WO2023185410A1 - 刺激电极导线的成像识别方法及相关装置 - Google Patents

刺激电极导线的成像识别方法及相关装置 Download PDF

Info

Publication number
WO2023185410A1
WO2023185410A1 PCT/CN2023/080446 CN2023080446W WO2023185410A1 WO 2023185410 A1 WO2023185410 A1 WO 2023185410A1 CN 2023080446 W CN2023080446 W CN 2023080446W WO 2023185410 A1 WO2023185410 A1 WO 2023185410A1
Authority
WO
WIPO (PCT)
Prior art keywords
training
preset
image
imaging
mark detection
Prior art date
Application number
PCT/CN2023/080446
Other languages
English (en)
French (fr)
Inventor
姜传江
Original Assignee
苏州景昱医疗器械有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 苏州景昱医疗器械有限公司 filed Critical 苏州景昱医疗器械有限公司
Publication of WO2023185410A1 publication Critical patent/WO2023185410A1/zh

Links

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61N—ELECTROTHERAPY; MAGNETOTHERAPY; RADIATION THERAPY; ULTRASOUND THERAPY
    • A61N1/00—Electrotherapy; Circuits therefor
    • A61N1/18—Applying electric currents by contact electrodes
    • A61N1/32—Applying electric currents by contact electrodes alternating or intermittent currents
    • A61N1/36—Applying electric currents by contact electrodes alternating or intermittent currents for stimulation
    • A61N1/372—Arrangements in connection with the implantation of stimulators
    • A61N1/37211—Means for communicating with stimulators
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61N—ELECTROTHERAPY; MAGNETOTHERAPY; RADIATION THERAPY; ULTRASOUND THERAPY
    • A61N1/00—Electrotherapy; Circuits therefor
    • A61N1/02—Details
    • A61N1/04—Electrodes
    • A61N1/05—Electrodes for implantation or insertion into the body, e.g. heart electrode
    • A61N1/0526—Head electrodes
    • A61N1/0529—Electrodes for brain stimulation
    • A61N1/0534—Electrodes for deep brain stimulation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61N—ELECTROTHERAPY; MAGNETOTHERAPY; RADIATION THERAPY; ULTRASOUND THERAPY
    • A61N1/00—Electrotherapy; Circuits therefor
    • A61N1/18—Applying electric currents by contact electrodes
    • A61N1/32—Applying electric currents by contact electrodes alternating or intermittent currents
    • A61N1/36—Applying electric currents by contact electrodes alternating or intermittent currents for stimulation
    • A61N1/372—Arrangements in connection with the implantation of stimulators
    • A61N1/37211—Means for communicating with stimulators
    • A61N1/37252—Details of algorithms or data aspects of communication system, e.g. handshaking, transmitting specific data or segmenting data

Definitions

  • This application relates to the technical field of implantable medical equipment, for example, to imaging identification methods and related devices for stimulation electrode leads.
  • DBS Deep Brain Electrical stimulation therapy
  • DBS Deep Brain Stimulation
  • the stimulating electrode used to apply electrical stimulation acts on the patient's head and stimulates a designated part of the brain to treat the patient's brain damage.
  • the other end of the stimulating electrode is connected to the neurostimulator through the stimulating electrode lead.
  • the imaging techniques include magnetic resonance imaging (MRI, Magnetic Resonance Imaging), computed tomography (CT, computed tomography), X-ray, fluorescence imaging, and three-dimensional imaging.
  • physicians desire to precisely place and orient stimulation electrode leads that can deliver stimulation within a patient (e.g., the brain) to deliver electrical stimulation to the intended site and avoid side effects. For example, it is desirable to deliver stimulation from a stimulating lead to a very small target site so as not to stimulate other nearby brain tissue; if stimulation is not delivered precisely to the desired target site, efficacy will be reduced and adjacent brain tissue will be damaged. Receive unnecessary excessive stimulation, causing pain to the patient.
  • the purpose of this application is to provide an imaging identification method and related devices for stimulation electrode leads, through imaging
  • the technology directly identifies the marks set on the electrode sheets, and solves the problem of low accuracy in identifying stimulation electrode leads by judging the confidence of each label.
  • the present application provides an imaging identification device for a stimulation electrode lead.
  • a plurality of electrode sheets are arranged on the outer peripheral surface of the stimulation electrode lead. At least some of the electrode sheets are respectively provided with marks, and the marks are used to identify the stimulation electrode leads.
  • the electrode sheet is identified during imaging; the device includes: an image acquisition module, used to use image acquisition equipment to collect the target image of the stimulation electrode lead in real time; a detection result module, used to obtain the mark corresponding to the target image Detection results, the mark detection results include one or more tags and their confidence and location information; a confidence judgment module, used to judge whether the confidence of each tag meets the preset conditions; when one or more tags When the confidence of all labels does not meet the preset conditions, the image acquisition module is called again; when the confidence of all labels meets the preset conditions, the result output module is called; the result output module is used to output the mark detection results to Default user device.
  • an image acquisition module used to use image acquisition equipment to collect the target image of the stimulation electrode lead in real time
  • a detection result module used to obtain the mark corresponding to the target image Detection results
  • the mark detection results include one or more tags and their confidence and location information
  • a confidence judgment module used to judge whether the confidence of each tag meets the preset conditions; when one or more tags When
  • the detection result module includes: an imaging recognition unit, used to perform imaging recognition on the target image using an imaging recognition model to obtain a mark detection result corresponding to the target image; wherein, the imaging recognition
  • the training process of the model is as follows: obtain a first training set, which includes a plurality of first training images and the annotation data of their corresponding mark detection results; use the first training set to test the preset first depth
  • the learning model is trained to obtain the imaging recognition model.
  • using the first training set to train a preset first deep learning model includes: for each image in the first training set a first training image, input the first training image into a preset first deep learning model, and obtain the prediction data of the mark detection result corresponding to the first training image; based on the corresponding first training image
  • the prediction data and annotation data of the corresponding mark detection results are used to update the model parameters of the preset first deep learning model; it is detected whether the preset first training end condition is met, and if so, the training is stopped and the The preset first deep learning model obtained by training is used as the imaging recognition model. If not, the next training data is used to continue training the preset first deep learning model.
  • the detection result module includes: a target detection unit, used to perform target detection on the target image to obtain one or more sub-images and their corresponding position information, each sub-image corresponding to a mark;
  • the sub-picture classification unit is used to label and classify each sub-picture and obtain the label corresponding to each sub-picture and its confidence level;
  • the labeling result unit is used to label and classify each sub-picture based on the corresponding label and its confidence level and location information, to obtain a mark detection result corresponding to the target image, where the mark detection result includes one or more labels and their confidence and location information.
  • the sub-picture classification unit includes: a sub-picture classification sub-unit, which is used to perform label classification on each sub-picture using a label classification model to obtain a label classification result corresponding to each sub-picture; wherein, the label
  • the training process of the classification model is as follows: obtain a second training set, which includes a plurality of second training images and their corresponding annotation data of labeled classification results; use the second training set to compare the preset second
  • the deep learning model is trained to obtain the label classification model.
  • using the second training set to train a preset second deep learning model includes: for each item in the second training set a second training image, input the second training image into a preset second deep learning model, and obtain prediction data of the mark detection results corresponding to the second training image; based on the second training image
  • the prediction data and annotation data of the corresponding mark detection results are used to update the model parameters of the preset second deep learning model; it is detected whether the preset second training end condition is met, and if so, the training is stopped and the The preset second deep learning model obtained by training is used as the label classification model. If not, the next training data is used to continue training the preset second deep learning model.
  • the apparatus further includes: a result display module, configured to display the target image and its corresponding mark detection result using the user equipment.
  • the present application provides an imaging identification method for stimulating electrode leads.
  • a plurality of electrode sheets are arranged on the outer peripheral surface of the stimulating electrode leads. At least some of the electrode sheets are respectively provided with marks, and the marks are used for imaging during imaging.
  • Identify the electrode sheet; the method includes: S101: Use an image acquisition device to collect the target image of the stimulation electrode lead in real time; S102: Obtain the mark detection result corresponding to the target image, and the mark detection result includes a or multiple tags and their confidence and location information; S103: Determine whether the confidence of each tag meets the preset conditions; when the confidence of one or more tags does not meet the preset conditions, re-execute step S101 to Obtain a new target image; when the confidence levels of all labels meet the preset conditions, perform step S104; S104: Output the mark detection results to the preset user equipment.
  • the step S102 includes: using an imaging recognition model to perform imaging recognition on the target image to obtain a mark detection result corresponding to the target image; wherein, the training process of the imaging recognition model is as follows: Obtain a first training set, the first training set includes a plurality of first training images and the annotation data of the corresponding mark detection results; using the first training set to train the preset first deep learning model to obtain the imaging recognition model.
  • using the first training set to train a preset first deep learning model includes: for each image in the first training set a first training image, input the first training image into a preset first deep learning model, and obtain the prediction data of the mark detection result corresponding to the first training image; based on the corresponding first training image
  • the prediction data and annotation data of the corresponding mark detection results are used to update the model parameters of the preset first deep learning model; it is detected whether the preset first training end condition is met, and if so, the training is stopped and the The preset first deep learning model obtained by training is used as the imaging recognition model. If not, the next training data is used to continue training the preset first deep learning model.
  • obtaining the mark detection result corresponding to the target image includes: performing target detection on the target image to obtain one or more sub-images and their corresponding position information, each sub-image corresponding to a mark. ; Perform mark classification on each sub-image to obtain the label corresponding to each sub-image and its confidence level; obtain the mark detection result corresponding to the target image based on the label corresponding to each sub-image, its confidence level and location information, and the mark
  • the detection results include one or more tags along with their confidence and location information.
  • the step S202 includes: using a mark classification model to perform mark classification on each sub-image to obtain a mark classification result corresponding to each sub-image; wherein, the training process of the mark classification model is as follows: Obtain the second A training set, the second training set includes a plurality of second training images and the annotation data of their corresponding label classification results; the second training set is used to train a preset second deep learning model to obtain the labeling Classification model.
  • using the second training set to train a preset second deep learning model includes: for each item in the second training set a second training image, input the second training image into a preset second deep learning model, and obtain prediction data of the mark detection results corresponding to the second training image; based on the second training image
  • the prediction data and annotation data of the corresponding mark detection results are used to update the model parameters of the preset second deep learning model; it is detected whether the preset second training end condition is met, and if so, the training is stopped and the The preset second deep learning model obtained by training is used as the label classification model. If not, the next training data is used to continue training the preset second deep learning model.
  • the method further includes step S105: using the user equipment to display the The target image and its corresponding mark detection results are described.
  • the present application provides an electronic device for imaging and identifying stimulation electrode leads.
  • the electrode sheets are arranged on the outer peripheral surface of the stimulation electrode leads. At least part of the electrode sheets are respectively provided with marks, and the marks are used for The electrode sheet is identified during imaging;
  • the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.
  • the present application provides a computer-readable storage medium that stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.
  • Detection is based on real-time collected target images, and the confidence of the tags in the detection results is judged through preset conditions (the tags are used to indicate the identification of the electrode pads), and the image is recalled when the confidence of each tag does not meet the preset conditions.
  • the acquisition module performs image acquisition, detection and judgment until a mark detection result that meets the confidence condition is obtained. Compared with related technologies, a more accurate stimulation electrode lead identification result is obtained.
  • the doctor refers to the identification results of the stimulation electrode leads, which does not require the doctor to make complex logical judgments. It is highly intelligent, and even inexperienced doctors can accurately deliver stimulation to the desired target point, shortening the time for doctors to place and orient the stimulation electrode leads. This improves the doctor's efficiency in accurately placing and orienting the stimulation electrode leads, reduces the patient's pain during the doctor's placement and directional stimulation of the electrode leads, and thereby improves the efficacy of electrical stimulation therapy for patients.
  • the marked electrode pads can determine the electrode position after imaging recognition, and can also be used to generate stimulation signals. There is no need to preset markers in the non-electrode pad areas. components, which can reduce the manufacturing cost and difficulty of manufacturing the stimulation electrode leads.
  • Figure 1 is a schematic structural diagram of an imaging identification device for stimulating electrode leads provided by an embodiment of the present application
  • Figure 2 is a partial perspective view of a stimulation electrode lead provided by an embodiment of the present application.
  • Figure 3 is a partial structure of a stimulation electrode lead in a flattened state provided by an embodiment of the present application.
  • FIG. 4 is a partial structural schematic diagram of another stimulation electrode lead in a flattened state provided by an embodiment of the present application.
  • FIG. 5 is a partial structural schematic diagram of another stimulation electrode lead in a flattened state provided by an embodiment of the present application.
  • FIG. 6 is a partial structural schematic diagram of another stimulation electrode lead in a flattened state provided by an embodiment of the present application.
  • FIG. 7 is a partial structural schematic diagram of another stimulation electrode lead in a flattened state provided by an embodiment of the present application.
  • Figure 8 is a schematic structural diagram of a detection result module provided by an embodiment of the present application.
  • FIG. 9 is a schematic structural diagram of another imaging identification device for stimulating electrode leads provided by an embodiment of the present application.
  • Figure 10 is a schematic flow chart of an imaging identification method for stimulation electrode leads provided by an embodiment of the present application.
  • Figure 11 is a schematic flowchart of obtaining mark detection results provided by an embodiment of the present application.
  • Figure 12 is a schematic flow chart of yet another imaging identification method for stimulation electrode leads provided by an embodiment of the present application.
  • Figure 13 is a schematic structural diagram of an electronic device provided by an embodiment of the present application.
  • Figure 14 is a schematic structural diagram of a computer-readable storage medium provided by an embodiment of the present application.
  • the implantable neurostimulation system mainly includes a stimulator implanted in the patient's body (i.e., an implantable neurostimulator, a nerve stimulation device) and a program-controlled device installed outside the patient's body.
  • a stimulator implanted in the patient's body i.e., an implantable neurostimulator, a nerve stimulation device
  • a program-controlled device installed outside the patient's body.
  • Relevant neuromodulation technology mainly involves implanting electrodes in specific parts of the body's tissues (i.e., target points) through stereotaxic surgery, and a stimulator implanted in the patient's body sends electrical pulses to the target point through the electrodes to regulate the corresponding neural structures. And the electrical activity and functions of the network, thereby improving symptoms and relieving pain.
  • the stimulator can be an implantable nerve electrical stimulation device, an implantable cardiac electrical stimulation system (also known as a pacemaker), an implantable drug delivery device (Implantable Drug Delivery System, referred to as IDDS) and a lead. Any type of switching device.
  • Implantable neuroelectric stimulation devices include, for example, Deep Brain Stimulation (DBS), Cortical Nerve Stimulation (CNS), and Spinal Cord Stimulation. , referred to as SCS), implanted sacral nerve electrical stimulation system (Sacral Nerve Stimulation, referred to as SNS), implanted vagus nerve electrical stimulation system (Vagus Nerve Stimulation, referred to as VNS), etc.
  • the stimulator can include an IPG (implantable pulse generator, implantable pulse generator), extension wires, and stimulation electrode wires.
  • IPG implantable pulse generator
  • the IPG is installed in the patient's body, receives program-controlled instructions sent by the program-controlled equipment, and relies on sealed batteries and circuits to provide controllable information to the tissues in the body.
  • the electrical stimulation energy is delivered through implanted extension leads and stimulation electrode leads to deliver one or two controllable specific electrical stimulations to specific areas of tissue in the body.
  • the extension lead is used in conjunction with the IPG as a transmission medium for electrical stimulation signals to transmit the electrical stimulation signals generated by the IPG to the stimulation electrode leads.
  • the stimulating electrode lead may be a neurostimulating electrode that delivers electrical stimulation to specific areas of tissue in the body through multiple electrode contacts.
  • the stimulator is provided with one or more stimulation electrode leads on one or both sides, and multiple electrode contacts are provided on the stimulation electrode leads.
  • the electrode contacts can be evenly or non-uniformly arranged in the circumferential direction of the stimulation electrode leads.
  • the electrode contacts may be arranged in an array of 4 rows and 3 columns (12 electrode contacts in total) in the circumferential direction of the stimulation electrode lead.
  • Electrode contacts may include stimulation contacts and/or signal acquisition contacts.
  • the electrode contacts may be, for example, sheet-shaped, ring-shaped, dot-shaped, etc., and the electrode contacts may be electrode sheets in the above-mentioned shapes.
  • the stimulated internal tissue may be the patient's brain tissue, and the stimulated site may be a specific part of the brain tissue.
  • the stimulated parts are generally different, the number of stimulation contacts used (single source or multiple sources), one or more channels (single channel or multi-channel) specific electrical stimulation signals
  • the application and stimulation parameter data are also different.
  • the embodiments of this application do not limit the applicable disease types, which may be the disease types applicable to deep brain stimulation (DBS), spinal cord stimulation (SCS), pelvic stimulation, gastric stimulation, peripheral nerve stimulation, and functional electrical stimulation.
  • the program-controlled device when the program-controlled device and the stimulator establish a program-controlled connection, can be used to adjust the stimulation parameters of the stimulator (different stimulation parameters correspond to different electrical stimulation signals), and the stimulator can also be used to sense the deep brain of the patient.
  • the bioelectrical activity is used to collect electrophysiological signals, and the stimulation parameters of the electrical stimulation signal of the stimulator can be continuously adjusted through the collected electrophysiological signals.
  • Stimulation parameters can include: frequency (for example, the number of electrical stimulation pulse signals per unit time 1 s, the unit is Hz), pulse width (the duration of each pulse, the unit is ⁇ s), amplitude (generally expressed in voltage, that is, The intensity of each pulse, in V), timing (for example, it can be continuous or triggered), stimulation mode (including one or more of current mode, voltage mode, timed stimulation mode and cyclic stimulation mode), physician control upper limit One or more of the upper and lower limits (the range that the doctor can adjust) and the upper and lower limits of the patient's control (the range that the patient can adjust independently).
  • frequency for example, the number of electrical stimulation pulse signals per unit time 1 s, the unit is Hz
  • pulse width the duration of each pulse, the unit is ⁇ s
  • amplitude generally expressed in voltage, that is, The intensity of each pulse, in V
  • timing for example, it can be continuous or triggered
  • stimulation mode including one or more of current mode, voltage mode, timed stimulation mode and cyclic stimulation mode
  • each stimulation parameter of the stimulator can be adjusted in current mode or voltage mode.
  • the program-controlled equipment may be a doctor-programmed equipment (that is, a program-controlled equipment used by a doctor) or a patient-programmed equipment (that is, a program-controlled equipment used by a patient).
  • the doctor's program-controlled equipment may be, for example, a tablet computer, a notebook computer, a desktop computer, a mobile phone, or other intelligent terminal equipment equipped with program-controlled software.
  • the patient's program-controlled equipment can be, for example, tablet computers, laptops, desktop computers, mobile phones and other intelligent terminal devices equipped with program-controlled software.
  • the patient's program-controlled equipment can also be other electronic equipment with program-controlled functions (such as chargers, data sets with program-controlled functions). collection equipment).
  • the embodiments of this application do not limit the data interaction between the doctor's program-controlled equipment and the stimulator.
  • the doctor remotely programs the device
  • the doctor's program-controlled equipment can interact with the stimulator through the server and the patient's program-controlled equipment.
  • the doctor performs face-to-face programming with the patient offline
  • the doctor's program-controlled equipment can interact with the stimulator through the patient's program-controlled equipment, and the doctor's program-controlled equipment can also directly interact with the stimulator.
  • the patient programming device may include a host (in communication with the server) and a host (in communication with the server). Stimulator communication) slave machine, the connection between the host machine and the slave machine that can communicate.
  • the doctor's program-controlled equipment can interact with the server through the 3G/4G/5G network
  • the server can interact with the host through the 3G/4G/5G network
  • the host can interact with the slave through the Bluetooth protocol/WIFI protocol/USB protocol.
  • the slave machine can interact with the stimulator through the 401MHz-406MHz operating frequency band/2.4GHz-2.48GHz operating frequency band, and the doctor's program-controlled equipment can directly conduct data with the stimulator through the 401MHz-406MHz operating frequency band/2.4GHz-2.48GHz operating frequency band. Interaction.
  • the Chinese patent with publication number CN112604159A discloses a segmented electrode that can identify the electrode orientation by setting additional marks, and determine the position and direction of the electrode through the correspondence between the predefined mark direction and the electrode stimulation piece.
  • This The method requires doctors to have strong logical judgment ability, and errors are prone to occur during the judgment process, causing unnecessary pain to patients.
  • an embodiment of the present application provides an imaging identification device for a stimulation electrode lead.
  • a plurality of electrode sheets are arranged on the outer peripheral surface of the stimulation electrode lead. At least some of the electrode sheets are respectively provided with marks, and the marks are used to identify the stimulation electrode leads. The electrode pads are identified during imaging.
  • the above device can realize an identification method that is different from related technologies, (using a manufacturing process such as a flexible film circuit) to directly use the structure of the mark on the electrode sheet to identify the differently marked electrodes in the stimulation electrode lead through the image obtained by the imaging technology.
  • the position and orientation of the piece that is to say, compared with the additional setting of marks in the related art, the orientation can be identified using the marks respectively set on the electrode sheets during imaging.
  • FIGs 3 to 7 they are schematic diagrams of the stimulation electrode leads in a flat state. Markings provided on the electrode sheets can be used to distinguish and identify each electrode sheet. For example, the difference in shape of each electrode piece in the electrode piece is used as a mark, or the location of the connection points on the electrode piece is different as a mark to distinguish different electrode pieces ( Figure 3, Figure 4, Figure 6 and Figure 7), or the connection points on the electrode piece are The different shapes are used as marks to distinguish different electrode sheets ( Figure 5), or a combination of different marking arrangements on the above electrode sheets. Doctors can implant stimulating electrode leads in the deep brain area of the patient. The surface of the stimulating electrode lead can be provided with multiple electrode pads arranged in a regular matrix, and use the neurostimulator to stimulate the multiple electrode pads set on the stimulating electrode lead to release stimulation.
  • the electrode sheets can be identified by marking electrodes in different rows and columns.
  • the electrode sheets have 4 rows and 3 columns in each row (12 output labels, namely electrode sheet No. 1 to electrode sheet No. 12).
  • the electrode sheets in the first row and the first column can be labeled. and the second row and second column electrode pads; in another embodiment, the second row, third column electrode pads and the fourth row, second column electrode pads can be marked; in yet another embodiment, the first row, third column electrode pads can be marked Three columns of electrode pads and the third row The first column electrode pad; in another embodiment, the first row, second column electrode pad and the fourth row, third column electrode pad can be marked.
  • the electrode sheets have 5 rows and 4 columns in each row (20 output labels, namely electrode sheet No. 1 to electrode sheet No. 20).
  • the electrodes in the first row and the first column can be labeled. and the second row and second column electrode sheets; in another embodiment, the second row, third column electrode sheet and the fourth row, second column electrode sheet can be marked; in yet another embodiment, the first row can be marked The third column electrode pad and the third row first column electrode pad; in another embodiment, the first row second column electrode pad and the fifth row third column electrode pad can be marked.
  • the electrode sheets have 5 rows and 4 columns in each row (20 output labels, namely electrode sheet No. 1 to electrode sheet No. 20).
  • the electrodes in the first row and the first column can be labeled. electrode sheets, the second row and second column electrode sheets, and the third row and third column electrode sheets; in another embodiment, the second row and third column electrode sheets, the fourth row and second column electrode sheets, the fifth row and third column electrode sheets can be marked Four columns of electrode sheets; in another embodiment, the electrode sheets in the first row and third column, the electrode sheets in the third row and first column, and the electrode sheets in the fourth row and second column can be marked; in yet another embodiment, the electrode sheets in the fourth row and second column can be marked The first row and second column electrode pads, the fourth row and third column electrode pads, and the fifth row and first column electrode pads.
  • the objects who perform the identification operation on the stimulation electrode leads through the imaging identification device can be the patient's first doctor, consulting experts and other people who treat the patient.
  • the patients in the embodiments of this application may be patients with Parkinson's disease, patients with depression, obsessive-compulsive disorder and other mental illnesses, or may be patients with drug addiction or detoxification patients.
  • the electrical stimulation of the stimulator can be delivered to specific areas of the human body to apply stimulation therapy.
  • the stimulating electrode leads can release electrical stimulation to the neural structures of the brain to excite or inhibit cell activities, which can effectively treat spastic diseases (eg, epilepsy), pain, migraine, mental illness (eg, severe Depression (MDD), bipolar disorder, anxiety disorder, post-traumatic stress disorder, mild depression, obsessive-compulsive disorder (OCD), behavioral disorders, mood disorders, memory disorders, mental status disorders, mobility disorders (e.g., special tremor or Parkinson's disease), Huntington's disease, Alzheimer's disease, drug addiction, autism, or other neurological or psychiatric diseases and impairments.
  • spastic diseases eg, epilepsy
  • MMDD severe Depression
  • OCD obsessive-compulsive disorder
  • behavioral disorders e.g., anxiety disorder, post-traumatic stress disorder, mild depression, obsessive-compulsive disorder (OCD)
  • behavioral disorders e.g., special
  • the device includes an image acquisition module 101, a detection result module 102, a confidence judgment module 103 and a result output module 104.
  • the image acquisition module 101 is used to collect the target image of the stimulation electrode lead in real time using an image acquisition device.
  • Image acquisition equipment may include capabilities such as magnetic resonance imaging (MRI), computed tomography Imaging equipment for imaging technologies such as CT, X-ray, fluorescence imaging, and three-dimensional imaging.
  • MRI magnetic resonance imaging
  • CT computed tomography Imaging equipment for imaging technologies such as CT, X-ray, fluorescence imaging, and three-dimensional imaging.
  • Figure 2 is a partial perspective view of a stimulation electrode lead obtained through X-rays.
  • the detection result module 102 is used to obtain the mark detection result corresponding to the target image.
  • the mark detection result includes one or more tags and their confidence and location information.
  • the position information can be the coordinate value corresponding to the mark. Through the position information, the position of the corresponding mark on the target image can be accurately obtained.
  • the confidence judgment module 103 is used to judge whether the confidence of each label meets the preset conditions; when the confidence of one or more labels does not meet the preset conditions, re-call the image acquisition module; when the confidence of all labels When the preset conditions are met, the result output module is called.
  • the result output module 104 is used to output the mark detection result to a preset user equipment.
  • the user equipment used to receive the marker detection results can adopt a programmable controller existing in the related technology, that is, the user equipment can be a separate hardware device that can interact with the stimulator through a wireless network or a wired network.
  • Electronic devices such as tablets, computers, mobile phones or smart wearable devices, etc. Users can use this program-controlled device to receive mark detection results.
  • the user equipment is equipped with a computer program (ie, software), and when the computer program is executed by the processor, it can realize the function of receiving the mark detection results in the embodiment of the present application.
  • the preset condition is, for example, a numerical range preset condition. In one embodiment, the preset condition is that the confidence level is not less than the preset confidence level.
  • the preset confidence levels are, for example, 0.95, 0.97, 0.94, and 0.98.
  • This application can set the same or different preset conditions for different patients.
  • the same preset condition is set for different patients, that is, the confidence level is not less than 0.96.
  • differentiated preset conditions can be set according to different patient conditions or treatment stages to achieve humanized and customized diagnosis and treatment of patients.
  • the doctor uses the above-mentioned imaging recognition device of the stimulating electrode leads to determine the position and direction of the electrode pads of the stimulating electrode leads implanted in the bodies of patients Zhang San, Li Si and Wang Wu. Refer to Table 1 below, and the specific judgment conditions are as follows.
  • doctor's consultation time and patient's treatment time are very precious, so doctors and patients prefer to position the stimulation electrode leads to effective desired target points to reduce re-diagnosis and treatment caused by poor positioning results. Therefore, by judging whether the confidence of each label meets the preset conditions, even if the confidence of a label does not meet the preset conditions, the image acquisition module will be called again, so that the doctor can treat the patient through stimulating electrode leads later.
  • the stimulation is delivered to a very small target point without irritating adjacent brain tissue, saving more treatment time and thus reducing the patient's discomfort during treatment.
  • the stimulation electrode lead identification result can be obtained more accurately and with a higher degree of intelligence.
  • the doctor refers to the identification results of the stimulation electrode leads without the need for the doctor to make complex logical judgments.
  • Even an inexperienced doctor can accurately deliver the stimulation to the desired target point, shortening the time for the doctor to place and orient the stimulation electrode leads, and improving the efficiency of the stimulation.
  • the doctor's efficiency in accurately placing and orienting the stimulation electrode leads reduces the patient's pain during the doctor's placement and direction of the stimulation electrode lead and improves the efficacy of electrical stimulation for the patient.
  • the marked electrode pads can determine the electrode position after imaging recognition, and can also be used to generate stimulation signals. There is no need to preset markers in the non-electrode pad areas. components, which can reduce the manufacturing cost and difficulty of manufacturing the stimulation electrode leads.
  • any two electrode sheets among the plurality of electrode sheets are Insulated, the plurality of electrode sheets include a plurality of stimulation electrode sheets and a plurality of collection electrode sheets.
  • the stimulation electrode lead can not only be used to release electrical stimulation energy, but can also be used to collect bioelectrical signals from tissues in the body.
  • the device may further include a result display module 105, which is configured to display the target image and its corresponding mark detection result using the user equipment.
  • the result display module 105 may include a display, a projector and other equipment modules that provide display functions.
  • the display of the mark detection result and the target image on the display module can be understood as displaying the target image on the interface of the display module, and the position information is used to correspond the mark detection result and the image displayed on the target image.
  • the corresponding label, confidence level, etc. can be displayed on the electrode sheet.
  • the displayed labels may be electrode sheet 1, electrode sheet 2...electrode sheet N, etc., and the marked confidence levels may be 0.91, 0.94, 0.98, etc.
  • the target image of the stimulation electrode lead collected in real time by the image acquisition device, the label of the mark of the electrode sheet in the stimulation electrode lead, and its confidence and position information can be intuitively displayed on the display module.
  • the detection result module may include an imaging recognition unit.
  • the imaging recognition unit may be used to perform imaging recognition on the target image using an imaging recognition model to obtain a mark detection result corresponding to the target image.
  • the training process of the imaging recognition model is as follows:
  • the first training set includes annotation data of a plurality of first training images and their corresponding mark detection results
  • the first training set is used to train a preset first deep learning model to obtain the imaging recognition model.
  • the degree of intelligence is higher; when the trained imaging recognition model is applied to the imaging recognition of stimulation electrode leads in actual scenes, the recognition accuracy is high.
  • using the first training set to train a preset first deep learning model may include:
  • the first deep learning model is used to obtain prediction data of the mark detection results corresponding to the first training image; based on the prediction data and annotation data of the mark detection results corresponding to the first training image, the prediction data is The model parameters of the first deep learning model are updated; it is detected whether the preset first training end condition is met, and if so, the training is stopped, and the preset first deep learning model obtained by training is used as the Imaging recognition model, if not, use the next training data to continue training the preset first deep learning model.
  • the imaging recognition model can be trained with a large amount of training data and can predict corresponding mark detection for a variety of input data. As a result, it has a wide range of applications and a high level of intelligence. Through design, establish an appropriate number of neuron computing nodes and a multi-layer computing hierarchy, and select the appropriate input layer and output layer to obtain the preset first deep learning model. Through the learning of the preset first deep learning model and tuning to establish the functional relationship from input to output. Although the functional relationship between input and output cannot be found 100%, it can be as close as possible to the realistic correlation relationship. The imaging recognition model trained thus can realize imaging The self-diagnosis function of identification is high, and the diagnostic results are highly reliable.
  • this application uses the first training set to train the first deep learning model, so that the final The recognition effect of the imaging recognition model is more consistent with the actual imaging results, and users get more satisfactory imaging recognition results of the stimulation electrode leads, improving the user experience.
  • the detection result module may also include a target detection unit 201 , a subgraph classification unit 202 and a labeling result unit 203 .
  • the target detection unit 201 is used to perform target detection on the target image to obtain one or more sub-images and their corresponding position information, each sub-image corresponding to a mark.
  • the main attributes of the target image can be reflected through sub-images, and compression, denoising and other processing of image data can be achieved.
  • the sub-picture classification unit 202 is used to label and classify each sub-picture, and obtain the label corresponding to each sub-picture and its confidence level.
  • Marking result unit 203 is used to obtain the mark detection result corresponding to the target image based on the label corresponding to each sub-image and its confidence and position information.
  • the mark detection result includes one or more labels and their confidence and position. information.
  • the objects in the target image are The mark is classified through each sub-image to obtain the label corresponding to each sub-image and its confidence level, and then the mark detection results including all labels, their confidence level and location information corresponding to the target image are obtained, which is highly intelligent.
  • the sub-picture classification unit may include a sub-picture classification sub-unit, which may be used to perform label classification on each sub-picture using a label classification model to obtain a label classification corresponding to each sub-picture. result.
  • the second training set including a plurality of second training images and their corresponding annotation data of the labeled classification results
  • the second training set is used to train the preset second deep learning model to obtain the mark classification model.
  • each subgraph is labeled and classified through the subgraph classification subunit, and the label classification results corresponding to each subgraph are obtained, which are used to train the label classification model, which can improve the robustness of the label classification model and effectively reduce its fitting risk.
  • using the second training set to train the preset second deep learning model may include the following steps:
  • the second training end condition at the end of training can be configured based on actual needs, and the trained label classification model has strong robustness and low overfitting risk.
  • the label classification model can be trained with a large amount of training data and can predict corresponding label detection for a variety of input data. As a result, it has a wide range of applications and a high level of intelligence.
  • build Establish an appropriate number of neuron computing nodes and a multi-layer computing hierarchy, and select the appropriate input layer and output layer to obtain the preset second deep learning model.
  • the label classification model trained can realize self-processing of imaging recognition. Diagnostic function, and the diagnostic results are highly reliable.
  • an embodiment of the present application also provides an imaging identification method for stimulation electrode leads. Since the imaging identification method of the stimulating electrode lead plays the same or similar role as the imaging identification device of the stimulating electrode lead mentioned above, it will not be described again here.
  • a plurality of electrode sheets are arranged on the outer peripheral surface of the stimulation electrode lead, and at least some of the electrode sheets are respectively provided with marks, and the marks are used to identify the electrode sheets during imaging.
  • the method includes steps S101 to S104.
  • Step S101 Use an image acquisition device to collect the target image of the stimulation electrode lead in real time.
  • Step S102 Obtain the mark detection result corresponding to the target image.
  • the mark detection result includes one or more tags, their confidence level and location information.
  • Step S103 Determine whether the confidence of each label meets the preset conditions; when the confidence of one or more labels does not meet the preset conditions, re-execute step S101 to obtain a new target image; when the confidence of all labels When the preset conditions are satisfied, step S104 is executed.
  • Step S104 Output the mark detection result to a preset user equipment.
  • step S102 may include: performing imaging recognition on the target image using an imaging recognition model to obtain a mark detection result corresponding to the target image.
  • the training process of the imaging recognition model is as follows: obtain a first training set, the first training set includes annotation data of a plurality of first training images and their corresponding mark detection results; use the first training set to The preset first deep learning model is trained to obtain the imaging recognition model.
  • using the first training set to train a preset first deep learning model may include:
  • obtaining the mark detection result corresponding to the target image may include step S201 to step S203.
  • Step S201 Perform target detection on the target image to obtain one or more sub-images and their corresponding position information. Each sub-image corresponds to a mark.
  • Step S202 Perform label classification on each sub-picture to obtain the label corresponding to each sub-picture and its confidence level.
  • Step S203 Based on the label corresponding to each sub-image and its confidence level and location information, obtain the mark detection result corresponding to the target image.
  • the mark detection result includes one or more labels, their confidence level and location information.
  • step S202 may include:
  • the second training set including a plurality of second training images and their corresponding annotation data of the labeled classification results
  • the second training set is used to train the preset second deep learning model to obtain the mark classification model.
  • using the second training set to train a preset second deep learning model may include:
  • the method may further include step S105.
  • Step S105 Use the user equipment to display the target image and its corresponding mark detection result.
  • an embodiment of the present application also provides an electronic device 200 for imaging and identifying stimulation electrode leads.
  • the stimulation electrode leads are arranged on the outer peripheral surface of the stimulation electrode leads. At least part of the electrode sheets are respectively provided with Mark, the mark is used to identify the electrode sheet during imaging.
  • the electronic device 200 includes one or more memories 210, one or more processors 220, and a bus 230 connecting different platform systems.
  • Memory 210 may include readable media in the form of volatile memory, such as random access memory (RAM) 211 and/or cache memory 212, and may further include read only memory (ROM) 213.
  • RAM random access memory
  • ROM read only memory
  • the memory 210 also stores a computer program.
  • the computer program can be executed by the processor 220, so that the processor 220 executes the steps of the above-mentioned method in the embodiment of the present application.
  • the specific implementation manner is the same as that described in the above-mentioned method embodiment.
  • the technical effects achieved are the same, and some contents will not be repeated again.
  • Memory 210 may also include utilities 214 having one or more program modules 215, such program modules 215 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of these examples.
  • program modules 215 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of these examples.
  • One or some combination may include the implementation of a network environment.
  • the processor 220 can execute the above-mentioned computer program, and can execute the utility tool 214.
  • Bus 230 may be a local bus representing one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or using any of a variety of bus structures.
  • the electronic device 200 may also communicate with one or more external devices 240, such as a keyboard, a pointing device, a Bluetooth device, etc., and may also communicate with one or more devices capable of interacting with the electronic device 200, and/or with a device that enables the electronic device 200 to 200 is any device capable of communicating with one or more other computing devices (eg, router, modem, etc.). This communication may occur through the input/output interface 250.
  • the electronic device 200 may also communicate with one or more networks (eg, a local area network (LAN), a wide area network (WAN), and/or a public network, such as the Internet) through the network adapter 260.
  • Network adapter 260 may communicate with other modules of electronic device 200 via bus 230.
  • Embodiments of the present application also provide a computer-readable storage medium, the specific implementation manner of which is consistent with the implementation manners and technical effects achieved described in the embodiments of the above method, and part of the content will not be described again.
  • the computer-readable storage medium is used to store a computer program; when the computer program is executed, the steps of the above method in the embodiment of the present application are implemented.
  • Figure 14 shows a program product 300 provided by this embodiment for implementing the above method, which can use a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer.
  • a readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.
  • Program product 300 may take the form of any combination of one or more readable media.
  • the readable medium may be a readable signal medium or a readable storage medium.
  • the readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: electrical connection with one or more conductors, portable disk, hard disk, random access memory (RAM), read only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.
  • a computer-readable storage medium may include a data signal propagated in baseband or as part of a carrier wave carrying the readable program code therein. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above.
  • a readable storage medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transport the program for use by or in connection with an instruction execution system, apparatus, or device.
  • Program code contained on a readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.
  • Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., as well as conventional procedural programming languages. Such as "C" language or similar programming language.
  • the program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server execute on.
  • the remote computing devices may be connected over any kind of network, including a local area network (LAN) or a wide area network (WAN).
  • AN local area network
  • AN local area network
  • AN connected to a user computing device, or may be connected to an external computing device (eg, via an Internet service provider).

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Neurology (AREA)
  • Neurosurgery (AREA)
  • Psychology (AREA)
  • Cardiology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Electrotherapy Devices (AREA)
  • Image Analysis (AREA)

Abstract

本申请提供了一种刺激电极导线的成像识别方法及相关装置，装置包括：图像采集模块，用于利用图像采集设备实时采集刺激电极导线的目标图像；检测结果模块，用于获取目标图像对应的标记检测结果，标记检测结果包括一个或多个标签及其置信度和位置信息；置信度判断模块，用于判断每个标签的置信度是否满足预设条件；当一个或多个标签的置信度不满足预设条件时，重新调用图像采集模块；当所有标签的置信度均满足预设条件时，调用结果输出模块；结果输出模块，用于将标记检测结果输出至预设的用户设备。通过成像识别装置，以获得更精确的刺激电极导线识别结果。

Description

刺激电极导线的成像识别方法及相关装置
本申请要求于2022年03月31日提交的申请号为202210342589.8的中国专利的优先权，上述中国专利通过全文引用的形式并入。
技术领域
本申请涉及可植入医疗设备技术领域，例如涉及刺激电极导线的成像识别方法及相关装置。
背景技术
相关技术中，对于脑深部神经电刺激治疗(DBS，Deep Brain Stimulation)，涉及将电刺激递送到大脑的特定区域中的神经结构以激发或抑制细胞活动，可以有效处理例如慢性疼痛，帕金森病，特发性震颤等运动障碍、癫痫，以及诸如抑郁症和强迫症等精神疾病。具体地，用于施加电刺激的刺激电极作用在患者的头部并刺激大脑的指定部位，对患者大脑损伤起到治疗作用，同时刺激电极的另一端通过刺激电极导线连接神经刺激器。目前，为了满足将刺激电极导线准确植入在大脑内的期望部位处，避免对大脑的其他部位产生副作用，通常使用各种成像技术辅助刺激电极导线以使其相对精确地植入大脑内的期望部位处。所述成像技术如磁共振成像(MRI，Magnetic Resonance Imaging)、计算机断层摄影(CT，Computed Tomography)、X射线、荧光成像以及立体成像。
在具体应用中，医生期望将可以释放刺激的刺激电极导线在患者(例如大脑)内精确放置和定向，以将电刺激递送到预期部位并避免副作用。例如，期望将刺激电极导线的刺激递送到非常小的目标点位进而不会刺激邻近的其它大脑组织；如果没有精确地将刺激递送到期望目标点位，会使疗效降低，并且邻近大脑组织会接受到不必要的过量刺激，造成患者的痛苦。
因此，亟需设计一种新的刺激电极导线的成像识别装置，以辅助医生提高刺激电极导线放置到患者体内的精度。
发明内容
本申请的目的在于提供刺激电极导线的成像识别方法及相关装置，通过成像 技术直接识别设置在电极片上的标记，通过每个标签的置信度的判断，解决了对刺激电极导线识别精度低的问题。
本申请的目的采用以下技术方案实现：
第一方面，本申请提供了一种刺激电极导线的成像识别装置，多个电极片设置于所述刺激电极导线的外周面，至少部分所述电极片上分别设置有标记，所述标记用于在成像时对所述电极片进行识别；所述装置包括：图像采集模块，用于利用图像采集设备实时采集所述刺激电极导线的目标图像；检测结果模块，用于获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息；置信度判断模块，用于判断每个标签的置信度是否满足所述预设条件；当一个或多个标签的置信度不满足预设条件时，重新调用图像采集模块；当所有标签的置信度均满足所述预设条件时，调用结果输出模块；结果输出模块，用于将所述标记检测结果输出至预设的用户设备。
在一种实现方式中，所述检测结果模块包括：成像识别单元，用于利用成像识别模型对所述目标图像进行成像识别，得到所述目标图像对应的标记检测结果；其中，所述成像识别模型的训练过程如下：获取第一训练集，所述第一训练集包括多个第一训练图像及其对应的标记检测结果的标注数据；利用所述第一训练集对预设的第一深度学习模型进行训练，得到所述成像识别模型。
在一种实现方式中，在所述成像识别模型的训练过程中，所述利用所述第一训练集对预设的第一深度学习模型进行训练，包括：针对所述第一训练集中的每个第一训练图像，将所述第一训练图像输入预设的第一深度学习模型，得到与所述第一训练图像相对应的标记检测结果的预测数据；基于与所述第一训练图像相对应的标记检测结果的预测数据以及标注数据，对所述预设的第一深度学习模型的模型参数进行更新；检测是否满足预设的第一训练结束条件，如果是，则停止训练，并将训练得到的所述预设的第一深度学习模型作为所述成像识别模型，如果否，则利用下一个所述训练数据继续训练所述预设的第一深度学习模型。
在一种实现方式中，所述检测结果模块包括：目标检测单元，用于对所述目标图像进行目标检测，得到一个或多个子图及其对应的位置信息，每个子图分别对应一个标记；子图分类单元，用于对每个子图进行标记分类，得到每个子图对应的标签及其置信度；标记结果单元，用于基于每个子图对应的标签及其置信度 和位置信息，获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息。
在一种实现方式中，所述子图分类单元包括：子图分类子单元，用于利用标记分类模型对每个子图进行标记分类，得到每个子图对应的标记分类结果；其中，所述标记分类模型的训练过程如下：获取第二训练集，所述第二训练集包括多个第二训练图像及其对应的标记分类结果的标注数据；利用所述第二训练集对预设的第二深度学习模型进行训练，得到所述标记分类模型。
在一种实现方式中，在所述标记分类模型的训练过程中，所述利用所述第二训练集对预设的第二深度学习模型进行训练，包括：针对所述第二训练集中的每个第二训练图像，将所述第二训练图像输入预设的第二深度学习模型，得到与所述第二训练图像相对应的标记检测结果的预测数据；基于与所述第二训练图像相对应的标记检测结果的预测数据以及标注数据，对所述预设的第二深度学习模型的模型参数进行更新；检测是否满足预设的第二训练结束条件，如果是，则停止训练，并将训练得到的所述预设的第二深度学习模型作为所述标记分类模型，如果否，则利用下一个所述训练数据继续训练所述预设的第二深度学习模型。
在一种实现方式中，所述装置还包括：结果显示模块，用于利用所述用户设备显示所述目标图像及其对应的标记检测结果。
第二方面，本申请提供了一种刺激电极导线的成像识别方法，多个电极片设置于刺激电极导线的外周面，至少部分所述电极片上分别设置有标记，所述标记用于在成像时对所述电极片进行识别；所述方法包括：S101：利用图像采集设备实时采集所述刺激电极导线的目标图像；S102：获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息；S103：判断每个标签的置信度是否满足所述预设条件；当一个或多个标签的置信度不满足预设条件时，重新执行步骤S101以获得新的目标图像；当所有标签的置信度均满足所述预设条件时，执行步骤S104；S104：将所述标记检测结果输出至预设的用户设备。
在一种实现方式中，所述步骤S102包括：利用成像识别模型对所述目标图像进行成像识别，得到所述目标图像对应的标记检测结果；其中，所述成像识别模型的训练过程如下：获取第一训练集，所述第一训练集包括多个第一训练图像 及其对应的标记检测结果的标注数据；利用所述第一训练集对预设的第一深度学习模型进行训练，得到所述成像识别模型。
在一种实现方式中，在所述成像识别模型的训练过程中，所述利用所述第一训练集对预设的第一深度学习模型进行训练，包括：针对所述第一训练集中的每个第一训练图像，将所述第一训练图像输入预设的第一深度学习模型，得到与所述第一训练图像相对应的标记检测结果的预测数据；基于与所述第一训练图像相对应的标记检测结果的预测数据以及标注数据，对所述预设的第一深度学习模型的模型参数进行更新；检测是否满足预设的第一训练结束条件，如果是，则停止训练，并将训练得到的所述预设的第一深度学习模型作为所述成像识别模型，如果否，则利用下一个所述训练数据继续训练所述预设的第一深度学习模型。
在一种实现方式中，所述获取所述目标图像对应的标记检测结果包括：对所述目标图像进行目标检测，得到一个或多个子图及其对应的位置信息，每个子图分别对应一个标记；对每个子图进行标记分类，得到每个子图对应的标签及其置信度；基于每个子图对应的标签及其置信度和位置信息，获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息。
在一种实现方式中，所述步骤S202包括：利用标记分类模型对每个子图进行标记分类，得到每个子图对应的标记分类结果；其中，所述标记分类模型的训练过程如下：获取第二训练集，所述第二训练集包括多个第二训练图像及其对应的标记分类结果的标注数据；利用所述第二训练集对预设的第二深度学习模型进行训练，得到所述标记分类模型。
在一种实现方式中，在所述标记分类模型的训练过程中，所述利用所述第二训练集对预设的第二深度学习模型进行训练，包括：针对所述第二训练集中的每个第二训练图像，将所述第二训练图像输入预设的第二深度学习模型，得到与所述第二训练图像相对应的标记检测结果的预测数据；基于与所述第二训练图像相对应的标记检测结果的预测数据以及标注数据，对所述预设的第二深度学习模型的模型参数进行更新；检测是否满足预设的第二训练结束条件，如果是，则停止训练，并将训练得到的所述预设的第二深度学习模型作为所述标记分类模型，如果否，则利用下一个所述训练数据继续训练所述预设的第二深度学习模型。
在一种实现方式中，所述方法还包括步骤S105：利用所述用户设备显示所 述目标图像及其对应的标记检测结果。
第三方面，本申请提供了一种电子设备，用于对刺激电极导线进行成像识别，电极片设置于刺激电极导线的外周面，至少部分所述电极片上分别设置有标记，所述标记用于在成像时对所述电极片进行识别；所述电子设备包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现上述方法的步骤。
第四方面，本申请提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时实现上述方法的步骤。
采用本申请提供的刺激电极导线的成像识别方法及相关装置，至少具有以下优点：
基于实时采集的目标图像进行检测，通过预设条件对检测结果中标签的置信度进行判断(标签用于指示电极片的标识)，在每个标签的置信度不满足预设条件时重新调用图像采集模块进行图像采集、检测和判断，直至获得满足置信度条件的标记检测结果，相比相关技术，获得更精确的刺激电极导线识别结果。医生参考刺激电极导线的识别结果，无需医生进行复杂的逻辑判断，智能化程度高，即便不是经验丰富的医生也能精确地将刺激递送到期望目标点位，缩短了医生放置和定向刺激电极导线的时间，提高了医生精确地放置和定向刺激电极导线的效率，减轻了患者在医生放置和定向刺激电极导线期间的痛苦，进而提高了电刺激治疗对患者的疗效。
同时，由于标记是直接设置在电极片上的，带有标记的电极片经成像识别本身能够起到确定电极方位的作用，同时还可以用于产生刺激信号，非电极片区域不需要另外预设标记部件，可以减少刺激电极导线的制造成本和降低刺激电极导线的制造难度。
附图说明
下面结合附图和实施例对本申请进一步说明。
图1是本申请实施例提供的一种刺激电极导线的成像识别装置的结构示意图；
图2是本申请实施例提供的一种刺激电极导线的局部透视示意图；
图3是本申请实施例提供的一种刺激电极导线的平坦化状态下的部分结构 示意图；
图4是本申请实施例提供的又一种刺激电极导线的平坦化状态下的部分结构示意图；
图5是本申请实施例提供的又一种刺激电极导线的平坦化状态下的部分结构示意图；
图6是本申请实施例提供的又一种刺激电极导线的平坦化状态下的部分结构示意图；
图7是本申请实施例提供的又一种刺激电极导线的平坦化状态下的部分结构示意图；
图8是本申请实施例提供的一种检测结果模块的结构示意图；
图9是本申请实施例提供的又一种刺激电极导线的成像识别装置的结构示意图；
图10是本申请实施例提供的一种刺激电极导线的成像识别方法的流程示意图；
图11是本申请实施例提供的一种获取标记检测结果的流程示意图；
图12是本申请实施例提供的又一种刺激电极导线的成像识别方法的流程示意图；
图13是本申请实施例提供的一种电子设备的结构示意图；
图14是本申请实施例提供的一种计算机可读存储介质的结构示意图。
具体实施方式
下面，结合附图以及具体实施方式，对本申请做进一步描述，需要说明的是，在不相冲突的前提下，以下描述的各实施例之间或各技术特征之间可以任意组合形成新的实施例。
本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”、“第三”、“第四”等(如果存在)是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本申请的实施例例如能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“对应于”以及他们的任何变形，意图在于覆盖不排他的 包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
下面，首先对本申请实施例的其中一个应用领域(即植入式神经刺激器)进行简单说明。
植入式神经刺激系统主要包括植入患者体内的刺激器(即植入式神经刺激器，一种神经刺激装置)以及设置于患者体外的程控设备。相关的神经调控技术主要是通过立体定向手术在生物体的组织的特定部位(即靶点)植入电极，并由植入患者体内的刺激器经电极向靶点发放电脉冲，调控相应神经结构和网络的电活动及其功能，从而改善症状、缓解病痛。其中，刺激器可以是植入式神经电刺激装置、植入式心脏电刺激系统(又称心脏起搏器)、植入式药物输注装置(Implantabl e Drug Delivery System，简称I DDS)和导线转接装置中的任意一种。植入式神经电刺激装置例如是脑深部电刺激系统(Deep Brain Stimulation，简称DBS)、植入式脑皮层刺激系统(Cortical Nerve Stimulation，简称CNS)、植入式脊髓电刺激系统(Spinal Cord Stimulation，简称SCS)、植入式骶神经电刺激系统(Sacral Nerve Stimulation，简称SNS)、植入式迷走神经电刺激系统(Vagus Nerve Stimulation，简称VNS)等。
刺激器可以包括IPG(implantable pulse generator，植入式脉冲发生器)、延伸导线和刺激电极导线，IPG设置于患者体内，接收程控设备发送的程控指令，依靠密封电池和电路向体内组织提供可控制的电刺激能量，通过植入的延伸导线和刺激电极导线，为体内组织的特定区域递送一路或两路可控制的特定电刺激。延伸导线配合IPG使用，作为电刺激信号的传递媒体，将IPG产生的电刺激信号，传递给刺激电极导线。刺激电极导线可以是神经刺激电极，刺激电极导线通过多个电极触点，向体内组织的特定区域递送电刺激。刺激器设置有单侧或双侧的一路或多路刺激电极导线，刺激电极导线上设置有多个电极触点，电极触点可以均匀排列或者非均匀排列在刺激电极导线的周向上。作为一个示例，电极触点可以以4行3列的阵列(共计12个电极触点)排列在刺激电极导线的周向上。电极触点可以包括刺激触点和/或信号采集触点。电极触点例如可以采用片状、环状、点状等形状，电极触点可以是上述形状的电极片。
在一些可能的方式中，受刺激的体内组织可以是患者的脑组织，受刺激的部位可以是脑组织的特定部位。当患者的疾病类型不同时，受刺激的部位一般来说是不同的，所使用的刺激触点(单源或多源)的数量、一路或多路(单通道或多通道)特定电刺激信号的运用以及刺激参数数据也是不同的。本申请实施例对适用的疾病类型不做限定，其可以是脑深部刺激(DBS)、脊髓刺激(SCS)、骨盆刺激、胃刺激、外周神经刺激、功能性电刺激所适用的疾病类型。
本申请实施例中，程控设备和刺激器建立程控连接时，可以利用程控设备调整刺激器的刺激参数(不同的刺激参数所对应的电刺激信号不同)，也可以通过刺激器感测患者脑深部的生物电活动以采集得到电生理信号，并可以通过所采集到的电生理信号来继续调节刺激器的电刺激信号的刺激参数。
刺激参数可以包括：频率(例如是单位时间1s内的电刺激脉冲信号个数，单位为Hz)、脉宽(每个脉冲的持续时间，单位为μs)、幅值(一般用电压表述，即每个脉冲的强度，单位为V)、时序(例如可以是连续或者触发)、刺激模式(包括电流模式、电压模式、定时刺激模式和循环刺激模式中的一种或多种)、医生控制上限及下限(医生可调节的范围)和患者控制上限及下限(患者可自主调节的范围)中的一种或多种。
在一个具体应用场景中，可以在电流模式或者电压模式下对刺激器的各刺激参数进行调节。
程控设备可以是医生程控设备(即医生使用的程控设备)或者患者程控设备(即患者使用的程控设备)。医生程控设备例如可以是搭载有程控软件的平板电脑、笔记本电脑、台式计算机、手机等智能终端设备。患者程控设备例如可以是搭载有程控软件的平板电脑、笔记本电脑、台式计算机、手机等智能终端设备，患者程控设备还可以是其他具有程控功能的电子设备(例如是具有程控功能的充电器、数据采集设备)。
本申请实施例对医生程控设备和刺激器的数据交互不进行限制，当医生远程程控时，医生程控设备可以通过服务器、患者程控设备与刺激器进行数据交互。当医生线下和患者面对面进行程控时，医生程控设备可以通过患者程控设备与刺激器进行数据交互，医生程控设备还可以直接与刺激器进行数据交互。
在一些可能的方式中，患者程控设备可以包括(与服务器通信的)主机和(与 刺激器通信的)子机，主机和子机可通信的连接。其中，医生程控设备可以通过3G/4G/5G网络与服务器进行数据交互，服务器可以通过3G/4G/5G网络与主机进行数据交互，主机可以通过蓝牙协议/WIFI协议/USB协议与子机进行数据交互，子机可以通过401MHz-406MHz工作频段/2.4GHz-2.48GHz工作频段与刺激器进行数据交互，医生程控设备可以通过401MHz-406MHz工作频段/2.4GHz-2.48GHz工作频段与刺激器直接进行数据交互。
公开号为CN112604159A的中国专利，公开了一种分片式电极，可以通过额外设置标记来识别电极方位，通过预先定义的标记方向与电极刺激片的对应关系来判定电极的位置及方向，但是这样的方式需要医生具有很强的逻辑判断能力，判断过程中容易出现错误，给患者带来不必要的痛苦。
参见图1，本申请实施例提供了一种刺激电极导线的成像识别装置，多个电极片设置于刺激电极导线的外周面，至少部分所述电极片上分别设置有标记，所述标记用于在成像时对所述电极片进行识别。
上述装置可以实现一种与相关技术不同的识别方式，(使用例如柔性薄膜电路的制造工艺)直接利用电极片上设置标记的结构，通过成像技术得到的图像来识别刺激电极导线的中不同标记的电极片的位置及方向。也就是说，相比相关技术中的额外设置标记，成像时可以利用电极片上分别设置的标记识别方位。
参见图3至图7，为刺激电极导线的平坦化状态下的示意图，电极片上设置标记可用于对各个电极片进行区分识别。例如，电极片中每个电极片的形状差异作为标记，或电极片上的连接点位置不同作为区分不同电极片的标记(图3、图4、图6和图7)，或电极片上的连接点的形状不同作为区分不同电极片的标记(图5)，或者以上电极片不同的标记设置方式的结合。医生可以通过在患者的脑深部区域植入刺激电极导线，刺激电极导线的表面可以设有多个规则矩阵排列的电极片，通过神经刺激器使刺激电极导线上设置的多个电极片释放刺激。
当电极片规则排列时，可以利用在不同行分别标记不同列电极的方式来对电极片进行识别。在一个实施例中，电极片有4行，每行3列(输出标签12个，即1号电极片～12号电极片)，在一个实施例中，可以标记第一行第一列电极片和第二行第二列电极片；在另一个实施例中，可以标记第二行第三列电极片和第四行第二列电极片；在又一个实施例中，可以标记第一行第三列电极片和第三行 第一列电极片；在又一个实施例中，可以标记第一行第二列电极片和第四行第三列电极片。
在另一个实施例中，电极片有5行，每行4列(输出标签20个，即1号电极片～20号电极片)，在一个实施例中，可以标记第一行第一列电极片和第二行第二列电极片；在另一个实施例中，可以标记第二行第三列电极片和第四行第二列电极片；在又一个实施例中，可以标记第一行第三列电极片和第三行第一列电极片；在又一个实施例中，可以标记第一行第二列电极片和第五行第三列电极片。
在又一个实施例中，电极片有5行，每行4列(输出标签20个，即1号电极片～20号电极片)，在一个实施例中，可以标记第一行第一列电极片、第二行第二列电极片和第三行第三列电极片；在另一个实施例中，可以标记第二行第三列电极片、第四行第二列电极片、第五行第四列电极片；在又一个实施例中，可以标记第一行第三列电极片、第三行第一列电极片和第四行第二列电极片；在又一个实施例中，可以标记第一行第二列电极片、第四行第三列电极片和第五行第一列电极片。
其中，对刺激电极导线通过成像识别装置进行识别操作的对象可以是患者的首诊医生、会诊专家等对患者进行治疗的人。本申请实施例中的患者，可以是帕金森患者，或者抑郁症患者、强迫症患者等精神疾病类患者，还可以是药物成瘾症患者或者戒毒人员。
通过刺激电极导线，可以将刺激器的电刺激递送到人体特定区域中施加刺激治疗。在本实施例中，刺激电极导线可以对大脑的的神经结构释放电刺激以激发或抑制细胞活动，可以有效地处理例如痉挛疾病(例如，癫痫)、疼痛、偏头痛、精神疾病(例如，重度抑郁症(MDD))、躁郁症、焦虑症、创伤后压力心理障碍症、轻郁症、强迫症(OCD)、行为障碍、情绪障碍、记忆障碍、心理状态障碍、移动障碍(例如，特发性震颤或帕金森氏病)、亨廷顿病、阿尔茨海默症、药物成瘾症、自闭症或其他神经学或精神科疾病和损害。
所述装置包括图像采集模块101、检测结果模块102、置信度判断模块103和结果输出模块104。
图像采集模块101，用于利用图像采集设备实时采集所述刺激电极导线的目标图像。图像采集设备可以包括能够实现如磁共振成像(MRI)、计算机断层摄 影(CT)、X射线、荧光成像以及立体成像等成像技术的成像设备。参见图2，为通过X射线所获得的一种刺激电极导线的局部透视示意图。
检测结果模块102，用于获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息。位置信息可以是标记对应的坐标值，通过位置信息可以准确获得对应标记在目标图像上的位置。
置信度判断模块103，用于判断每个标签的置信度是否满足所述预设条件；当一个或多个标签的置信度不满足预设条件时，重新调用图像采集模块；当所有标签的置信度均满足所述预设条件时，调用结果输出模块。
结果输出模块104，用于将所述标记检测结果输出至预设的用户设备。其中，用于接收标记检测结果的用户设备可以采用相关技术中存在的程控器，即用户设备可以是单独的硬件设备，该硬件设备是一种能够与刺激器通过无线网络或有线网络进行数据交互的电子设备，例如是平板电脑、计算机、手机或者智能穿戴设备等，用户可以利用这种程控设备进行标记检测结果的接收。总的来说，用户设备中搭载了计算机程序(即软件)，所述计算机程序被处理器执行时能够实现本申请实施例中标记检测结果的接收的作用。
本申请对预设条件不做限定，预设条件例如是数值范围预设条件，在一个实施例中，所述预设条件是置信度不小于预设置信度。预设置信度例如是0.95、0.97、0.94、0.98。
本申请可以针对不同患者设置相同或者不同的预设条件。在一个实施例中，为不同患者设置相同的预设条件，即置信度不小于0.96。
在另一个实施例中，可以根据患者病症的不同或治疗阶段的不同设置差异化的预设条件，以实现对患者人性化、定制性的诊断和治疗。例如，医生通过上述刺激电极导线的成像识别装置对患者张三、李四和王五植入体内的刺激电极导线的电极片位置和方向进行判断。参见下表1，具体判断情况如下。
表1

一般而言，医生的就诊时间和患者的治疗时间都是很宝贵的，所以医生和患者更期望将刺激电极导线定位到有效的期望目标点，以减少定位效果不佳造成的重新诊治。因此通过判断每个标签的置信度是否满足所述预设条件，即便一个标签的置信度不满足预设条件时，也会重新调用图像采集模块，为后期医生对患者通过刺激电极导线向患者治疗时将刺激递送到非常小的目标点位且不刺激邻近大脑组织，节省更多的治疗时间，进而减轻了患者治疗中的不适。
由此，基于实时采集的目标图像进行检测，通过预设条件对检测结果中标签的置信度进行判断，标签用于指示电极片的标识，在每个标签的置信度不满足预设条件时重新调用图像采集模块101进行图像采集、检测和判断，直至获得满足置信度条件的标记检测结果，相比相关技术，能获得刺激电极导线识别结果更精确，智能化程度高。医生参考刺激电极导线的识别结果，无需医生进行复杂的逻辑判断，即便不是经验丰富的医生也能精确地将刺激递送到期望目标点位，缩短了医生放置和定向刺激电极导线的时间，提高了医生精确地放置和定向刺激电极导线的效率，减轻了患者在医生放置和定向刺激电极导线期间的痛苦，提高对患者电刺激的疗效。
同时，由于标记是直接设置在电极片上的，带有标记的电极片经成像识别本身能够起到确定电极方位的作用，同时还可以用于产生刺激信号，非电极片区域不需要另外预设标记部件，可以减少刺激电极导线的制造成本和降低刺激电极导线的制造难度。
在一些可选的实施例实施例中，所述多个电极片中任意两个电极片之间彼此 绝缘，所述多个电极片包括多个刺激电极片和多个采集电极片。此时，刺激电极导线不仅可以用于释放电刺激能量，还可以用于采集生物体内组织的生物电信号。
参见图9，所述装置还可以包括结果显示模块105，所述结果显示模块105用于利用所述用户设备显示所述目标图像及其对应的标记检测结果。其中，结果显示模块105可以包括显示器、投影仪等提供显示功能的设备模块。
其中，标记检测结果和目标图像在显示模块显示可以理解为，在显示模块的界面上显示目标图像，位置信息用于将标记检测结果和目标图像上显示的图像进行对应，在目标图像所显示的电极片上可以显示其所对应的标签、置信度等。其中，所显示的标签可以是电极片1、电极片2……电极片N等，所标注的置信度可以是0.91、0.94、0.98等。
由此，通过结果显示模块105的设置，图像采集设备实时采集的刺激电极导线的目标图像、刺激电极导线中电极片的标记的标签及其置信度和位置信息可以直观的在显示模块上展示，一方面使刺激电极导线实时信息更便于医生参考；另一方面使患者或其家人直观了解诊治的进程，缓解患者及其家人的紧张情绪，促进医患之间保持信任和理解的关系。
在一些实施例中，所述检测结果模块可以包括成像识别单元。所述成像识别单元可以用于利用成像识别模型对所述目标图像进行成像识别，得到所述目标图像对应的标记检测结果。
其中，所述成像识别模型的训练过程如下：
获取第一训练集，所述第一训练集包括多个第一训练图像及其对应的标记检测结果的标注数据；
利用所述第一训练集对预设的第一深度学习模型进行训练，得到所述成像识别模型。
由此，通过成像识别模型，相比传统人工的对目标图像进行识别，智能化程度高；将训练得到的成像识别模型应用到实际场景中刺激电极导线的成像识别上，识别准确度高。
具体地，在所述成像识别模型的训练过程中，所述利用所述第一训练集对预设的第一深度学习模型进行训练，可以包括：
针对所述第一训练集中的每个第一训练图像，将所述第一训练图像输入预设 的第一深度学习模型，得到与所述第一训练图像相对应的标记检测结果的预测数据；基于与所述第一训练图像相对应的标记检测结果的预测数据以及标注数据，对所述预设的第一深度学习模型的模型参数进行更新；检测是否满足预设的第一训练结束条件，如果是，则停止训练，并将训练得到的所述预设的第一深度学习模型作为所述成像识别模型，如果否，则利用下一个所述训练数据继续训练所述预设的第一深度学习模型。
利用第一训练集对预设的第一深度学习模型进行训练，可以得到训练好的成像识别模型，成像识别模型可以由大量的训练数据训练得到，能够针对多种输入数据预测得到相应的标记检测结果，适用范围广，智能化水平高。通过设计，建立适量的神经元计算节点和多层运算层次结构，选择合适的输入层和输出层，就可以得到预设的第一深度学习模型，通过该预设的第一深度学习模型的学习和调优，建立起从输入到输出的函数关系，虽然不能100％找到输入与输出的函数关系，但是可以尽可能地逼近现实的关联关系，由此训练得到的成像识别模型，可以实现对成像识别的自我诊断功能，且诊断结果可靠性高。
由此，相较于传统的识别系统，往往仅通过分析和与现有的图像及其标记检测结果进行比对，本申请利用第一训练集对第一深度学习模型进行训练，使得最终形成的成像识别模型的识别效果与实际的成像结果更加匹配，用户得到更为满意的刺激电极导线的成像识别结果，提升用户体验。
参见图8，在一些实施例中，所述检测结果模块还可以包括目标检测单元201、子图分类单元202和标记结果单元203。
目标检测单元201，用于对所述目标图像进行目标检测，得到一个或多个子图及其对应的位置信息，每个子图分别对应一个标记。通过子图可以体现出目标图像的主要属性，可实现对图数据的压缩，去噪等处理。
子图分类单元202，用于对每个子图进行标记分类，得到每个子图对应的标签及其置信度。
标记结果单元203，用于基于每个子图对应的标签及其置信度和位置信息，获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息。
由此，基于目标检测单元、子图分类单元和标记结果单元，将目标图像中的 标记通过每个子图进行标记分类，以获取每个子图对应的标签及其置信度，进而获取目标图像对应的包括所有标签及其置信度和位置信息的标记检测结果，智能化程度高。
在一些实施例中，所述子图分类单元可以包括子图分类子单元，所述子图分类子单元可以用于利用标记分类模型对每个子图进行标记分类，得到每个子图对应的标记分类结果。
其中，所述标记分类模型的训练过程如下：
获取第二训练集，所述第二训练集包括多个第二训练图像及其对应的标记分类结果的标注数据；
利用所述第二训练集对预设的第二深度学习模型进行训练，得到所述标记分类模型。
由此，通过子图分类子单元对每个子图进行标记分类，得到每个子图对应的标记分类结果，用于训练标记分类模型，可以提高标记分类模型的鲁棒性，有效降低其拟合风险。
具体地，在所述标记分类模型的训练过程中，所述利用所述第二训练集对预设的第二深度学习模型进行训练，可以包括以下步骤：
针对所述第二训练集中的每个第二训练图像，将所述第二训练图像输入预设的第二深度学习模型，得到与所述第二训练图像相对应的标记检测结果的预测数据；
基于与所述第二训练图像相对应的标记检测结果的预测数据以及标注数据，对所述预设的第二深度学习模型的模型参数进行更新；
检测是否满足预设的第二训练结束条件，如果是，则停止训练，并将训练得到的所述预设的第二深度学习模型作为所述标记分类模型，如果否，则利用下一个所述训练数据继续训练所述预设的第二深度学习模型。
由此，训练结束的第二训练结束条件可基于实际需求配置，训练得到的标记分类模型具有较强的鲁棒性和较低的过拟合风险。
利用第二训练集对预设的第二深度学习模型进行训练，可以得到训练好的标记分类模型，标记分类模型可以由大量的训练数据训练得到，能够针对多种输入数据预测得到相应的标记检测结果，适用范围广，智能化水平高。通过设计，建 立适量的神经元计算节点和多层运算层次结构，选择合适的输入层和输出层，就可以得到预设的第二深度学习模型，通过该预设的第二深度学习模型的学习和调优，建立起从输入到输出的函数关系，虽然不能100％找到输入与输出的函数关系，但是可以尽可能地逼近现实的关联关系，由此训练得到的标记分类模型，可以实现对成像识别的自我诊断功能，且诊断结果可靠性高。
参见图10，本申请实施例还提供了一种刺激电极导线的成像识别方法。由于刺激电极导线的成像识别方法所起的作用与上述刺激电极导线的成像识别装置相同或相似，在此不予赘述。
其中，多个电极片设置于刺激电极导线的外周面，至少部分所述电极片上分别设置有标记，所述标记用于在成像时对所述电极片进行识别。
所述方法包括步骤S101至S104。
步骤S101：利用图像采集设备实时采集所述刺激电极导线的目标图像。
步骤S102：获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息。
步骤S103：判断每个标签的置信度是否满足所述预设条件；当一个或多个标签的置信度不满足预设条件时，重新执行步骤S101以获得新的目标图像；当所有标签的置信度均满足所述预设条件时，执行步骤S104。
步骤S104：将所述标记检测结果输出至预设的用户设备。
在一些实施例中，所述步骤S102可以包括：利用成像识别模型对所述目标图像进行成像识别，得到所述目标图像对应的标记检测结果。
其中，所述成像识别模型的训练过程如下：获取第一训练集，所述第一训练集包括多个第一训练图像及其对应的标记检测结果的标注数据；利用所述第一训练集对预设的第一深度学习模型进行训练，得到所述成像识别模型。
在一些实施例中，在所述成像识别模型的训练过程中，所述利用所述第一训练集对预设的第一深度学习模型进行训练，可以包括：
针对所述第一训练集中的每个第一训练图像，将所述第一训练图像输入预设的第一深度学习模型，得到与所述第一训练图像相对应的标记检测结果的预测数据；基于与所述第一训练图像相对应的标记检测结果的预测数据以及标注数据，对所述预设的第一深度学习模型的模型参数进行更新；检测是否满足预设的第一 训练结束条件，如果是，则停止训练，并将训练得到的所述预设的第一深度学习模型作为所述成像识别模型，如果否，则利用下一个所述训练数据继续训练所述预设的第一深度学习模型。
参见图11，在一些实施例中，所述获取所述目标图像对应的标记检测结果可以包括步骤S201～步骤步骤S203。
步骤S201：对所述目标图像进行目标检测，得到一个或多个子图及其对应的位置信息，每个子图分别对应一个标记。
步骤S202：对每个子图进行标记分类，得到每个子图对应的标签及其置信度。
步骤S203：基于每个子图对应的标签及其置信度和位置信息，获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息。
在一些实施例中，所述步骤S202可以包括：
利用标记分类模型对每个子图进行标记分类，得到每个子图对应的标记分类结果。
其中，所述标记分类模型的训练过程如下：
获取第二训练集，所述第二训练集包括多个第二训练图像及其对应的标记分类结果的标注数据；
利用所述第二训练集对预设的第二深度学习模型进行训练，得到所述标记分类模型。
在一些实施例中，在所述标记分类模型的训练过程中，所述利用所述第二训练集对预设的第二深度学习模型进行训练，可以包括：
针对所述第二训练集中的每个第二训练图像，将所述第二训练图像输入预设的第二深度学习模型，得到与所述第二训练图像相对应的标记检测结果的预测数据；
基于与所述第二训练图像相对应的标记检测结果的预测数据以及标注数据，对所述预设的第二深度学习模型的模型参数进行更新；
检测是否满足预设的第二训练结束条件，如果是，则停止训练，并将训练得到的所述预设的第二深度学习模型作为所述标记分类模型，如果否，则利用下一 个所述训练数据继续训练所述预设的第二深度学习模型。
参见图12，在一些实施例中，所述方法还可以包括步骤S105。
步骤S105：利用所述用户设备显示所述目标图像及其对应的标记检测结果。
参见图13，本申请实施例还提供了一种电子设备200，用于对刺激电极导线进行成像识别，所述刺激电极导线设置于刺激电极导线的外周面，至少部分所述电极片上分别设置有标记，所述标记用于在成像时对所述电极片进行识别。
电子设备200包括一个或多个存储器210、一个或多个处理器220以及连接不同平台系统的总线230。
存储器210可以包括易失性存储器形式的可读介质，例如随机存取存储器(R AM)211和/或高速缓存存储器212，还可以进一步包括只读存储器(ROM)213。
其中，存储器210还存储有计算机程序，计算机程序可以被处理器220执行，使得处理器220执行本申请实施例中上述方法的步骤，其具体实现方式与上述方法实施例中记载的实施方式、所达到的技术效果一致，部分内容不再赘述。
存储器210还可以包括具有一个或多个程序模块215的实用工具214，这样的程序模块215包括但不限于：操作系统、一个或者多个应用程序、其它程序模块以及程序数据，这些示例中的每一个或某种组合中可能包括网络环境的实现。
相应的，处理器220可以执行上述计算机程序，以及可以执行实用工具214。
总线230可以为表示几类总线结构中的一种或多种，包括存储器总线或者存储器控制器、外围总线、图形加速端口、处理器或者使用多种总线结构中的任意总线结构的局域总线。
电子设备200也可以与一个或多个外部设备240例如键盘、指向设备、蓝牙设备等通信，还可与一个或者多个能够与该电子设备200交互的设备通信，和/或与使得该电子设备200能与一个或多个其它计算设备进行通信的任何设备(例如路由器、调制解调器等)通信。这种通信可以通过输入输出接口250进行。并且，电子设备200还可以通过网络适配器260与一个或者多个网络(例如局域网(L AN)，广域网(WAN)和/或公共网络，例如因特网)通信。网络适配器260可以通过总线230与电子设备200的其它模块通信。应当明白，尽管图中未示出，可以结合电子设备200使用其它硬件和/或软件模块，包括但不限于：微代码、设备驱动器、冗余处理器、外部磁盘驱动阵列、RAID系统、磁带驱动器以及数据备 份存储平台等。
本申请实施例还提供了一种计算机可读存储介质，其具体实现方式与上述方法的实施例中记载的实施方式、所达到的技术效果一致，部分内容不再赘述。
该计算机可读存储介质用于存储计算机程序；所述计算机程序被执行时实现本申请实施例中上述方法的步骤。
图14示出了本实施例提供的用于实现上述方法的程序产品300，其可以采用便携式紧凑盘只读存储器(CD-ROM)并包括程序代码，并可以在终端设备，例如个人电脑上运行。然而，本发明的程序产品300不限于此，在本文件中，可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。程序产品300可以采用一个或多个可读介质的任意组合。可读介质可以是可读信号介质或者可读存储介质。可读存储介质例如可以为但不限于电、磁、光、电磁、红外线、或半导体的系统、装置或器件，或者任意以上的组合。可读存储介质的更具体的例子(非穷举的列表)包括：具有一个或多个导线的电连接、便携式盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。
计算机可读存储介质可以包括在基带中或者作为载波一部分传播的数据信号，其中承载了可读程序代码。这种传播的数据信号可以采用多种形式，包括但不限于电磁信号、光信号或上述的任意合适的组合。可读存储介质还可以是可读存储介质以外的任何可读介质，该可读介质可以发送、传播或者传输用于由指令执行系统、装置或者器件使用或者与其结合使用的程序。可读存储介质上包含的程序代码可以用任何适当的介质传输，包括但不限于无线、有线、光缆、RF等等，或者上述的任意合适的组合。可以以一种或多种程序设计语言的任意组合来编写用于执行本发明操作的程序代码，程序设计语言包括面向对象的程序设计语言诸如Java、C++等，还包括常规的过程式程序设计语言诸如“C”语言或类似的程序设计语言。程序代码可以完全地在用户计算设备上执行、部分地在用户设备上执行、作为一个独立的软件包执行、部分在用户计算设备上部分在远程计算设备上执行、或者完全在远程计算设备或服务器上执行。在涉及远程计算设备的情形中，远程计算设备可以通过任意种类的网络，包括局域网(LAN)或广域网(W  AN)，连接到用户计算设备，或者，可以连接到外部计算设备(例如利用因特网服务提供商来通过因特网连接)。

Claims (10)

  1. 一种刺激电极导线的成像识别装置，多个电极片设置于刺激电极导线的外周面，至少部分所述电极片上分别设置有标记，所述标记用于在成像时对所述电极片进行识别；
    所述装置包括：
    图像采集模块，用于利用图像采集设备实时采集所述刺激电极导线的目标图像；
    检测结果模块，用于获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息；
    置信度判断模块，用于判断每个标签的置信度是否满足预设条件；当一个或多个标签的置信度不满足所述预设条件时，重新调用图像采集模块；当所有标签的置信度均满足所述预设条件时，调用结果输出模块；
    结果输出模块，用于将所述标记检测结果输出至预设的用户设备。
  2. 根据权利要求1所述的刺激电极导线的成像识别装置，其中，所述检测结果模块包括：
    成像识别单元，用于利用成像识别模型对所述目标图像进行成像识别，得到所述目标图像对应的标记检测结果；
    其中，所述成像识别模型的训练过程如下：
    获取第一训练集，所述第一训练集包括多个第一训练图像及其对应的标记检测结果的标注数据；
    利用所述第一训练集对预设的第一深度学习模型进行训练，得到所述成像识别模型。
  3. 根据权利要求2所述的刺激电极导线的成像识别装置，其中，在所述成像识别模型的训练过程中，所述利用所述第一训练集对预设的第一深度学习模型进行训练，包括：
    针对所述第一训练集中的每个第一训练图像，将所述第一训练图像输入预设的第一深度学习模型，得到与所述第一训练图像相对应的标记检测结果的预测数据；
    基于与所述第一训练图像相对应的标记检测结果的预测数据以及标注数据，对所述预设的第一深度学习模型的模型参数进行更新；
    检测是否满足预设的第一训练结束条件，如果是，则停止训练，并将训练得到的所述预设的第一深度学习模型作为所述成像识别模型，如果否，则利用下一个所述训练数据继续训练所述预设的第一深度学习模型。
  4. 根据权利要求1所述的刺激电极导线的成像识别装置，其中，所述检测结果模块包括：
    目标检测单元，用于对所述目标图像进行目标检测，得到一个或多个子图及其对应的位置信息，每个子图分别对应一个标记；
    子图分类单元，用于对每个子图进行标记分类，得到每个子图对应的标签及其置信度；
    标记结果单元，用于基于每个子图对应的标签及其置信度和位置信息，获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息。
  5. 根据权利要求4所述的刺激电极导线的成像识别装置，其中，所述子图分类单元包括：
    子图分类子单元，用于利用标记分类模型对每个子图进行标记分类，得到每个子图对应的标记分类结果；
    其中，所述标记分类模型的训练过程如下：
    获取第二训练集，所述第二训练集包括多个第二训练图像及其对应的标记分类结果的标注数据；
    利用所述第二训练集对预设的第二深度学习模型进行训练，得到所述标记分类模型。
  6. 根据权利要求5所述的刺激电极导线的成像识别装置，其中，在所述标记分类模型的训练过程中，所述利用所述第二训练集对预设的第二深度学习模型进行训练，包括：
    针对所述第二训练集中的每个第二训练图像，将所述第二训练图像输入预设的第二深度学习模型，得到与所述第二训练图像相对应的标记检测结果的预测数 据；
    基于与所述第二训练图像相对应的标记检测结果的预测数据以及标注数据，对所述预设的第二深度学习模型的模型参数进行更新；
    检测是否满足预设的第二训练结束条件，如果是，则停止训练，并将训练得到的所述预设的第二深度学习模型作为所述标记分类模型，如果否，则利用下一个所述训练数据继续训练所述预设的第二深度学习模型。
  7. 根据权利要求1所述的刺激电极导线的成像识别装置，其中，所述装置还包括：
    结果显示模块，用于利用所述用户设备显示所述目标图像及其对应的标记检测结果。
  8. 一种刺激电极导线的成像识别方法，多个电极片设置于刺激电极导线的外周面，至少部分所述电极片上分别设置有标记，所述标记用于在成像时对所述电极片进行识别；
    所述方法包括：
    S101：利用图像采集设备实时采集所述刺激电极导线的目标图像；
    S102：获取所述目标图像对应的标记检测结果，所述标记检测结果包括一个或多个标签及其置信度和位置信息；
    S103：判断每个标签的置信度是否满足预设条件；当一个或多个标签的置信度不满足所述预设条件时，重新执行步骤S101以获得新的目标图像；当所有标签的置信度均满足所述预设条件时，执行步骤S104；
    S104：将所述标记检测结果输出至预设的用户设备。
  9. 一种电子设备，用于对刺激电极导线进行成像识别，多个电极片设置于刺激电极导线的外周面，至少部分所述电极片上分别设置有标记，所述标记用于在成像时对所述电极片进行识别；
    所述电子设备包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现权利要求8所述方法的步骤。
  10. 一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时实现权利要求8所述方法的步骤。
PCT/CN2023/080446 2022-03-31 2023-03-09 刺激电极导线的成像识别方法及相关装置 WO2023185410A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210342589.8A CN114712712A (zh) 2022-03-31 2022-03-31 刺激电极导线的成像识别方法及相关装置
CN202210342589.8 2022-03-31

Publications (1)

Publication Number Publication Date
WO2023185410A1 true WO2023185410A1 (zh) 2023-10-05

Family

ID=82242147

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/080446 WO2023185410A1 (zh) 2022-03-31 2023-03-09 刺激电极导线的成像识别方法及相关装置

Country Status (2)

Country Link
CN (1) CN114712712A (zh)
WO (1) WO2023185410A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118266949A (zh) * 2024-06-03 2024-07-02 之江实验室 一种基于深度学习的脑深部锋电位信号检测方法和装置

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114712712A (zh) * 2022-03-31 2022-07-08 苏州景昱医疗器械有限公司 刺激电极导线的成像识别方法及相关装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011112843A1 (en) * 2010-03-12 2011-09-15 Inspire Medical Systems, Inc. Method and system for identifying a location for nerve stimulation
CN103027673A (zh) * 2011-09-30 2013-04-10 财团法人交大思源基金会 刺激目标范围标定方法
CN112604159A (zh) * 2020-12-11 2021-04-06 苏州景昱医疗器械有限公司 分片式电极及其标识方法
CN113240125A (zh) * 2021-01-13 2021-08-10 深延科技(北京)有限公司 模型训练方法及装置、标注方法及装置、设备及存储介质
CN113244533A (zh) * 2021-06-24 2021-08-13 景昱医疗器械(长沙)有限公司 参数调整方法、装置、电子设备及计算机可读存储介质
CN114712712A (zh) * 2022-03-31 2022-07-08 苏州景昱医疗器械有限公司 刺激电极导线的成像识别方法及相关装置

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111210024B (zh) * 2020-01-14 2023-09-15 深圳供电局有限公司 模型训练方法、装置、计算机设备和存储介质
CN113239950A (zh) * 2021-01-13 2021-08-10 深延科技(北京)有限公司 标注方法、装置、电子设备及存储介质
CN113180701B (zh) * 2021-07-01 2024-06-25 中国人民解放军军事科学院军事医学研究院 一种用于图像标签标注的脑电信号深度学习方法
CN113656706A (zh) * 2021-08-31 2021-11-16 平安医疗健康管理股份有限公司 基于多模态深度学习模型的信息推送方法及装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011112843A1 (en) * 2010-03-12 2011-09-15 Inspire Medical Systems, Inc. Method and system for identifying a location for nerve stimulation
CN103027673A (zh) * 2011-09-30 2013-04-10 财团法人交大思源基金会 刺激目标范围标定方法
CN112604159A (zh) * 2020-12-11 2021-04-06 苏州景昱医疗器械有限公司 分片式电极及其标识方法
CN113240125A (zh) * 2021-01-13 2021-08-10 深延科技(北京)有限公司 模型训练方法及装置、标注方法及装置、设备及存储介质
CN113244533A (zh) * 2021-06-24 2021-08-13 景昱医疗器械(长沙)有限公司 参数调整方法、装置、电子设备及计算机可读存储介质
CN114712712A (zh) * 2022-03-31 2022-07-08 苏州景昱医疗器械有限公司 刺激电极导线的成像识别方法及相关装置

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118266949A (zh) * 2024-06-03 2024-07-02 之江实验室 一种基于深度学习的脑深部锋电位信号检测方法和装置

Also Published As

Publication number Publication date
CN114712712A (zh) 2022-07-08

Similar Documents

Publication Publication Date Title
WO2023185410A1 (zh) 刺激电极导线的成像识别方法及相关装置
US11648398B2 (en) Therapy program selection for electrical stimulation therapy based on a volume of tissue activation
US20130023950A1 (en) Method and system to facilitate neurostimulator programming based on pre-existing therapy profiles
WO2024021960A1 (zh) 脉冲发生器、医疗系统及计算机可读存储介质
US20130268019A1 (en) Electrical stimulation programming
US20170252570A1 (en) Systems and methods for communication during remote programming
CN105682734A (zh) 用于并入来自图像的引线信息的系统
CN102573989A (zh) 在植入式医疗设备中存储治疗区域的图像
US9604067B2 (en) Techniques and methods for storing and transferring registration, atlas, and lead information between medical devices
WO2023005353A1 (zh) 基于多模态数据的配置信息获取装置及相关设备
WO2024001723A1 (zh) 控制设备、医疗系统及计算机可读存储介质
WO2023071378A1 (zh) 植入式神经刺激器和植入式神经刺激系统
US11400294B2 (en) Method and apparatus for neurostimulation device therapy management
WO2023124390A1 (zh) 程控设备和植入式神经刺激系统
WO2022262495A1 (zh) 家用医疗设备的控制方法及相关装置
WO2024041496A1 (zh) 充电提醒装置、植入式神经刺激系统及存储介质
WO2023241338A1 (zh) 程控设备、程控系统、电子设备和计算机可读存储介质
WO2024067449A1 (zh) 参数调节装置及其方法、程控设备、医疗系统、存储介质
WO2024083193A1 (zh) 脉冲发生器、刺激器、医疗系统及计算机可读存储介质
WO2023000787A1 (zh) 视频处理方法、装置、电子设备及计算机可读存储介质
WO2023000788A1 (zh) 参数比对方法、装置、电子设备及计算机可读存储介质
WO2016137926A1 (en) Topical nerve stimulator and sensor for control of autonomic function
WO2024001695A1 (zh) 数据采集方法、装置、系统及计算机可读存储介质
CN117065217A (zh) 电位信号采集设备及其方法、医疗系统、可读存储介质
CN116895065A (zh) 核团标注装置及方法、可穿戴的xr设备、相关装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23777788

Country of ref document: EP

Kind code of ref document: A1