Intelligent diagnosis method and system for fetal head MR image

Abstract

The application provides a fetal head MR image intelligent diagnosis method, comprising the following steps: extracting a preset image sequence from a patient DICOM image corresponding to an examination item; judging whether the image sequence is a T2WI image quality meeting a first preset condition, and if yes, defining the T2WI image as a first image; performing regional segmentation on the first image according to a first segmentation rule, and correcting the first image; defining the corrected first image as a second image; performing regional segmentation on the second image based on a second segmentation rule; analyzing the first image or the second image based on one or more preset diagnosis types and coordinate data of a region corresponding to the diagnosis type, automatically generating diagnosis data of each diagnosis type; and generating a diagnosis impression according to the diagnosis data by an image structured report. The application further discloses a fetal head MR image intelligent diagnosis system. The application can comprehensively, rapidly and intelligently provide image diagnosis data, and immediately know important dangerous signs.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical information, more particularly, to a fetal head MR image intelligent diagnosis method and system. BACKGROUND

[0002] The common fetal central nervous system malformations are neural tube defects, ventriculomegaly, microcephaly, agenesis of the corpus callosum, posterior fossa malformations and neuronal migration abnormalities. Early pregnancy usually relies on ultrasound to detect fetal brain malformations. However, ultrasound is not clear for fetal brain parenchyma display, and it is not easy to diagnose some non-specific manifestations of central nervous system malformations, especially in late pregnancy due to fetal skull ossification, fetal head into the pelvis, ultrasound waves are difficult to penetrate and cannot be satisfactorily displayed. At this time, fetal MRI examination can be used to make up for the deficiency of ultrasound. The clinical indications for fetal MRI examination are: after 18 weeks of gestation, when ultrasound finds suspicious lesions that cannot be determined, fetal MRI can be used for further examination. However, the use of fetal MRI in actual work has certain limitations. First, the brain structure of different age fetuses changes greatly, requiring the diagnostician to have sufficient experience. Second, when assessing brain development, multiple data need to be measured, which is time-consuming and has poor repeatability. The existing diagnosis method cannot quickly and accurately complete the diagnosis of the image task, affecting the work efficiency and quality of the doctor, and also causing the doctor to repeat the work, cannot immediately detect important dangerous signs, and delay the best time for treatment. SUMMARY

[0003] Therefore, the main purpose of the present application is to provide a fetal head MR image intelligent diagnosis method and system, which can solve the problem that the existing technology cannot give comprehensive, rapid and intelligent image diagnosis data, resulting in repeated work of image doctors, inability to immediately know important dangerous signs and delay of the best time for treatment.

[0004] To achieve the above-mentioned purpose, the technical scheme of the present application is as follows:

[0005] In one aspect, the present application provides a fetal head MR image intelligent diagnosis method, comprising: extracting a pre-set image sequence from a patient DICOM image corresponding to an examination item; judging whether the image sequence is a T2WI image quality that meets a first pre-set condition, and if so, defining the T2WI image as a first image; performing regional segmentation on the first image according to a first segmentation rule, and correcting the first image, outputting the corrected first image and the coordinate data of each region; defining the corrected first image as a second image; performing regional segmentation on the second image based on a second segmentation rule, and outputting the coordinate data of each region; based on one or more pre-set diagnosis types and the coordinate data of the region corresponding to the diagnosis type, analyzing the first image or the second image, automatically generating diagnosis data of each diagnosis type, and sending the diagnosis data to an image structured report according to a second pre-set condition; and the image structured report generates a diagnosis impression according to the diagnosis data.

[0006] Preferably, the diagnosis data is one or more of qualitative judgment data, quantitative judgment data, lesion region image, lesion region coordinate, key image, and prompt information.

[0007] Preferably, before correcting the first image, the method further comprises: judging the fetal state, and if the fetal state is after reduction, terminating the diagnosis process of the patient.

[0008] Preferably, before correcting the first image, the method further comprises: segmenting the cerebral falx region on the first image and outputting the coordinate data of the cerebral falx region.

[0009] Preferably, before segmenting the second image, the method further comprises: judging the scanning direction of the T2WI image sequence.

[0010] Preferably, the diagnosis type includes: judging fetal position, segmenting umbilical cord, segmenting placenta, evaluating placenta state, measuring cranium, measuring ventricle, measuring cerebellar axis cistern, measuring cerebellar structure, evaluating corpus callosum development, judging cerebral development malformation, and evaluating gestational age.

[0011] In another aspect, the present application also provides a fetal head MR image intelligent diagnosis system, which comprises a sequence extraction module, an image quality judgment module, a first segmentation module, a second segmentation module, an auxiliary diagnosis module and a structured report module, wherein the sequence extraction module is connected with the image quality judgment module and is used to extract a preset image sequence from a patient DICOM image corresponding to an examination item; the image quality judgment module is connected with the sequence extraction module, the first segmentation module and the auxiliary diagnosis module respectively and is used to judge whether the image sequence is a T2WI image whose quality meets a first preset condition, and if yes, defines the T2WI image as a first image; the first segmentation module is connected with the image quality judgment module, the auxiliary diagnosis module and the second segmentation module respectively and is used to perform regional segmentation on the first image according to a first segmentation rule, correct the first image, output the corrected first image and coordinate data of each region, and define the corrected first image as a second image; the second segmentation module is connected with the first segmentation module and the auxiliary diagnosis module respectively and is used to perform regional segmentation on the second image based on a second segmentation rule and output coordinate data of each region; the auxiliary diagnosis module is connected with the image quality judgment module, the first segmentation module, the second segmentation module and the structured report module respectively and is used to analyze the first image or the second image based on one or more preset diagnosis types and coordinate data of a region corresponding to the diagnosis type, automatically generate diagnosis data of each diagnosis type, and send the diagnosis data to an image structured report according to a second preset condition; and the structured report module is connected with the auxiliary diagnosis module and is used to generate a diagnosis impression according to the diagnosis data.

[0012] Preferably, the diagnosis data is one or more of qualitative judgment data, quantitative judgment data, a lesion region image, lesion region coordinates, a key image and prompt information.

[0013] Preferably, the first segmentation module further comprises a first judgment unit, which is used to judge a fetal state before correcting the first image, and if the fetal state is after a fetal reduction operation, terminate the diagnosis process of the patient.

[0014] Preferably, the first segmentation module further comprises a partition unit, which is used to partition a cerebral falx region on the first image and output coordinate data of the cerebral falx region before correcting the first image.

[0015] Preferably, the first segmentation module further comprises a second judgment unit, which is used to judge a scanning direction of the image sequence as a T2WI image before segmenting the second image.

[0016] Preferably, the diagnosis types include judging fetal position, segmenting umbilical cord, segmenting placenta, evaluating placenta status, measuring cranium, measuring ventricle, measuring cerebellar axis cistern, measuring cerebellar structure, evaluating corpus callosum development, judging cerebral development malformation, and evaluating gestational age.

[0017] Technical effects of the present application:

[0018] The method of the present application applies AI model, radiomics model and rule-based program to the diagnosis of fetal head MRI, obtains an intelligent fetal head MRI diagnosis system, and connects the system to PACS / RIS, so as to automatically generate diagnosis prompt information after image acquisition is completed, and automatically transfer diagnosis data into a structured report, to make comprehensive qualitative and quantitative diagnosis. The method also has a warning and triage function for critical findings. When important dangerous signs are automatically detected, warning information is sent through the information system to prompt doctors to pay attention in time and implement rescue measures. BRIEF DESCRIPTION OF DRAWINGS

[0019] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and serve to explain the principles of the application. In the drawings:

[0020] Figure 1 A flow chart of a fetal head MR image intelligent diagnosis method according to an embodiment of the present application is shown;

[0021] Figure 2 A structural schematic diagram of a fetal head MR image intelligent diagnosis system according to an embodiment of the present application is shown;

[0022] Figure 3 A structural schematic diagram of a fetal head MR image intelligent diagnosis system according to an embodiment of the present application is shown;

[0023] Figure 4 A structural schematic diagram of a fetal head MR image intelligent diagnosis system according to an embodiment of the present application is shown;

[0024] Figure 5 A structural schematic diagram of a fetal head MR image intelligent diagnosis system according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0025] The present application will be described in detail below with reference to the accompanying drawings and embodiments.

[0026] Embodiment one

[0027] Figure 1 A flow chart of a fetal head MR image intelligent diagnosis method according to an embodiment of the present application is shown; as shown in Figure 1 the method includes the following steps:

[0028] Step S101, extracting a pre-set image sequence from a patient DICOM image consistent with the examination item;

[0029] Wherein, the patient DICOM image consistent with the examination item is to determine whether it is consistent with the examination item registered in the RIS. The tool used in this process is program and AI model recognition. Input patient DICOM, make a qualitative judgment, if it is consistent with the examination item (fetal head MR), extract the pre-set image sequence from the patient DICOM image; if it is not consistent with the examination item, the AI diagnosis process is aborted, and a prompt message is sent, which is handled by the relevant personnel and recorded in the database.

[0030] Wherein, the pre-set image sequence can be: T2WI, T1WI, DWI, ADC, axial positioning image, sagittal positioning image, coronal positioning image, etc.

[0031] Wherein, the tool applied in this step is AI model recognition; the output sequence properties, T2WI image, T1WI image, DWI image, ADC image, axial positioning image, sagittal positioning image, coronal positioning image. The above different images are used for the input of different AI function modules in the subsequent.

[0032] Step S102, determining whether the image quality of the image sequence as T2WI meets the first pre-set condition, if yes, defining the T2WI image as the first image;

[0033] Determine whether the T2WI image quality is qualified, wherein the first pre-set condition can be: scanning position is not correct, range is insufficient, obvious image artifact, signal-to-noise ratio is too low, etc.

[0034] If the image quality is qualified, the image can be used for the subsequent AI diagnosis process, and the qualitative judgment result is returned to the corresponding control of the structured report "technical evaluation (image quality)".

[0035] If the T2WI image quality is not qualified, the AI diagnosis process is aborted, and the qualitative judgment result is returned to the corresponding control of the structured report "technical evaluation (image quality)". Send a prompt message, which is handled by the relevant personnel and recorded in the database.

[0036] Step S103, according to the first segmentation rule, regionally segmenting the first image, and correcting the first image, outputting the corrected first image and the coordinate data of each region; defining the corrected first image as the second image;

[0037] Wherein, the first segmentation rule is to segment the skull and cranial cavity in the first image, and output the skull region coordinate data, cranial cavity region coordinate data and fetal number.

[0038] The skull region and the cranial cavity region are used to correct the brain image.

[0039] The skull region and the cranial cavity region are used to correct the brain image.

[0040] The number of fetuses is calculated according to the skull region and the cranial cavity region, and the quantitative result is returned to the corresponding control of the structured report "number of fetuses".

[0041] If the number of fetuses is 1, the diagnostic data of the subsequent AI function module is returned to the corresponding control of the structured report.

[0042] If the number of fetuses is greater than 1, the diagnostic data of the subsequent AI function module is returned to the corresponding control of the structured report for each fetus, such as "fetus 1: …", "fetus 2: …", and so on.

[0043] Before correcting the first image, the method further comprises judging the fetal status, and if the fetal status is post-reduction, terminating the diagnosis process of the patient.

[0044] The tool used in this step is AI model recognition, which makes a judgment on whether each fetus is normal according to a single fetus, and is divided into non-reduction post-reduction.

[0045] The process is as follows: input the first image, skull region coordinate data, and cranial cavity region coordinate data, output qualitative judgment, reduction, key image-reduction. If the qualitative judgment is "post-reduction", the corresponding control of the structured report "overall assessment: post-reduction" is returned, and the key image is generated according to the preset rule, the corresponding control of the structured report "key image" is returned, and the subsequent AI function module is not executed. If the qualitative judgment is "non-reduction post-reduction", no information is returned. Continue to execute the subsequent AI function module.

[0046] Before correcting the first image, the method further comprises segmenting the cerebral falx region on the first image and outputting the coordinate data of the cerebral falx region.

[0047] The tool used in this step is AI model recognition, which segments the cerebral falx region according to the fetus for non-reduction post-reduction. Input the first image, skull region coordinate data, and cranial cavity region coordinate data, output the coordinate data of the cerebral falx region, which is used to correct the brain image.

[0048] The process of correcting the head image is as follows:

[0049] For the latter, after non-reduction, according to a single fetus, the head image of each fetus is aligned so that the cerebral falx is in the middle sagittal position of the image. The first image, the coordinate data of the cerebral falx region, the skull region coordinate data, and the skull cavity region coordinate data are input, and the corrected head T2WI image, the corrected skull region, the corrected skull cavity region, and the corrected cerebral falx region are output.

[0050] The corrected head T2WI axial image is used for subsequent image recognition. The corrected skull region, the corrected skull cavity region are used for the mask of the subsequent AI model. The corrected skull region, the corrected skull cavity region, and the corrected cerebral falx region are used for subsequent measurement.

[0051] Step S104, based on the second segmentation rule, the second image is regionally segmented, and the coordinate data of each region is output;

[0052] Wherein, before segmenting the second image, the method further comprises: judging the image sequence as T2WI image scanning direction. The tool used in this step is AI model recognition. For the latter, after non-reduction, according to a single fetus, the T2WI image of each fetus is further judged for the scanning direction of the fetal head, which is divided into: axial, coronal, sagittal and oblique.

[0053] The second image, the corrected skull region coordinate data, the corrected skull cavity region coordinate data, and the corrected cerebral falx region coordinate data are input; and the corrected head T2WI axial image, the corrected head T2WI coronal image, and the corrected head T2WI sagittal image are output.

[0054] The corrected head T2WI axial image is used for measurement of brain circumference, brain biparietal diameter, head circumference, skull biparietal diameter, skull frontooccipital diameter, lateral ventricle triangular area diameter line, and cerebellar transverse diameter. The corrected head T2WI coronal image is used for measurement of cerebellar hemisphere transverse diameter. The corrected head T2WI sagittal image is used for measurement of cerebellar vermis superior-inferior diameter, anteroposterior diameter, and cross-sectional area.

[0055] Wherein, the second segmentation rule is to segment the skull-brain structure-axial, skull-brain structure-sagittal, skull-brain structure-coronal, umbilical cord, and placenta, which is as follows: the segmentation tool is AI model recognition.

[0056] Segmenting the skull-brain structure-axial:

[0057] For non-reduction cases, per fetus, segment the following structures in the corrected head T2WI axial image (second image) of each fetus: subarachnoid space, cortex, white matter, basal ganglia, right lateral ventricle, left lateral ventricle, third ventricle, fourth ventricle, corpus callosum, septum pellucidum, brainstem, cerebellum. Input the second image, corrected skull region, corrected skull cavity region; output different regions of the brain parenchyma, including: subarachnoid space region, cortex region, white matter region, basal ganglia region, right lateral ventricle region, left lateral ventricle region, third ventricle region, fourth ventricle region, corpus callosum region, septum pellucidum region, brainstem region, cerebellum region.

[0058] The right lateral ventricle region, left lateral ventricle region, third ventricle region, fourth ventricle region, and cerebellum region are used for subsequent brain structure measurement.

[0059] The cortex region, white matter region, basal ganglia region, corpus callosum region, septum pellucidum region, brainstem region, and cerebellum region are used for subsequent brain development assessment.

[0060] Segmenting cranial brain structures - sagittal view:

[0061] For non-reduction cases, per fetus, segment the following structures in the corrected head T2WI sagittal image of each fetus: corpus callosum, cisterna magna, superior cerebellar cistern, and cerebellar vermis. Input the second image, corrected skull region, corrected skull cavity region; output corpus callosum region, cisterna magna region, superior cerebellar cistern region, cerebellar vermis region.

[0062] The corpus callosum region is used for subsequent corpus callosum development assessment.

[0063] The cisterna magna region, superior cerebellar cistern region, and cerebellar vermis region are used for subsequent brain structure measurement.

[0064] Segmenting cranial brain structures - coronal view:

[0065] For non-reduction cases, per fetus, segment the following structures in the corrected head T2WI coronal image of each fetus: cerebellum. Input the corrected head T2WI coronal image, corrected skull region, corrected skull cavity region; output cerebellum region, which is used for subsequent brain structure measurement.

[0066] Segmenting umbilical cord:

[0067] For non-reduction cases, segment the neck umbilical cord per fetus. Input the first image, output neck umbilical cord region, number of umbilical cord around neck, key image - umbilical cord around neck.

[0068] If the number of umbilical cord around neck >= 1, return the corresponding control of "umbilical cord around neck" in the structured report, and generate the key image according to the preset rules, and return the "key image" of the corresponding control of the structured report.

[0069] If the number of umbilical cord around the collar >= 3, send a prompt message, and the relevant personnel are responsible for processing and recording in the database.

[0070] Placenta segmentation:

[0071] For the latter, the placenta is segmented according to the single fetus. The first image is input, and the placenta region is output, which is used for subsequent placenta state evaluation.

[0072] Step S105, based on one or more preset diagnostic types and the coordinates of the region corresponding to the diagnostic type, analyze the first image or the second image, automatically generate diagnostic data for each diagnostic type, and send the diagnostic data to the image structured report according to the second preset condition;

[0073] Among them, the diagnostic data is one or more of qualitative judgment data, quantitative judgment data, lesion region image, lesion region coordinates, key image, and prompt information.

[0074] Among them, the diagnostic type includes: judging fetal position, segmenting umbilical cord, segmenting placenta, evaluating placenta state, measuring cranium, measuring ventricle, measuring cerebellar axis cistern, measuring cerebellar structure, evaluating corpus callosum development, judging cerebral development malformation, and evaluating gestational age.

[0075] The specific implementation is as follows:

[0076] Judging fetal position:

[0077] The tool used in this step is AI model recognition. For the latter, the position is judged according to the single fetus, which is divided into: head position, breech position and transverse position. The axial positioning image, sagittal positioning image and coronal positioning image are input, and the fetal position is output. The qualitative judgment result is returned to the corresponding control of the structured report "fetal position".

[0078] Evaluating placenta state:

[0079] The tool used in this step is AI model recognition and imageomics. For the latter, the placenta state is evaluated according to the single fetus. The first image and the placenta region are input, and the qualitative judgment-placenta state and the key image-placenta abnormality are output.

[0080] If the qualitative judgment is "placenta abnormality", the corresponding control of the structured report "placenta evaluation" is returned, and the key image is generated according to the preset rule, the "key image" of the structured report corresponding control is returned, and a prompt message is sent, which is handled by the relevant personnel and recorded in the database. If the qualitative judgment is "placenta normal", the corresponding control of the structured report "placenta normal" is returned.

[0081] Quantitative measurement of cranium

[0082] This step uses the tool Program. For non-reduction, measure the whole brain size, evaluate the brain development, according to the single fetus. Input corrected skull area, corrected cortex area; output brain circumference, bi-parietal diameter, skull head circumference, skull bi-parietal diameter, skull fronto-occipital diameter, key image-whole brain measurement. Each measurement returns to the corresponding control of the structured report "whole brain measurement", and the key image is generated according to the preset rule, which returns to the corresponding control of the structured report.

[0083] Quantitative measurement of ventricles

[0084] This step uses the tool Program. For non-reduction, measure the volume of right lateral ventricle, left lateral ventricle, third ventricle, fourth ventricle, and bilateral triangular area diameter, evaluate the brain development, according to the single fetus. Input right lateral ventricle area, left lateral ventricle area, third ventricle area, fourth ventricle area; output right lateral ventricle volume, left lateral ventricle volume, third ventricle volume, fourth ventricle volume, right lateral ventricle triangular area diameter, left lateral ventricle triangular area diameter, key image-ventricle measurement. Each measurement returns to the corresponding control of the structured report "ventricle measurement", and the key image is generated according to the preset rule, which returns to the corresponding control of the structured report.

[0085] Quantitative measurement of cerebellar cistern

[0086] This step uses the tool Program. For non-reduction, measure the size of cerebellomedullary cistern and superior cerebellar cistern, evaluate the brain development, according to the single fetus. Input cerebellomedullary cistern area, superior cerebellar cistern area; output cerebellomedullary cistern anteroposterior diameter, superior cerebellar cistern anteroposterior diameter, key image-cerebellar cistern measurement. Each measurement returns to the corresponding control of the structured report "cerebellar cistern measurement", and the key image is generated according to the preset rule, which returns to the corresponding control of the structured report.

[0087] Quantitative measurement of cerebellar structure

[0088] The tool used in this step is program. For non-reduction surgery, based on the corrected T2WI axial, corrected T2WI sagittal and corrected T2WI coronal image results of each fetus, the size of cerebellar structure is measured, and brain development is evaluated. The corrected head T2WI axial image cerebellar region, the corrected head T2WI coronal image cerebellar region and the corrected head T2WI sagittal image cerebellar vermis region are input; the corrected head T2WI axial image cerebellar hemisphere transverse diameter, the corrected head T2WI coronal image cerebellar hemisphere transverse diameter, the corrected head T2WI sagittal image cerebellar vermis upper and lower diameter, the corrected head T2WI sagittal image cerebellar vermis front and back diameter, the corrected head T2WI sagittal image cerebellar vermis area, the key image-cerebellar structure measurement axial, the key image-cerebellar structure measurement coronal and the key image-cerebellar structure measurement sagittal are output. Each measurement value is returned to the corresponding control of the structured report "cerebellar structure measurement", and the key images are generated according to the preset rules, and the key image 1, the key image 2 and the key image 3 of the structured report corresponding control are returned.

[0089] Assess corpus callosum development

[0090] The tool used in this step is AI model recognition. For non-reduction surgery, the development of corpus callosum is judged according to single fetus. The corrected head T2WI sagittal image and the corpus callosum region are input, and the qualitative judgment and the key image-corpus callosum development anomaly are output. If the qualitative judgment is "corpus callosum development anomaly", the corresponding control of the structured report "corpus callosum development anomaly" is returned, the key image of the structured report corresponding control is generated according to the preset rules, and the prompt information is sent, which is processed by the relevant personnel and recorded in the database. If the qualitative judgment is "corpus callosum development normal", the corresponding control of the structured report "corpus callosum development normal" is returned.

[0091] Judge cerebral development malformation

[0092] The tool used in this step is AI model recognition. For non-reduction surgery, the development of brain is evaluated according to single fetus, and whether there is development malformation is judged. The corrected head T2WI axial image, the corrected skull region and the corrected cranial cavity region are input, and the qualitative judgment, the activation region of the classification model and the key image-cerebral development malformation are output. If the qualitative judgment is "cerebral development malformation", the corresponding control of the structured report "cerebral development malformation" is returned, the key image of the structured report corresponding control is generated according to the preset rules, and the prompt information is sent, which is processed by the relevant personnel and recorded in the database. If the qualitative judgment is "cerebral development normal", the corresponding control of the structured report "cerebral development normal" is returned.

[0093] Assess gestational age

[0094] The tool used in this step is an AI model to identify the gestational age of a single fetus after non-reduction. The input is the corrected head T2WI axial image, the corrected skull region, the corrected cranial cavity region, and the gestational age from the RIS registration. The output is the predicted gestational age and a qualitative judgment. If the qualitative judgment is "abnormal predicted gestational age", the corresponding control of "abnormal predicted gestational age" is returned to the structured report, and a prompt message is sent for the relevant personnel to handle and record in the database. If the qualitative judgment is "normal predicted gestational age", the corresponding control of "normal predicted gestational age" is returned to the structured report.

[0095] In step S106, the image structured report generates a diagnostic impression based on the diagnostic data.

[0096] All findings of the tools (AI models, imageomics models, and programs) are integrated to obtain the overall diagnostic impression.

[0097] A qualitative judgment is made based on the results returned by all intelligent tools. The tool used is a program, the input is all diagnostic data, and the output is the "diagnostic impression" of the structured report. Based on the rules built into the structured report, all diagnostic data are integrated to automatically obtain the final diagnosis and return it to the "diagnostic impression". All data and images are stored in the structured report database.

[0098] The embodiments of the present application apply AI models, imageomics models, and rule-based programs to the diagnosis of fetal head MRI to obtain an intelligent fetal head MRI diagnosis system. When connected to PACS / RIS, the system can automatically generate a diagnostic prompt message after image acquisition is completed and automatically transfer diagnostic data into a structured report to make a comprehensive qualitative and quantitative diagnosis. The method also has a critical finding warning and triage function. When important dangerous signs are automatically detected, warning information is sent through the information system to prompt doctors to pay attention and implement rescue measures in a timely manner.

[0099] Embodiment Two

[0100] Figure 2 A fetal head MR image intelligent diagnosis system structure schematic diagram according to Embodiment Two of the present application is shown. As shown in Figure 2 The system includes a sequence extraction module 10, an image quality judgment module 20, a first segmentation module 30, a second segmentation module 40, an auxiliary diagnosis module 50, and a structured report module 60, wherein,

[0101] The sequence extraction module 10 is connected to the image quality judgment module 20 and is used to extract pre-set image sequences from patient DICOM images consistent with the examination item;

[0102] The DICOM image of the patient consistent with the examination item is used to determine whether it is consistent with the examination item registered in the RIS. The tools used in this process are program and AI model identification. The patient DICOM is input, a qualitative judgment is made, if it is consistent with the examination item (fetal head MR), the pre-set image sequence is extracted from the patient DICOM image; if it is not consistent with the examination item, the AI diagnosis process is aborted, a prompt message is sent, the relevant personnel are responsible for processing, and are recorded in the database.

[0103] The pre-set image sequence can be T2WI, T1WI, DWI, ADC, axial positioning image, sagittal positioning image, coronal positioning image, etc.

[0104] The tool applied in this step is AI model identification; the output sequence properties, T2WI image, T1WI image, DWI image, ADC image, axial positioning image, sagittal positioning image, coronal positioning image. The above different images are used for the input of different AI function modules in the subsequent process.

[0105] The image quality judgment module 20 is connected with the sequence extraction module 10, the first segmentation module 30 and the auxiliary diagnosis module 50 respectively, and is used for judging whether the image quality of the T2WI image sequence meets the first preset condition, and if so, defining the T2WI image as a first image;

[0106] The T2WI image quality is identified, and the first preset condition can be that the scanning position is not correct, the range is insufficient, there are obvious image artifacts, the signal-to-noise ratio is too low, etc.

[0107] If the image quality is qualified, the image can be used for the subsequent AI diagnosis process, and the qualitative judgment result is returned to the corresponding control of the structured report "technical evaluation (image quality)".

[0108] If the T2WI image quality is not qualified, the AI diagnosis process is aborted, the qualitative judgment result is returned to the corresponding control of the structured report "technical evaluation (image quality)", a prompt message is sent, the relevant personnel are responsible for processing, and are recorded in the database.

[0109] The first segmentation module 30 is connected with the image quality judgment module 20, the auxiliary diagnosis module 50 and the second segmentation module 40 respectively, and is used for performing regional segmentation on the first image according to the first segmentation rule, correcting the first image, outputting the corrected first image and the coordinate data of each region, and defining the corrected first image as a second image.

[0110] The first segmentation rule is to segment the skull and skull cavity in the first image, and output skull region coordinate data, skull cavity region coordinate data and fetus number.

[0111] The skull cavity region is used for a mask for segmenting the cerebral falx.

[0112] The skull region and the skull cavity region are used for correcting the brain image.

[0113] The fetus number is calculated according to the skull region and the skull cavity region, and the quantitative result is returned to the corresponding control of the structured report “fetus number”.

[0114] If the fetus number is 1, the diagnosis data of the subsequent AI function module is returned to the corresponding control of the structured report.

[0115] If the fetus number is greater than 1, the diagnosis data of the subsequent AI function module is returned to the corresponding control of each fetus in the structured report, such as “fetus 1: …”, “fetus 2: …”, and so on.

[0116] The process of correcting the head image is as follows:

[0117] For non-reduction after the operation, the head image of each fetus is aligned according to the single fetus, so that the cerebral falx is in the mid-sagittal position of the image. The first image, the coordinate data of the cerebral falx region, the coordinate data of the skull region, and the coordinate data of the skull cavity region are input, and the corrected head T2WI image, the corrected skull region, the corrected skull cavity region, and the corrected cerebral falx region are output.

[0118] The corrected head T2WI axial image is used for subsequent image recognition. The corrected skull region, the corrected skull cavity region are used for the mask of the subsequent AI model. The corrected skull region, the corrected skull cavity region, and the corrected cerebral falx region are used for subsequent measurement.

[0119] The second segmentation module 40 is connected with the first segmentation module 30 and the auxiliary diagnosis module 50, and is used for performing regional segmentation on the second image based on a second segmentation rule, and outputting coordinate data of each region.

[0120] The second image, the corrected skull region coordinate data, the corrected skull cavity region coordinate data, and the corrected cerebral falx region coordinate data are input, and the corrected head T2WI axial image, the corrected head T2WI coronal image, and the corrected head T2WI sagittal image are output.

[0121] The corrected head T2WI axial image is used for measurement of brain circumference, brain biparietal diameter, head circumference, skull biparietal diameter, skull frontooccipital diameter, lateral ventricle triangular area diameter and cerebellar transverse diameter, etc. The corrected head T2WI coronal image is used for measurement of cerebellar hemisphere transverse diameter, etc. The corrected head T2WI sagittal image is used for measurement of cerebellar vermis superior-inferior diameter, anterior-posterior diameter and cross-sectional area, etc.

[0122] The second segmentation rule is to segment the cranial structure-axis, cranial structure-sagittal, cranial structure-coronal, umbilical cord, placenta, and the specific rules are as follows: the segmentation tool is an AI model recognition:

[0123] Segmentation of cranial structure-axis:

[0124] For non-reduction after the latter, according to a single fetus, in each fetal corrected head T2WI axial image (second image), the subarachnoid space, cortex, white matter, basal ganglia, right lateral ventricle, left lateral ventricle, third ventricle, fourth ventricle, corpus callosum, septum pellucidum, brainstem and cerebellum are segmented. The second image, the corrected skull region and the corrected cranial cavity region are input; and the different regions of the brain parenchyma are output, including: subarachnoid space region, cortex region, white matter region, basal ganglia region, right lateral ventricle region, left lateral ventricle region, third ventricle region, fourth ventricle region, corpus callosum region, septum pellucidum region, brainstem region and cerebellum region.

[0125] The right lateral ventricle region, the left lateral ventricle region, the third ventricle region, the fourth ventricle region and the cerebellum region are used for subsequent brain structure measurement.

[0126] The cortex region, the white matter region, the basal ganglia region, the corpus callosum region, the septum pellucidum region, the brainstem region and the cerebellum region are used for subsequent brain development evaluation.

[0127] Segmentation of cranial structure-sagittal:

[0128] For non-reduction after the latter, according to a single fetus, in each fetal corrected head T2WI sagittal image, the corpus callosum, the cerebellumoblongata pool, the cerebellar superior pool and the cerebellar vermis are segmented. The second image, the corrected skull region and the corrected cranial cavity region are input; and the corpus callosum region, the cerebellumoblongata pool region, the cerebellar superior pool region and the cerebellar vermis region are output.

[0129] The corpus callosum region is used for subsequent corpus callosum development evaluation.

[0130] The cerebellumoblongata pool region, the cerebellar superior pool region and the cerebellar vermis region are used for subsequent brain structure measurement.

[0131] Segmentation of cranial structure-coronal:

[0132] For non-reduction surgery, according to single fetus, the cerebellum is segmented in the corrected head T2WI coronal image of each fetus. The corrected head T2WI coronal image, the corrected skull region and the corrected cranial cavity region are input; and the cerebellum region is output, which is used for subsequent brain structure measurement.

[0133] Segmenting umbilical cord:

[0134] For non-reduction surgery, the neck umbilical cord is segmented according to single fetus. The first image is input; and the neck umbilical cord region, the number of umbilical cord around the neck and the key image-umbilical cord around the neck are output.

[0135] If the number of umbilical cord around the neck is greater than or equal to 1, the corresponding control of the structured report "umbilical cord around the neck" is returned, and the key image is generated according to the preset rule, and the "key image" of the corresponding control of the structured report is returned.

[0136] If the number of umbilical cord around the neck is greater than or equal to 3, a prompt message is sent, which is handled by the relevant personnel and recorded in the database.

[0137] Segmenting placenta:

[0138] For non-reduction surgery, the placenta is segmented according to single fetus. The first image is input; and the placenta region is output, which is used for subsequent placenta state evaluation.

[0139] The auxiliary diagnosis module 50 is connected with the image quality judgment module 20, the first segmentation module 30, the second segmentation module 40 and the structured report module 60, respectively, for analyzing the first image or the second image based on one or more preset diagnosis types and coordinate data of the region corresponding to the diagnosis type, automatically generating diagnosis data of each diagnosis type, and sending the diagnosis data to the image structured report according to a second preset condition.

[0140] The diagnosis data is one or more of qualitative judgment data, quantitative judgment data, lesion region image, lesion region coordinate, key image and prompt message.

[0141] The diagnosis types include: judging fetal position, segmenting umbilical cord, segmenting placenta, evaluating placenta state, measuring cranium, measuring brain ventricle, measuring cerebellar axial cistern, measuring cerebellar structure, evaluating corpus callosum development, judging cerebral development malformation and evaluating gestational age.

[0142] The specific implementation is as follows:

[0143] Judging fetal position:

[0144] The tool used in this step is AI model recognition. For non-reduction surgery, the fetal position is determined according to the individual fetus, which is divided into: head position, breech position and transverse position. Input the axial positioning image, sagittal positioning image and coronal positioning image, and output the fetal position. The qualitative judgment result is returned to the corresponding control of the structured report "fetal position".

[0145] Evaluate placental status:

[0146] The tool used in this step is AI model recognition and imageomics. For non-reduction surgery, the placental status is evaluated according to the individual fetus. Input the first image and the placental region, and output the qualitative judgment-placental status and the key image-placental abnormalities.

[0147] If the qualitative judgment is "placental abnormalities", the corresponding control of the structured report "placental evaluation" is returned, and the key image is generated according to the preset rules, the "key image" of the structured report is returned, and a prompt message is sent, which is handled by the relevant personnel and recorded in the database. If the qualitative judgment is "placental normal", the corresponding control of the structured report "placental normal" is returned.

[0148] Quantitative measurement of cranium and brain

[0149] The tool used in this step is a program. For non-reduction surgery, the overall size of the cranium and brain is measured according to the individual fetus to evaluate brain development. Input the corrected skull region and the corrected cortical region; Output the head circumference, biparietal diameter, skull circumference, skull biparietal diameter, skull frontooccipital diameter, and key image-cranial and brain overall measurement. Each measurement value is returned to the corresponding control of the structured report "cranium and brain overall measurement", and the key image is generated according to the preset rules and returned to the corresponding control of the structured report.

[0150] Quantitative measurement of ventricles

[0151] The tool used in this step is a program. For non-reduction surgery, the volume of the right ventricle, left ventricle, third ventricle and fourth ventricle and the bilateral triangular area diameter line are measured according to the individual fetus to evaluate brain development. Input the right ventricle region, left ventricle region, third ventricle region and fourth ventricle region, and output the right ventricle volume, left ventricle volume, third ventricle volume, fourth ventricle volume, right ventricle triangular area diameter line, left ventricle triangular area diameter line and key image-ventricle measurement. Each measurement value is returned to the corresponding control of the structured report "ventricle measurement", and the key image is generated according to the preset rules and returned to the corresponding control of the structured report.

[0152] Quantitative measurement of cerebellar cistern

[0153] The tool used in this step is Program. For non-reduction surgery, the size of the cerebellomedullary cistern and the superior cerebellar cistern is measured according to the individual fetus to evaluate brain development. The input is the area of the cerebellomedullary cistern and the area of the superior cerebellar cistern; the output is the superior-inferior diameter of the cerebellomedullary cistern, the superior-inferior diameter of the superior cerebellar cistern, and the key image-cerebral cistern measurement. Each measurement is returned to the corresponding control of the structured report "cerebral cistern measurement" and the key image is generated according to the preset rules and returned to the corresponding control of the structured report.

[0154] Quantitative measurement of cerebellar structures

[0155] The tool used in this step is Program. For non-reduction surgery, the size of the cerebellar structure is measured according to the individual fetus based on the corrected T2WI axial, corrected T2WI sagittal, and corrected T2WI coronal image results to evaluate brain development. The input is the corrected T2WI axial image cerebellar area, the corrected T2WI coronal image cerebellar area, and the corrected T2WI sagittal image cerebellar vermis area; the output is the corrected T2WI axial image cerebellar hemisphere transverse diameter, the corrected T2WI coronal image cerebellar hemisphere transverse diameter, the corrected T2WI sagittal image cerebellar vermis superior-inferior diameter, the corrected T2WI sagittal image cerebellar vermis anterior-posterior diameter, the corrected T2WI sagittal image cerebellar vermis area, the key image-cerebellar structure measurement axial, the key image-cerebellar structure measurement coronal, and the key image-cerebellar structure measurement sagittal. Each measurement is returned to the corresponding control of the structured report "cerebellar structure measurement" and the key image is generated according to the preset rules and returned to the corresponding control of the structured report key image 1, key image 2, and key image 3.

[0156] Evaluate corpus callosum development

[0157] The tool used in this step is AI model recognition. For non-reduction surgery, the corpus callosum development is judged according to the individual fetus. The input is the corrected head T2WI sagittal image and the corpus callosum area, and the output is the qualitative judgment and the key image-corpus callosum development abnormality. If the qualitative judgment is "corpus callosum development abnormality", the corresponding control of the structured report "corpus callosum development abnormality" is returned, and the key image is generated according to the preset rules and returned to the corresponding control of the structured report "key image", and a prompt message is sent for the relevant personnel to handle and record in the database. If the qualitative judgment is "corpus callosum development normal", the corresponding control of the structured report "corpus callosum development normal" is returned.

[0158] Judge cerebral development malformation

[0159] The tool used in this step is AI model recognition. For non-reduction fetuses, the brain development is evaluated according to a single fetus to determine whether there is a developmental deformity. The corrected head T2WI axial image, the corrected skull region, and the corrected cranial cavity region are input, and the qualitative judgment, the activation region of the classification model, and the key image of the brain development deformity are output. If the qualitative judgment is "brain development deformity", the corresponding control of the structured report "brain development deformity" is returned, and the key image is generated according to the preset rule, the "key image" of the structured report corresponding control is returned, and a prompt information is sent, which is processed by the relevant personnel and recorded in the database. If the qualitative judgment is "normal brain development", the corresponding control of the structured report "normal brain development" is returned.

[0160] Assess gestational age

[0161] The tool used in this step is AI model recognition. For non-reduction fetuses, the gestational age is determined according to a single fetus. The corrected head T2WI axial image, the corrected skull region, the corrected cranial cavity region, and the gestational age from RIS registration are input, and the predicted gestational age and the qualitative judgment are output. If the qualitative judgment is "predicted gestational age abnormality", the corresponding control of the structured report "predicted gestational age abnormality" is returned, and a prompt information is sent, which is processed by the relevant personnel and recorded in the database. If the qualitative judgment is "predicted gestational age normal", the corresponding control of the structured report "predicted gestational age normal" is returned.

[0162] The structured report module 60 is connected with the auxiliary diagnosis module 50, and is used for generating a diagnosis impression according to the diagnosis data.

[0163] All findings of the tools (AI model, imageomics model, and program) are integrated to obtain an overall diagnosis impression.

[0164] A qualitative judgment is made according to the returned results of all intelligent tools. The tool used is a program, all diagnosis data are input, and the "diagnosis impression" of the structured report is output. Based on the rules built in the structured report, all diagnosis data are integrated, and the final diagnosis is automatically obtained and returned to the "diagnosis impression". All data and images are stored in the structured report database.

[0165] In the embodiments of the present application, the AI model, the imageomics model, and the rule-based program are applied to the diagnosis of fetal head MRI to obtain an intelligent fetal head MRI diagnosis system. When connected to PACS / RIS, the diagnosis prompt information can be automatically generated after the image acquisition is completed, the diagnosis data can be automatically transmitted into the structured report, and a comprehensive qualitative and quantitative diagnosis can be made. The method also has a warning and triage function for critical findings. When important dangerous signs are automatically detected, a warning information is sent through the information system to prompt the doctor to pay attention in time and implement rescue measures.

[0166] Embodiment three

[0167] Figure 3 Fig. 2 shows a structure diagram of the fetal head MR image intelligent diagnosis system according to Embodiment Two of the present application; as shown in the figure, the first segmentation module 30 further includes a first judgment unit 302, which is configured to judge the fetal state before correcting the first image, and if the fetal state is post-reduction, the diagnosis process of the patient is terminated. Figure 3

[0168] The tool used in this step is AI model recognition, which judges whether each fetal state is normal according to a single fetus, and is divided into post-reduction and non-post-reduction.

[0169] The process is as follows: input the first image, skull region coordinate data, and cranial cavity region coordinate data, output qualitative judgment, reduction, and key image-reduction. If the qualitative judgment is "post-reduction", the corresponding control of the structured report "overall assessment: post-reduction" is returned, and the key image is generated according to the preset rule, the "key image" of the structured report corresponding control is returned, and the subsequent AI function module is not executed. If the qualitative judgment is "non-reduction", no information is returned. Continue to execute the subsequent AI function module.

[0170] Embodiment four

[0171] Figure 4 Fig. 2 shows a structure diagram of the fetal head MR image intelligent diagnosis system according to Embodiment Two of the present application; as shown in the figure, the first segmentation module 30 further includes a first judgment unit 302, which is configured to judge the fetal state before correcting the first image, and if the fetal state is post-reduction, the diagnosis process of the patient is terminated. Figure 4

[0172] The tool used in this step is AI model recognition, which judges whether each fetal state is normal according to a single fetus, and is divided into post-reduction and non-post-reduction.

[0173] Embodiment five

[0174] Figure 5 Fig. 2 shows a structure diagram of the fetal head MR image intelligent diagnosis system according to Embodiment Two of the present application; as shown in the figure, the first segmentation module 30 further includes a first judgment unit 302, which is configured to judge the fetal state before correcting the first image, and if the fetal state is post-reduction, the diagnosis process of the patient is terminated. Figure 5

[0175] ​​​The tool used in this step is AI model recognition. For non-reduction post-operation, according to single fetus, the scanning direction of the head of each fetus is further judged on the T2WI image of each fetus, and is divided into: axial position, coronal position, sagittal position and oblique position.

[0176] The second image, the corrected skull region coordinate data, the corrected cranial cavity region coordinate data and the corrected cerebral falx region coordinate data are input; and the corrected head T2WI axial position image, the corrected head T2WI coronal position image and the corrected head T2WI sagittal position image are output.

[0177] From the above description, it can be seen that the above-mentioned embodiments of the present application achieve the following technical effects: The embodiments of the present application apply AI model, imageomics model and rule-based program to the diagnosis of fetal head MRI, obtain fetal head MRI intelligent diagnosis system, connect it to PACS / RIS, can automatically generate diagnosis prompt information after image acquisition is completed, and automatically transmit diagnosis data into structured report, make comprehensive qualitative and quantitative diagnosis, the method also has the functions of early warning and triage of critical discovery, when important dangerous signs are automatically detected, warning information is sent through the information system to prompt the doctor to pay attention in time and implement rescue measures.

[0178] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present application can be realized by general computing devices, which can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be realized by program codes executable by computing devices, so that they can be stored in storage devices and executed by computing devices, or they can be respectively manufactured into individual integrated circuit modules, or multiple modules or steps among them can be manufactured into a single integrated circuit module to realize. Thus, the present application is not limited to any specific combination of hardware and software.

[0179] The above only describes the preferred embodiments of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A fetal head MR image intelligent diagnosis method, characterized in that, The method comprises the following steps: extracting a preset image sequence from a DICOM image of a patient consistent with the examination item; determining whether the image sequence meets a first preset condition for image quality of T2WI, and if so, defining the image of T2WI as a first image; performing regional segmentation on the first image according to a first segmentation rule, correcting the first image, outputting the corrected first image and coordinate data of each region, and defining the corrected first image as a second image; wherein the first segmentation rule is to segment the skull and cranial cavity in the first image, and output the skull region coordinate data, the cranial cavity region coordinate data, and the number of fetuses; wherein the step of correcting the first image is to: for non-reduction fetotomy or for each fetus, align the head image of each fetus so that the cerebral falx is in the median sagittal position of the image; input the first image, the coordinate data of the cerebral falx region, the coordinate data of the skull region, and the coordinate data of the cranial cavity region, and output the corrected head T2WI image, the corrected skull region, the corrected cranial cavity region, and the corrected cerebral falx region; performing regional segmentation on the second image based on a second segmentation rule, and outputting coordinate data of each region; wherein the second segmentation rule is to segment the cranial structure-axis, the cranial structure-sagittal, the cranial structure-coronal, the umbilical cord, and the placenta; segmenting the cranial structure-axis to output: the subarachnoid space region, the cortical region, the white matter region, the basal ganglia region, the right lateral ventricle region, the left lateral ventricle region, the third ventricle region, the fourth ventricle region, the corpus callosum region, the septum pellucidum region, the brainstem region, and the cerebellum region; segmenting the cranial structure-sagittal to output: the corpus callosum region, the cerebellum cistern region, the superior cerebellar cistern region, and the cerebellar vermis region; segmenting the cranial structure-coronal to output: the cerebellum region, which is used for subsequent brain structure measurement; segmenting the umbilical cord to output: the neck umbilical cord region and the number of umbilical cord around the neck, and the key image-umbilical cord around the neck; segmenting the placenta to output: the placenta region, which is used for subsequent placenta state evaluation; based on one or more preset diagnostic types and the coordinate data of the region corresponding to the diagnostic type, analyzing the first image or the second image, automatically generating diagnostic data of each diagnostic type, and sending the diagnostic data to an image structured report according to a second preset condition; wherein the diagnostic types include: judging fetal position, segmenting umbilical cord, segmenting placenta, evaluating placenta state, measuring cranial structure, measuring ventricle, measuring cerebellar axis cistern, measuring cerebellar structure, evaluating corpus callosum development, judging cerebral development malformation, and evaluating gestational age; the image structured report generates a diagnostic impression according to the diagnostic data.

2. The fetal head MR image intelligent diagnosis method according to claim 1, characterized in that, The diagnostic data is one or more of qualitative judgment data, quantitative judgment data, lesion region image, lesion region coordinate, key image, and prompt information.

3. The fetal head MR image intelligent diagnosis method according to claim 1, characterized in that, Before correcting the first image, the method further comprises: judging the fetal state, and if the fetal state is after reduction fetotomy, terminating the diagnosis process of the patient.

4. The intelligent fetal head MR image diagnostic method of claim 1, wherein, Before the first image is corrected, the method further comprises: segmenting a cerebral falx region on the first image, and outputting coordinate data of the cerebral falx region.

5. The intelligent fetal head MR image diagnostic method of claim 1, wherein, Before the second image is segmented, the method further comprises: judging an image scanning direction of the T2WI image sequence.

6. A fetal head MR image intelligent diagnosis system, characterized in that, The system comprises: a sequence extraction module, an image quality judgment module, a first segmentation module, a second segmentation module, an auxiliary diagnosis module, and a structured report module, wherein, The sequence extraction module is connected to the image quality judgment module, and is configured to extract a preset image sequence from a patient DICOM image corresponding to an examination item; The image quality judgment module is connected to the sequence extraction module, the first segmentation module, and the auxiliary diagnosis module, and is configured to judge whether the image quality of the T2WI image sequence meets a first preset condition, and if so, define the T2WI image as a first image; The first segmentation module is connected to the image quality judgment module, the auxiliary diagnosis module, and the second segmentation module, and is configured to perform regional segmentation on the first image according to a first segmentation rule, correct the first image, output the corrected first image and coordinate data of each region, and define the corrected first image as a second image; wherein the first segmentation rule is to segment a skull and a skull cavity in the first image, and output skull region coordinate data, skull cavity region coordinate data, and a number of fetuses; wherein the step of correcting the first image is to: for non-reduction fetuses, or for a single fetus, align the head image of each fetus, so that the cerebral falx is in the median sagittal position of the image; input the first image, the coordinate data of the cerebral falx region, the coordinate data of the skull region, and the coordinate data of the skull cavity region, and output a corrected head T2WI image, a corrected skull region, a corrected skull cavity region, and a corrected cerebral falx region; The second segmentation module is connected to the first segmentation module and the auxiliary diagnosis module, and is configured to perform regional segmentation on the second image based on a second segmentation rule, and output coordinate data of each region; wherein the second segmentation rule is to segment a skull-brain structure-axial position, a skull-brain structure-sagittal position, a skull-brain structure-coronal position, an umbilical cord, and a placenta; Segmenting the skull-brain structure-axial position outputs: a subarachnoid space region, a cortical region, a white matter region, a basal ganglia region, a right lateral ventricle region, a left lateral ventricle region, a third ventricle region, a fourth ventricle region, a corpus callosum region, a septum pellucidum region, a brainstem region, and a cerebellum region; Segmenting the skull-brain structure-sagittal position outputs: a corpus callosum region, a cerebellum medulla oblongata pool region, a cerebellum upper pool region, and a cerebellum vermis region; Segmenting the skull-brain structure-coronal position outputs: a cerebellum region, which is used for subsequent brain structure measurement; Segmenting the umbilical cord outputs: a neck umbilical cord region, a number of umbilical cord neck rings, and a key image-umbilical cord neck; Segmenting the placenta outputs: a placenta region, which is used for subsequent placenta state evaluation; The auxiliary diagnosis module is connected with the image quality judgment module, the first segmentation module, the second segmentation module and the structured report module respectively, and is configured to analyze the first image or the second image based on one or more preset diagnosis types and coordinate data of regions corresponding to the diagnosis types, automatically generate diagnosis data of each diagnosis type, and send the diagnosis data to an image structured report according to a second preset condition; wherein the diagnosis types include judging fetal position, segmenting umbilical cord, segmenting placenta, evaluating placenta state, measuring cranium, measuring ventricle, measuring cerebellar axis cistern, measuring cerebellar structure, evaluating corpus callosum development, judging cerebral development malformation and evaluating gestational age. The structured report module is connected with the auxiliary diagnosis module, and is configured to generate a diagnosis impression according to the diagnosis data.

7. The fetal head MR image intelligent diagnosis system according to claim 6, characterized in that, The diagnosis data is one or more of qualitative judgment data, quantitative judgment data, lesion region image, lesion region coordinate, key image and prompt information.

8. The fetal head MR image intelligent diagnosis system according to claim 6, characterized in that, The first segmentation module further includes a first judgment unit configured to judge fetal state before correcting the first image, and terminate the diagnosis process of the patient if the fetal state is after reduction of fetus.

9. The fetal head MR image intelligent diagnosis system according to claim 6, characterized in that, The first segmentation module further includes a partition unit configured to segment a cerebral falx region on the first image and output coordinate data of the cerebral falx region before correcting the first image.

10. The fetal head MR image intelligent diagnosis system according to claim 6, characterized in that, The first segmentation module further includes a second judgment unit configured to judge an image scanning direction of the image sequence as T2WI before segmenting the second image.

Images