9 As shown in the figure, the outermost rectangle is the sellar dura mater rectangle Q1Q2Q3Q4. The sellar dura mater target detection model preferably uses the Yolov3 model. Before calling it, a large number of historical neuroendoscopic white-light images are first obtained to calibrate the sellar dura mater rectangle range. This is then used as a training dataset to train the model. After training, the model has high accuracy in calibrating the sellar dura mater rectangle.

[0071] Similarly, to improve the safety of the posterior sellar structure group, the target opening area of the sellar dura mater needs to avoid high-risk structures within the group. Therefore, when calibrating the target opening area of the sellar dura mater, reference is made to the calibration data after segmentation of the posterior sellar structure group. In this step, since the sellar dura mater is translucent, the trained posterior sellar structure group segmentation model can be directly used to segment and calibrate the posterior sellar structure group in the fifth white light image, obtaining seventh calibration data, which is more accurate than the pre-stored fourth calibration data. Next, an initial region is determined based on the sixth calibration data, and then specific areas to be avoided within this initial region are determined based on the seventh calibration data. Combining these two data, the sellar dura mater target opening area is obtained. Performing the site removal operation within this area maximizes safety while meeting the requirements for pituitary tumor removal. This area is represented as a wireframe, obtaining eighth calibration data. The site located within the sellar dura mater target opening area is the sellar dura mater target removal site.

[0072] In one embodiment, S5 specifically includes: obtaining a sellar dura mater rectangular frame, a centroid position and a segmentation boundary line of the left carotid artery, and a centroid position and a segmentation boundary line of the right carotid artery from the fifth white light image; obtaining a third outward expansion arc of the segmentation boundary line of the left carotid artery based on the centroid position of the left carotid artery and the fifth outward expansion parameter, the third outward expansion arc intersecting with the sellar dura mater rectangular frame at a seventh intersection point and an eighth intersection point; obtaining a fourth outward expansion arc of the segmentation boundary line of the right carotid artery based on the centroid position of the right carotid artery and the fifth outward expansion parameter, the fourth outward expansion arc intersecting with the sellar dura mater rectangular frame at a ninth intersection point and a tenth intersection point; obtaining a second upper calibration line based on the sellar dura mater rectangular frame, the seventh intersection point, and the ninth intersection point, and obtaining a second lower calibration line based on the sellar dura mater rectangular frame, the eighth intersection point, and the tenth intersection point; obtaining eighth calibration data of the sellar dura mater target opening area based on the second upper calibration line, the third outward expansion arc, the second lower calibration line, and the fourth outward expansion arc.

[0073] like Figure 9 As shown, in the fifth white-light image, the sixth calibration data includes the sellar dura mater rectangular frame Q1Q2Q3Q4, and the seventh calibration data includes the segmentation boundary lines and centroid positions of each structure in the sellar posterior structure group. Based on the centroid position of the left carotid artery and the fifth expansion parameter, the segmentation boundary line of the left carotid artery is expanded by Δd3 to form a third expansion arc A1B1. The third expansion arc A1B1 intersects with the sellar dura mater rectangular frame Q1Q2Q3Q4 at the seventh intersection A1 and the eighth intersection B1. Based on the centroid position of the right carotid artery and the fifth expansion parameter, the segmentation boundary line of the right carotid artery is expanded by Δd3 to form a fourth expansion arc C1D1. The fourth expansion arc C1D1 intersects with the sellar dura mater rectangular frame Q1Q2Q3Q4 at the ninth intersection D1 and the tenth intersection C1.

[0074] The part of the rectangular frame Q1Q2Q3Q4 of the saddle dura mater located between A1 and D1 forms the second upper calibration line A1D1, and the part of the rectangular frame Q1Q2Q3Q4 of the saddle dura mater located between B1 and C1 forms the second lower calibration line B1C1. The second upper calibration line A1D1, the third outward expansion arc A1B1, the second lower calibration line B1C1, and the fourth outward expansion arc C1D1 are connected in sequence to form the eighth calibration data of the target opening area of the saddle dura mater.

[0075] The target dura mater opening area obtained through this method completely avoids the carotid artery during opening, maintaining a safe distance, thus balancing safety and operability. During the opening process, the segmented blood vessels within the target dura mater opening area are displayed as dotted lines to remind the physician to treat the vessels with caution.

[0076] S6: In response to the removal signal of the sellar dura mater in the target opening area of the sellar dura mater, determine the pituitary tumor information acquisition task and generate sixth operation prompt information; obtain the sixth white light image collected by the endoscope based on the sixth operation prompt information, call the trained tumor segmentation model to segment the pituitary tumor in the sixth white light image, and obtain the pituitary tumor information according to the segmentation result.

[0077] After the physician removes the sellar floor dura mater target removal site within the sellar floor dura mater target opening area based on the eighth calibration data, Figure 10 As shown, the pituitary gland and pituitary tumor will be exposed. The physician can feedback the removal results of the target removal site of the sellar floor dura mater to the intelligent navigation system and give the intelligent navigation system a removal signal. The intelligent navigation system responds to the removal signal and automatically determines the new operation task, that is, the pituitary tumor information acquisition task. The sixth operation prompt information is used to prompt the physician for the next operation based on the current position of the endoscope and the pituitary tumor information acquisition task to be performed, such as prompting "forward to the upper left or forward to the lower left", etc. The sixth operation prompt information will be calculated and updated in real time according to the current position of the endoscope to assist the physician in efficiently moving the endoscope to the appropriate position.

[0078] During the camera movement process, the intelligent navigation system receives, in real time, the sixth white-light image captured by the endoscope based on the sixth operation prompt information. It then uses the trained tumor segmentation model to segment the pituitary tumor in the sixth white-light image and calibrate the segmentation boundaries to obtain the segmentation result. The tumor segmentation model preferably uses the Unet++ model. Prior to invocation, a large number of historical neuroendoscopic white-light images are acquired to delineate the boundaries of various tumor types. This data set is then used to train the model. After training, the model demonstrates high accuracy in identifying and segmenting tumors.

[0079] According to the above segmentation results, the pituitary tumor information is obtained, and the doctor can determine the next plan based on this information. In each of the above steps of obtaining pituitary tumor information through the nasal cavity, the parts and regions are automatically identified and calibrated based on the trained model, so the accuracy of part identification is high, and the accuracy, rationality and safety of regional calibration are also improved. In addition, each step automatically determines the next operation task based on the recognition result and calibration content, and generates the next operation prompt information based on the operation task, so that the degree of dependence on the doctor's experience is reduced and the efficiency is improved. Therefore, this application can realize real-time intelligent navigation throughout the process, reduce the difficulty of information acquisition, and facilitate the rapid and accurate acquisition of pituitary tumor information, as well as the determination of subsequent plans based on this information.

[0080] In one embodiment, after S7, it also includes: obtaining a third area of the pituitary tumor based on the pituitary tumor information, and obtaining a fourth area of the target opening area of the sellar dura mater based on the eighth calibration data; judging whether the ratio of the third area to the fourth area is greater than a preset ratio threshold; if not, generating overall removal prompt information of the pituitary tumor; if so, generating block removal prompt information of the pituitary tumor, and in response to the i-th processing result of the block removal prompt information, cyclically executing the operation of obtaining the third area of the remaining pituitary tumor, the operation of judging the area ratio and the preset ratio threshold, and the operation of generating the removal prompt information until the judgment result is no.

[0081] After obtaining the pituitary tumor information, if it is confirmed that the pituitary tumor can be removed, a third area s3 of the pituitary tumor can be obtained based on the pituitary tumor information, and a fourth area s4 of the target opening area of the sellar floor dura mater can be obtained based on the eighth calibration data. This fourth area s4 is also the area of the area to be subsequently opened. It is determined whether the ratio s3 / s4 of the two is greater than a preset ratio threshold τ. If it is not greater than, it indicates that the pituitary tumor is relatively small relative to the current opening area and the pituitary tumor can be removed as a whole. At this time, a prompt message for the removal of the pituitary tumor as a whole is generated. If it is greater than, it indicates that the pituitary tumor is relatively large relative to the current opening area and needs to be removed in blocks. At this time, a prompt message for the removal of the pituitary tumor in blocks is generated.

[0082] Before the pituitary tumor is removed, this application will automatically calculate prompt information for overall removal or piecemeal removal, providing effective guidance for the removal of the pituitary tumor.

[0083] After generating a prompt message for the removal of a pituitary tumor, the physician will provide feedback to the intelligent navigation system after each block is removed. In response to the i-th processing result, the intelligent navigation system loops through the image acquisition and tumor segmentation steps described above, obtaining a third area s3 of the remaining pituitary tumor and repeating the operation of determining the size of the ratio s3 / s4 relative to a preset ratio threshold τ. If the determination result is not greater than, a prompt message for the removal of the entire pituitary tumor is generated. If the determination result is greater than, a prompt message for the removal of the entire pituitary tumor is still generated. After removing a block, the system enters the next loop until the determination result is not greater than, generating a prompt message for the removal of the entire pituitary tumor.

[0084] Through the above-mentioned method, the present application always reminds the physician to remove the pituitary tumor in pieces when the area is large. When the area of the pituitary tumor is reduced to a certain extent, the physician is reminded that the tumor can be removed as a whole. This can ensure the safety of each removal and improve the removal efficiency.

[0085] In one embodiment, after the step of generating overall removal prompt information of the pituitary tumor, it also includes: in response to the removal signal of the pituitary tumor, determining the invasive tumor identification task, and generating seventh operation prompt information; obtaining a seventh white light image collected by the endoscope based on the seventh operation prompt information, calling the trained tumor invasion identification model to identify the invasive tumor in the seventh white light image, and obtaining an identification result; when the identification result is that there is no invasive tumor, generating no invasion prompt information; when the identification result is that there is an invasive tumor, determining the invasive tissue tumor removal risk assessment task according to the identification result, and generating eighth operation prompt information.

[0086] After the physician removes the pituitary tumor based on the overall removal prompt information of the pituitary tumor, the physician can feedback the pituitary tumor removal result to the intelligent navigation system, giving the intelligent navigation system a removal signal. The intelligent navigation system responds to the removal signal and automatically determines a new operation task, namely the invasive tumor identification task. The seventh operation prompt information is used to prompt the physician for the next operation based on the current position of the endoscope and the invasive tumor identification task to be performed, such as prompting "Look for the invasive tumor in front to the upper left or in front to the lower left". The seventh operation prompt information will be calculated and updated in real time according to the current position of the endoscope to assist the physician in efficiently moving the endoscope to the appropriate position.

[0087] During the mirror movement process, the intelligent navigation system receives in real time the seventh white light image collected by the endoscope based on the seventh operation prompt information, calls the trained tumor invasion recognition model to identify the invasive tumor in the seventh white light, and obtains the recognition result. When it is recognized that there is no invasion, a non-invasion prompt information is generated and the operation can be ended. When it is recognized that there is an invasive tumor, a new operation task is automatically determined, that is, the invasive tumor removal risk assessment task. The eighth operation prompt information is used to prompt the physician's next operation based on the current position of the endoscope and the invasive tumor removal risk assessment task to be performed, such as prompting "go forward to the upper left or forward to the lower left to assess the risk of invasive tumor removal". The eighth operation prompt information will be calculated and updated in real time according to the current position of the endoscope to assist the physician in efficiently moving the mirror to the appropriate position. In this way, the invasion of pituitary tumors can be identified, which is beneficial to the physician's next decision.

[0088] In one embodiment, after the step of generating the eighth operation prompt information, it also includes: obtaining an eighth white light image collected by the endoscope based on the eighth operation prompt information; calling the trained invasive tumor removal risk assessment model to assess the invasive tumor removal risk in the eighth white light image to obtain an assessment result; when the assessment result is that the risk is lower than the threshold, generating invasive tumor removal prompt information; when the assessment result is that the risk is higher than the threshold, generating do not remove prompt information.

[0089] During the mirror movement process, the intelligent navigation system receives the eighth white light image collected by the endoscope based on the eighth operation prompt information in real time, calls the trained risk assessment model to assess the risk of removing the invading tumor in the eighth white light image, and obtains the assessment result. If the assessment result is low risk, the invading tumor removal prompt information is generated, then Figure 11 As shown, based on this information, the physician can remove the invading tumor. During the removal process, similar procedures can be performed as during pituitary tumor removal, determining whether to remove the tumor in sections based on area. If the assessment result is high risk, a "Do Not Remove" message is generated, informing the physician that removal is not currently possible. This approach allows for more accurate risk assessment of the invading tumor, ensuring surgical safety.

[0090] It can be seen from the above embodiments that the intelligent navigation system of the present application can provide effective reference for doctors, whether in the process of obtaining pituitary tumor information or in the process of removing pituitary tumors, to assist doctors in completing tasks quickly and efficiently.

[0091] Based on the method described in the above embodiment, this embodiment will be further described from the perspective of the pituitary tumor information acquisition device. Figure 12 , the pituitary tumor information acquisition device may include:

[0092] A first acquisition module 10 is configured to acquire a first white-light image captured by the endoscope within the nasal cavity of a target case based on first operation prompt information, wherein the first operation prompt information is generated based on a superior turbinate search task, call a trained superior turbinate recognition model to identify the superior turbinate in the first white-light image to obtain first recognition information, determine a sphenoid sinus opening search task based on the first recognition information, and generate second operation prompt information;

[0093] a second acquisition module 20 configured to acquire a second white-light image acquired by the endoscope based on the second operation prompt information, call the trained sphenoid sinus opening recognition model to recognize the sphenoid sinus opening in the second white-light image to obtain second recognition information, determine a sphenoid sinus target opening area calibration task based on the second recognition information, and generate third operation prompt information;

[0094] a third acquisition module 30, configured to acquire a third white light image acquired by the endoscope based on the third operation prompt information, and call the trained sphenoid sinus target opening region segmentation model to segment and calibrate the sphenoid sinus target opening region in the third white light image to obtain first calibration data;

[0095] The first obtaining module 40 is configured to determine a sellar floor bone target opening area calibration task in response to a sphenoid sinus removal signal within the sphenoid sinus target opening area, and generate fourth operation prompt information; obtain a fourth white light image acquired by the endoscope based on the fourth operation prompt information, call a trained sellar floor target detection model to identify and calibrate the sellar floor in the fourth white light image to obtain second calibration data, and call pre-stored third calibration data of the sellar floor and fourth calibration data of the sellar floor posterior structure group, the third calibration data and the fourth labeling data being obtained based on a trans-pituitarism MRI coronal scan image of the target case; and obtain fifth calibration data of the sellar floor bone target opening area based on the second calibration data, the third calibration data, and the fourth calibration data;

[0096] a second obtaining module 50 for determining a sellar floor dura mater target opening area calibration task in response to a sellar floor bone removal signal in the sellar floor bone target opening area, and generating fifth operation prompt information; obtaining a fifth white light image acquired by the endoscope based on the fifth operation prompt information, calling a trained sellar floor dura mater target detection model to identify and calibrate the sellar floor dura mater in the fifth white light image to obtain sixth calibration data, calling a trained sellar floor posterior structure group segmentation model to segment and calibrate the sellar floor posterior structure group in the fifth white light image to obtain seventh calibration data, and obtaining eighth calibration data of the sellar floor dura mater target opening area based on the sixth calibration data and the seventh calibration data;

[0097] The third obtaining module 60 is used to determine the pituitary tumor information acquisition task in response to the removal signal of the sellar dura mater in the sellar dura mater target opening area, and generate sixth operation prompt information; obtain the sixth white light image collected by the endoscope based on the sixth operation prompt information, call the trained tumor segmentation model to segment the pituitary tumor in the sixth white light image, and obtain the pituitary tumor information according to the segmentation result.

[0098] In one embodiment, the pituitary tumor information acquisition device further includes:

[0099] A fourth acquisition module is used to acquire a trans-pituitarism MRI coronal scan image of the target case;

[0100] a fourth obtaining module, configured to call the trained sellar floor target detection model to identify and calibrate the sellar floor in the trans-pituitary MRI coronal scan image to obtain the third calibration data, wherein the third calibration data includes a first sellar floor rectangular frame, a first centroid position of the first sellar floor rectangular frame, and a first area of the first sellar floor rectangular frame;

[0101] The fifth obtaining module is used to call the trained sellar floor posterior structure group segmentation model to segment and calibrate the sellar floor posterior structure group in the trans-pituitary MRI coronal scan image to obtain the fourth calibration data, and the fourth calibration data includes the segmentation boundary line and centroid position of each structure in the sellar floor posterior structure group.

[0102] In one embodiment, the second calibration data includes a second saddle bottom rectangular frame, a second centroid position of the second saddle bottom rectangular frame, and a second area of the second saddle bottom rectangular frame, and the first obtaining module 40 includes:

[0103] a first determining submodule, configured to determine relative translation parameters of the first saddle bottom rectangular frame and the second saddle bottom rectangular frame based on a positional deviation between the first centroid and the second centroid, and determine first position parameters of the centroid of each structure in the saddle bottom rear structure group in the fourth white light image based on the relative translation parameters;

[0104] a second determining submodule, configured to determine relative scaling parameters of the first saddle bottom rectangular frame and the second saddle bottom rectangular frame according to a ratio of the first area to the second area, and determine second position parameters of segmentation boundaries of each structure in the posterior saddle bottom structure group in the fourth white light image according to the relative scaling parameters;

[0105] A migration submodule is configured to migrate the segmentation boundary line and the centroid position of the posterior sellar floor structure group to the fourth white light image according to the first position parameter and the second position parameter.

[0106] In one embodiment, the first obtaining module 40 further includes:

[0107] a first acquisition submodule, configured to acquire, from the fourth white light image, a second sellar floor rectangular frame, a centroid position and a lower segmentation boundary line of the optic nerve, a centroid position and a segmentation boundary line of the pituitary gland, a centroid position and a segmentation boundary line of the left carotid artery, and a centroid position and a segmentation boundary line of the right carotid artery;

[0108] a first obtaining submodule, configured to obtain a first expansion line of a lower segmentation boundary line of the optic nerve according to the centroid position and the first expansion parameter of the optic nerve; obtain a first expansion arc of the segmentation boundary line of the left carotid artery according to the centroid position and the second expansion parameter, wherein the first expansion arc intersects with the first expansion line at a first intersection point and intersects with the second sellar floor rectangular frame at a second intersection point; obtain a second expansion arc of the segmentation boundary line of the right carotid artery according to the centroid position and the second expansion parameter, wherein the second expansion arc intersects with the first expansion line at a third intersection point and intersects with the second sellar floor rectangular frame at a fourth intersection point; obtain a second expansion line of the lower segmentation boundary line of the pituitary gland according to the centroid position and the third expansion parameter of the pituitary gland, and obtain a fifth intersection point and a sixth intersection point of a horizontal line passing through the lowest point of the second expansion line and the second sellar floor rectangular frame;

[0109] a second obtaining submodule, configured to obtain a first upper calibration line according to the first outward expansion arc, the second outward expansion arc, the first outward expansion line, the first intersection point, and the third intersection point; obtain a left calibration line according to the second saddle bottom rectangular frame, the second intersection point, and the fifth intersection point; obtain a right calibration line according to the second saddle bottom rectangular frame, the fourth intersection point, and the sixth intersection point; and obtain a first lower calibration line according to the horizontal line, the fifth intersection point, and the sixth intersection point;

[0110] The third obtaining submodule is configured to obtain fifth calibration data of the sellar floor bone target opening area according to the first upper calibration line, the first lower calibration line, the left calibration line, and the right calibration line.

[0111] In one embodiment, the first obtaining module 40 further includes:

[0112] a fourth obtaining submodule, configured to obtain a center line of the pituitary stalk from the fourth white light image, and extend the center line to obtain a center extension line located below the first lower calibration line;

[0113] The fifth obtaining submodule is used to obtain a third expansion line and a fourth expansion line located on both sides of the central extension line according to the central extension line and the fourth expansion parameter, and obtain calibration data of the arterial risk area according to the third expansion line and the fourth expansion line.

[0114] In one embodiment, the sixth calibration data includes a rectangular frame of the sellar floor dura mater, the seventh calibration data includes a segmentation boundary line and a centroid position of each structure in the sellar floor posterior structure group, and the second obtaining module 50 includes:

[0115] a second acquisition submodule, configured to acquire, from the fifth white light image, a rectangular frame of the sellar floor dura mater, a centroid position and a segmentation boundary line of the left carotid artery, and a centroid position and a segmentation boundary line of the right carotid artery;

[0116] a sixth obtaining submodule, configured to obtain, based on the centroid position of the left carotid artery and the fifth expansion parameter, a third expansion arc for the segmentation boundary line of the left carotid artery, wherein the third expansion arc intersects with the rectangular frame of the sellar floor dura mater at a seventh intersection point and an eighth intersection point; and obtain, based on the centroid position of the right carotid artery and the fifth expansion parameter, a fourth expansion arc for the segmentation boundary line of the right carotid artery, wherein the fourth expansion arc intersects with the rectangular frame of the sellar floor dura mater at a ninth intersection point and a tenth intersection point.

[0117] a seventh obtaining submodule, configured to obtain a second upper calibration line according to the sella serrata dura mater rectangular frame, the seventh intersection point, and the ninth intersection point, and to obtain a second lower calibration line according to the sella serrata dura mater rectangular frame, the eighth intersection point, and the tenth intersection point;

[0118] The eighth obtaining submodule is configured to obtain eighth calibration data of the sellar floor dura mater target opening area according to the second upper calibration line, the third outward expansion arc, the second lower calibration line, and the fourth outward expansion arc.

[0119] In one embodiment, the pituitary tumor information acquisition device further includes:

[0120] a sixth obtaining module, configured to obtain a third area of the pituitary tumor according to the pituitary tumor information, and obtain a fourth area of the target opening region of the sellar floor dura mater according to the eighth calibration data;

[0121] a first determining module, configured to determine whether a ratio of the third area to the fourth area is greater than a preset ratio threshold;

[0122] A first generating module is configured to generate a prompt message for overall removal of the pituitary tumor if no;

[0123] The second generation module is used to generate a block removal prompt information for the pituitary tumor, and in response to the i-th processing result of the block removal prompt information, cyclically execute the operation of obtaining the third area of the remaining pituitary tumor, the operation of judging the area ratio and the preset ratio threshold, and the operation of generating the removal prompt information until the judgment result is no.

[0124] In one embodiment, the pituitary tumor information acquisition device further includes:

[0125] a third generating module, configured to determine an invading tumor identification task in response to the pituitary tumor removal signal, and generate seventh operation prompt information;

[0126] a seventh obtaining module, configured to obtain a seventh white light image collected by the endoscope based on the seventh operation prompt information, call the trained tumor invasion recognition model to identify the invading tumor in the seventh white light image, and obtain a recognition result;

[0127] A fourth generating module is configured to generate non-invasion prompt information when the recognition result is that there is no invasion of the tumor;

[0128] The fifth generating module is configured to determine a risk assessment task for removing the invaded tissue tumor according to the recognition result when the recognition result indicates that an invaded tumor exists, and to generate eighth operation prompt information.

[0129] In one embodiment, the pituitary tumor information acquisition device further includes:

[0130] a fifth acquisition module, configured to acquire an eighth white light image acquired by the endoscope based on the eighth operation prompt information;

[0131] an eighth obtaining module, configured to call the trained invasion tumor removal risk assessment model to assess the invasion tumor removal risk in the eighth white light image to obtain an assessment result;

[0132] A sixth generating module, configured to generate a prompt message for removing the invading tumor when the assessment result indicates that the risk is lower than a threshold;

[0133] The seventh generating module is used to generate a do not remove prompt message when the assessment result shows that the risk is higher than the threshold.

[0134] Different from the existing technology, the pituitary tumor information acquisition device provided by the present application automatically identifies and calibrates the parts and regions based on the trained model in each step of obtaining pituitary tumor information through the nasal cavity. The accuracy of part identification is high, and the accuracy, rationality and safety of regional calibration are also improved. In addition, each step automatically determines the next operation task based on the recognition result and calibration content, and generates the next operation prompt information based on the operation task, so that the dependence on the physician's experience is reduced and the efficiency is improved. Therefore, the present application can realize real-time intelligent navigation throughout the process, reduce the difficulty of information acquisition, and facilitate the rapid and accurate acquisition of pituitary tumor information, as well as the determination of subsequent plans based on the information.

[0135] Accordingly, the embodiment of the present application further provides an electronic device, such as Figure 13As shown, the electronic device may include components such as a radio frequency (RF) circuit 1001, a memory 1002 including one or more computer-readable storage media, an input unit 1003, a display unit 1004, a sensor 1005, an audio circuit 1006, a WiFi module 1007, a processor 1008 including one or more processing cores, and a power supply 1009. It will be understood by those skilled in the art that Figure 13 The electronic device structure shown in the figure does not constitute a limitation of the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange components differently.

[0136] Radio frequency circuit 1001 can be used to receive and transmit signals during information transmission or calls. Specifically, it receives downlink information from the base station and transmits it to one or more processors 1008 for processing. It also transmits uplink data to the base station. Memory 1002 can be used to store software programs and modules. Processor 1008 executes the software programs and modules stored in memory 1002 to perform various functional applications and obtain pituitary tumor information. Input unit 1003 can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical, or trackball signal input related to user settings and function control.

[0137] The display unit 1004 may be used to display information input by a client or information provided to a client, as well as various graphical client interfaces of the server. These graphical client interfaces may be composed of graphics, text, icons, videos, or any combination thereof.

[0138] The electronic device may further include at least one sensor 1005, such as a light sensor, a motion sensor, or other sensors. The audio circuit 1006 may include a speaker, which may provide an audio interface between the user and the electronic device.

[0139] WiFi is a short-range wireless transmission technology. Electronic devices can help customers send and receive emails, browse web pages, and follow up streaming media through WiFi module 1007. It provides customers with wireless broadband Internet follow-up. Figure 13 A WiFi module 1007 is shown, but it is understandable that it is not an essential component of the electronic device and can be omitted as needed without changing the essence of the application.

[0140] The processor 1008 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire mobile phone. By running or executing software programs and / or modules stored in the memory 1002 and calling data stored in the memory 1002, it performs various functions of the electronic device and processes data, thereby monitoring the mobile phone as a whole.

[0141] The electronic device also includes a power supply 1009 (such as a battery) for supplying power to various components. Preferably, the power supply can be logically connected to the processor 1008 through a power management system, thereby managing charging, discharging, and power consumption through the power management system.

[0142] Although not shown, the electronic device may also include a camera, a Bluetooth module, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 1008 in the server will load the executable files corresponding to the processes of one or more application programs into the memory 1002 according to the following instructions, and the processor 1008 will run the application programs stored in the memory 1002, thereby achieving the following functions:

[0143] Obtaining a first white-light image captured by the endoscope within the nasal cavity of the target case based on first operation prompt information, the first operation prompt information being generated based on a superior turbinate search task, calling a trained superior turbinate recognition model to identify the superior turbinate in the first white-light image to obtain first recognition information, determining a sphenoid sinus opening search task based on the first recognition information, and generating second operation prompt information;

[0144] obtaining a second white-light image acquired by the endoscope based on the second operation prompt information, calling the trained sphenoid sinus opening recognition model to recognize the sphenoid sinus opening in the second white-light image to obtain second recognition information, determining a sphenoid sinus target opening area calibration task based on the second recognition information, and generating third operation prompt information;

[0145] acquiring a third white light image acquired by the endoscope based on the third operation prompt information, and calling the trained sphenoid sinus target opening region segmentation model to segment and calibrate the sphenoid sinus target opening region in the third white light image to obtain first calibration data;

[0146] In response to the sphenoid sinus removal signal in the sphenoid sinus target opening area, a sellar floor bone target opening area calibration task is determined, and fourth operation prompt information is generated; a fourth white light image acquired by the endoscope based on the fourth operation prompt information is acquired, a trained sellar floor target detection model is called to identify and calibrate the sellar floor in the fourth white light image to obtain second calibration data, and pre-stored third calibration data of the sellar floor and fourth calibration data of the sellar floor posterior structure group are called, the third calibration data and the fourth labeling data are obtained based on the MRI coronal scan image of the target case through the pituitary gland; fifth calibration data of the sellar floor bone target opening area is obtained according to the second calibration data, the third calibration data and the fourth calibration data;

[0147] In response to the sellar floor bone removal signal in the sellar floor bone target opening area, a sellar floor dura mater target opening area calibration task is determined, and fifth operation prompt information is generated; a fifth white light image acquired by the endoscope based on the fifth operation prompt information is acquired, a trained sellar floor dura mater target detection model is called to identify and calibrate the sellar floor dura mater in the fifth white light image to obtain sixth calibration data, a trained sellar floor posterior structure group segmentation model is called to segment and calibrate the sellar floor posterior structure group in the fifth white light image to obtain seventh calibration data, and eighth calibration data of the sellar floor dura mater target opening area is obtained based on the sixth calibration data and the seventh calibration data;

[0148] In response to the removal signal of the sellar dura mater in the sellar dura mater target opening area, the pituitary tumor information acquisition task is determined and the sixth operation prompt information is generated; the sixth white light image collected by the endoscope based on the sixth operation prompt information is obtained, the trained tumor segmentation model is called to segment the pituitary tumor in the sixth white light image, and the pituitary tumor information is obtained according to the segmentation result.

[0149] In the above embodiments, the description of each embodiment has its own focus. For the part that is not described in detail in a certain embodiment, please refer to the detailed description above and will not be repeated here.

[0150] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be accomplished by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0151] To this end, an embodiment of the present application provides a computer-readable storage medium storing a plurality of instructions, which can be loaded by a processor to implement the following functions:

[0152] Obtaining a first white-light image captured by the endoscope within the nasal cavity of the target case based on first operation prompt information, the first operation prompt information being generated based on a superior turbinate search task, calling a trained superior turbinate recognition model to identify the superior turbinate in the first white-light image to obtain first recognition information, determining a sphenoid sinus opening search task based on the first recognition information, and generating second operation prompt information;

[0153] obtaining a second white-light image acquired by the endoscope based on the second operation prompt information, calling the trained sphenoid sinus opening recognition model to recognize the sphenoid sinus opening in the second white-light image to obtain second recognition information, determining a sphenoid sinus target opening area calibration task based on the second recognition information, and generating third operation prompt information;

[0154] acquiring a third white light image acquired by the endoscope based on the third operation prompt information, and calling the trained sphenoid sinus target opening region segmentation model to segment and calibrate the sphenoid sinus target opening region in the third white light image to obtain first calibration data;

[0155] In response to the sphenoid sinus removal signal in the sphenoid sinus target opening area, a sellar floor bone target opening area calibration task is determined, and fourth operation prompt information is generated; a fourth white light image acquired by the endoscope based on the fourth operation prompt information is acquired, a trained sellar floor target detection model is called to identify and calibrate the sellar floor in the fourth white light image to obtain second calibration data, and pre-stored third calibration data of the sellar floor and fourth calibration data of the sellar floor posterior structure group are called, the third calibration data and the fourth labeling data are obtained based on the MRI coronal scan image of the target case through the pituitary gland; fifth calibration data of the sellar floor bone target opening area is obtained according to the second calibration data, the third calibration data and the fourth calibration data;

[0156] In response to the sellar floor bone removal signal in the sellar floor bone target opening area, a sellar floor dura mater target opening area calibration task is determined, and fifth operation prompt information is generated; a fifth white light image acquired by the endoscope based on the fifth operation prompt information is acquired, a trained sellar floor dura mater target detection model is called to identify and calibrate the sellar floor dura mater in the fifth white light image to obtain sixth calibration data, a trained sellar floor posterior structure group segmentation model is called to segment and calibrate the sellar floor posterior structure group in the fifth white light image to obtain seventh calibration data, and eighth calibration data of the sellar floor dura mater target opening area is obtained based on the sixth calibration data and the seventh calibration data;

[0157] In response to the removal signal of the sellar dura mater in the sellar dura mater target opening area, the pituitary tumor information acquisition task is determined and the sixth operation prompt information is generated; the sixth white light image collected by the endoscope based on the sixth operation prompt information is obtained, the trained tumor segmentation model is called to segment the pituitary tumor in the sixth white light image, and the pituitary tumor information is obtained according to the segmentation result.

[0158] The above is a detailed introduction to a pituitary tumor information acquisition method, device, electronic device and computer-readable storage medium provided in the embodiments of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the technical solutions and core ideas of the present application; ordinary technicians in this field should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or replace some of the technical features therein with equivalents; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for obtaining pituitary tumor information, characterized in that: include: Obtaining a first white-light image captured by the endoscope within the nasal cavity of the target case based on first operation prompt information, the first operation prompt information being generated based on a superior turbinate search task, calling a trained superior turbinate recognition model to identify the superior turbinate in the first white-light image to obtain first recognition information, determining a sphenoid sinus opening search task based on the first recognition information, and generating second operation prompt information; obtaining a second white-light image acquired by the endoscope based on the second operation prompt information, calling the trained sphenoid sinus opening recognition model to recognize the sphenoid sinus opening in the second white-light image to obtain second recognition information, determining a sphenoid sinus target opening area calibration task based on the second recognition information, and generating third operation prompt information; acquiring a third white light image acquired by the endoscope based on the third operation prompt information, and calling the trained sphenoid sinus target opening region segmentation model to segment and calibrate the sphenoid sinus target opening region in the third white light image to obtain first calibration data; In response to the sphenoid sinus removal signal in the sphenoid sinus target opening area, a sellar floor bone target opening area calibration task is determined, and fourth operation prompt information is generated; a fourth white light image acquired by the endoscope based on the fourth operation prompt information is acquired, a trained sellar floor target detection model is called to identify and calibrate the sellar floor in the fourth white light image to obtain second calibration data, and pre-stored third calibration data of the sellar floor and fourth calibration data of the sellar floor posterior structure group are called, the third calibration data and the fourth calibration data being obtained based on the MRI coronal scan image of the target case through the pituitary gland; fifth calibration data of the sellar floor bone target opening area is obtained based on the second calibration data, the third calibration data, and the fourth calibration data; In response to the sellar floor bone removal signal in the sellar floor bone target opening area, determining the sellar floor dura mater target opening area calibration task, and generating fifth operation prompt information; acquiring a fifth white-light image acquired by the endoscope based on the fifth operation prompt information, calling a trained sellar floor dura mater target detection model to identify and calibrate the sellar floor dura mater in the fifth white-light image to obtain sixth calibration data, calling a trained sellar floor posterior structure group segmentation model to segment and calibrate the sellar floor posterior structure group in the fifth white-light image to obtain seventh calibration data, and obtaining eighth calibration data of the sellar floor dura mater target opening area based on the sixth calibration data and the seventh calibration data; In response to the removal signal of the sellar dura mater in the sellar dura mater target opening area, the pituitary tumor information acquisition task is determined and the sixth operation prompt information is generated; the sixth white light image collected by the endoscope based on the sixth operation prompt information is obtained, the trained tumor segmentation model is called to segment the pituitary tumor in the sixth white light image, and the pituitary tumor information is obtained according to the segmentation result.

2. The method for obtaining pituitary tumor information according to claim 1, further comprising, before the step of obtaining a first white light image acquired by the endoscope in the nasal cavity based on the first operation prompt information: Acquiring a transpituitary MRI coronal scan image of the target case; calling the trained sellar floor target detection model to identify and calibrate the sellar floor in the trans-pituitary MRI coronal scan image to obtain the third calibration data, wherein the third calibration data includes a first sellar floor rectangular frame, a first centroid position of the first sellar floor rectangular frame, and a first area of the first sellar floor rectangular frame; The trained sellar floor posterior structure group segmentation model is called to segment and calibrate the sellar floor posterior structure group in the trans-pituitary MRI coronal scan image to obtain the fourth calibration data, wherein the fourth calibration data includes the segmentation boundary line and centroid position of each structure in the sellar floor posterior structure group.

3. The method for obtaining pituitary tumor information according to claim 2, wherein: The second calibration data includes a second saddle bottom rectangular frame, a second centroid position of the second saddle bottom rectangular frame, and a second area of the second saddle bottom rectangular frame. The step of obtaining fifth calibration data of the saddle bottom bone target opening area according to the second calibration data, the third calibration data, and the fourth calibration data includes: determining relative translation parameters of the first saddle bottom rectangular frame and the second saddle bottom rectangular frame based on a positional deviation between the first centroid and the second centroid, and determining first position parameters of the centroid of each structure in the saddle bottom rear structure group in the fourth white light image based on the relative translation parameters; determining relative scaling parameters of the first saddle bottom rectangular frame and the second saddle bottom rectangular frame according to a ratio of the first area to the second area, and determining second position parameters of segmentation boundaries of each structure in the posterior saddle bottom structure group in the fourth white light image according to the relative scaling parameters; The segmentation boundary line and the centroid position of the sellar floor posterior structure group are transferred to the fourth white light image according to the first position parameter and the second position parameter.

4. The method for obtaining pituitary tumor information according to claim 3, wherein: After the step of transferring the segmentation boundary line and the centroid position of the posterior sellar floor structure group to the fourth white light image, the method includes: Acquire from the fourth white light image a second sellar floor rectangular frame, the centroid position and lower segmentation boundary line of the optic nerve, the centroid position and segmentation boundary line of the pituitary gland, the centroid position and segmentation boundary line of the left carotid artery, and the centroid position and segmentation boundary line of the right carotid artery; A first expansion line of the lower segmentation boundary line of the optic nerve is obtained based on the centroid position and the first expansion parameter of the optic nerve; a first expansion arc of the segmentation boundary line of the left carotid artery is obtained based on the centroid position and the second expansion parameter, the first expansion arc intersecting with the first expansion line at a first intersection and intersecting with the second sellar floor rectangular frame at a second intersection; a second expansion arc of the segmentation boundary line of the right carotid artery is obtained based on the centroid position and the second expansion parameter, the second expansion arc intersecting with the first expansion line at a third intersection and intersecting with the second sellar floor rectangular frame at a fourth intersection; a second expansion line of the lower segmentation boundary line of the pituitary gland is obtained based on the centroid position and the third expansion parameter of the pituitary gland, and fifth and sixth intersections of a horizontal line passing through the lowest point of the second expansion line and the second sellar floor rectangular frame are obtained; Obtain a first upper demarcation line based on the first outward expansion arc, the second outward expansion arc, the first outward expansion line, the first intersection point, and the third intersection point; obtain a left demarcation line based on the second saddle bottom rectangular frame, the second intersection point, and the fifth intersection point; obtain a right demarcation line based on the second saddle bottom rectangular frame, the fourth intersection point, and the sixth intersection point; and obtain a first lower demarcation line based on the horizontal line, the fifth intersection point, and the sixth intersection point; Fifth calibration data of the sellar floor bone target opening area is obtained according to the first upper calibration line, the first lower calibration line, the left calibration line, and the right calibration line.

5. The method for obtaining pituitary tumor information according to claim 4, characterized in that: After the step of obtaining fifth calibration data of the sellar floor bone target opening area according to the first upper calibration line, the first lower calibration line, the left calibration line, and the right calibration line, the method further includes: Acquire a center line of the pituitary stalk from the fourth white light image, and extend the center line to obtain a center extension line located below the first lower calibration line; According to the central extension line and the fourth expansion parameter, a third expansion line and a fourth expansion line located on both sides of the central extension line are obtained, and according to the third expansion line and the fourth expansion line, calibration data of the arterial risk area is obtained.

6. The method for acquiring pituitary tumor information according to claim 1, wherein the sixth calibration data includes a rectangular frame of the sellar floor dura mater, and the seventh calibration data includes segmentation boundaries and centroid positions of each structure in the sellar floor posterior structure group. Based on the sixth and seventh calibration data, eighth calibration data of the sellar floor dura mater target opening area is obtained. The step of determining the sellar floor dura mater target removal site based on the eighth calibration data comprises: Acquire the sellar floor dura mater rectangular frame, the centroid position and segmentation boundary line of the left carotid artery, and the centroid position and segmentation boundary line of the right carotid artery from the fifth white light image; A third outward expansion arc of the segmentation boundary line of the left carotid artery is obtained based on the centroid position of the left carotid artery and the fifth outward expansion parameter, wherein the third outward expansion arc intersects the rectangular frame of the sellar floor dura mater at a seventh intersection point and an eighth intersection point. A fourth outward expansion arc of the segmentation boundary line of the right carotid artery is obtained based on the centroid position of the right carotid artery and the fifth outward expansion parameter, wherein the fourth outward expansion arc intersects the rectangular frame of the sellar floor dura mater at a ninth intersection point and a tenth intersection point. Obtain a second upper demarcation line according to the rectangular frame of the sella serrata dura mater, the seventh intersection point, and the ninth intersection point; and obtain a second lower demarcation line according to the rectangular frame of the sella serrata dura mater, the eighth intersection point, and the tenth intersection point; Eighth calibration data of the target opening area of the sellar floor dura mater is obtained according to the second upper calibration line, the third outward expansion arc, the second lower calibration line, and the fourth outward expansion arc.

7. The method for obtaining pituitary tumor information according to claim 1, further comprising: obtaining a third area of the pituitary tumor according to the pituitary tumor information, and obtaining a fourth area of the sellar floor dura mater target opening region according to the eighth calibration data; Determining whether a ratio of the third area to the fourth area is greater than a preset ratio threshold; If not, generating a prompt message for overall removal of the pituitary tumor; If so, generate block removal prompt information for the pituitary tumor, and in response to the i-th processing result of the block removal prompt information, loop through the operations of obtaining the third area of the remaining pituitary tumor, judging the area ratio and the preset ratio threshold, and generating the removal prompt information until the judgment result is no.

8. The method for obtaining pituitary tumor information according to claim 7, further comprising, after the step of generating the overall removal prompt information of the pituitary tumor: In response to the pituitary tumor removal signal, determining an invading tumor identification task and generating seventh operation prompt information; acquiring a seventh white light image collected by the endoscope based on the seventh operation prompt information, calling a trained tumor invasion recognition model to identify the invading tumor in the seventh white light image, and obtaining a recognition result; When the identification result is that there is no invasion of the tumor, generating non-invasion prompt information; When the recognition result indicates that an invading tumor exists, a risk assessment task for removing the invading tissue tumor is determined according to the recognition result, and eighth operation prompt information is generated.

9. The method for obtaining pituitary tumor information according to claim 8, further comprising, after the step of generating the eighth operation prompt information: acquiring an eighth white light image collected by the endoscope based on the eighth operation prompt information; Invoking the trained invasion tumor removal risk assessment model to assess the invasion tumor removal risk in the eighth white light image to obtain an assessment result; When the assessment result shows that the risk is lower than a threshold, generating a prompt message for removing the invading tumor; When the assessment result shows that the risk is higher than the threshold, a do not remove prompt message is generated.

10. A device for acquiring pituitary tumor information, characterized in that: include: a first acquisition module, configured to acquire a first white-light image captured by the endoscope within the nasal cavity of a target case based on first operation prompt information, the first operation prompt information being generated based on a superior turbinate search task, calling a trained superior turbinate recognition model to identify the superior turbinate in the first white-light image to obtain first recognition information, determining a sphenoid sinus opening search task based on the first recognition information, and generating second operation prompt information; a second acquisition module, configured to acquire a second white light image acquired by the endoscope based on the second operation prompt information, call the trained sphenoid sinus opening recognition model to recognize the sphenoid sinus opening in the second white light image to obtain second recognition information, determine a sphenoid sinus target opening area calibration task based on the second recognition information, and generate third operation prompt information; a third acquisition module, configured to acquire a third white light image acquired by the endoscope based on the third operation prompt information, and call the trained sphenoid sinus target opening area segmentation model to segment and calibrate the sphenoid sinus target opening area in the third white light image to obtain first calibration data; The first obtaining module is configured to determine a sellar floor bone target opening area calibration task in response to a sphenoid sinus removal signal in the sphenoid sinus target opening area, and generate fourth operation prompt information; obtain a fourth white light image acquired by the endoscope based on the fourth operation prompt information, call a trained sellar floor target detection model to identify and calibrate the sellar floor in the fourth white light image to obtain second calibration data, and call pre-stored third calibration data of the sellar floor and fourth calibration data of the sellar floor posterior structure group, the third calibration data and the fourth calibration data being obtained based on a trans-pituitar MRI coronal scan image of the target case; and obtain fifth calibration data of the sellar floor bone target opening area based on the second calibration data, the third calibration data and the fourth calibration data; a second obtaining module, configured to determine a sellar floor dura mater target opening area calibration task in response to a sellar floor bone removal signal in the sellar floor bone target opening area, and generate fifth operation prompt information; acquiring a fifth white-light image acquired by the endoscope based on the fifth operation prompt information, calling a trained sellar floor dura mater target detection model to identify and calibrate the sellar floor dura mater in the fifth white-light image to obtain sixth calibration data, calling a trained sellar floor posterior structure group segmentation model to segment and calibrate the sellar floor posterior structure group in the fifth white-light image to obtain seventh calibration data, and obtaining eighth calibration data of the sellar floor dura mater target opening area based on the sixth calibration data and the seventh calibration data; The third obtaining module is used to determine the pituitary tumor information acquisition task in response to the removal signal of the sellar dura mater in the sellar dura mater target opening area, and generate sixth operation prompt information; obtain the sixth white light image collected by the endoscope based on the sixth operation prompt information, call the trained tumor segmentation model to segment the pituitary tumor in the sixth white light image, and obtain the pituitary tumor information according to the segmentation result.

Images