CN111739645B - Training method of immune-related pneumonia prediction model - Google Patents
Training method of immune-related pneumonia prediction model Download PDFInfo
- Publication number
- CN111739645B CN111739645B CN202010407606.2A CN202010407606A CN111739645B CN 111739645 B CN111739645 B CN 111739645B CN 202010407606 A CN202010407606 A CN 202010407606A CN 111739645 B CN111739645 B CN 111739645B
- Authority
- CN
- China
- Prior art keywords
- parameters
- value
- image
- frequency
- values
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
- 206010035664 Pneumonia Diseases 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000012549 training Methods 0.000 title claims abstract description 31
- 210000004072 lung Anatomy 0.000 claims description 56
- 238000009826 distribution Methods 0.000 claims description 29
- 230000008030 elimination Effects 0.000 claims description 8
- 238000003379 elimination reaction Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 8
- 238000009169 immunotherapy Methods 0.000 claims description 7
- 238000001959 radiotherapy Methods 0.000 claims description 7
- 230000007170 pathology Effects 0.000 claims description 4
- 230000000391 smoking effect Effects 0.000 claims description 4
- 238000011282 treatment Methods 0.000 abstract description 9
- 238000003745 diagnosis Methods 0.000 abstract description 3
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 12
- 238000004590 computer program Methods 0.000 description 9
- 230000008901 benefit Effects 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 238000003860 storage Methods 0.000 description 5
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 206010067484 Adverse reaction Diseases 0.000 description 3
- 206010058467 Lung neoplasm malignant Diseases 0.000 description 3
- 230000006838 adverse reaction Effects 0.000 description 3
- 238000002512 chemotherapy Methods 0.000 description 3
- 201000005202 lung cancer Diseases 0.000 description 3
- 208000020816 lung neoplasm Diseases 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 230000004083 survival effect Effects 0.000 description 3
- 201000009030 Carcinoma Diseases 0.000 description 2
- 229940076838 Immune checkpoint inhibitor Drugs 0.000 description 2
- 102000037984 Inhibitory immune checkpoint proteins Human genes 0.000 description 2
- 108091008026 Inhibitory immune checkpoint proteins Proteins 0.000 description 2
- 230000004075 alteration Effects 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 239000012274 immune-checkpoint protein inhibitor Substances 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 208000002154 non-small cell lung carcinoma Diseases 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000002560 therapeutic procedure Methods 0.000 description 2
- 208000029729 tumor suppressor gene on chromosome 11 Diseases 0.000 description 2
- 208000010576 undifferentiated carcinoma Diseases 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 102000008096 B7-H1 Antigen Human genes 0.000 description 1
- 108010074708 B7-H1 Antigen Proteins 0.000 description 1
- 208000009956 adenocarcinoma Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 201000011510 cancer Diseases 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000007596 consolidation process Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000002651 drug therapy Methods 0.000 description 1
- 102000052116 epidermal growth factor receptor activity proteins Human genes 0.000 description 1
- 108700015053 epidermal growth factor receptor activity proteins Proteins 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000001900 immune effect Effects 0.000 description 1
- 238000002649 immunization Methods 0.000 description 1
- 230000003053 immunization Effects 0.000 description 1
- 239000002955 immunomodulating agent Substances 0.000 description 1
- 230000001024 immunotherapeutic effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- YOHYSYJDKVYCJI-UHFFFAOYSA-N n-[3-[[6-[3-(trifluoromethyl)anilino]pyrimidin-4-yl]amino]phenyl]cyclopropanecarboxamide Chemical compound FC(F)(F)C1=CC=CC(NC=2N=CN=C(NC=3C=C(NC(=O)C4CC4)C=CC=3)C=2)=C1 YOHYSYJDKVYCJI-UHFFFAOYSA-N 0.000 description 1
- 229960003301 nivolumab Drugs 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 206010041823 squamous cell carcinoma Diseases 0.000 description 1
- 230000008685 targeting Effects 0.000 description 1
- 230000001225 therapeutic effect Effects 0.000 description 1
- 238000002759 z-score normalization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
Landscapes
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Engineering & Computer Science (AREA)
- Public Health (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention mainly discloses a training method of an immune-related pneumonia prediction model, which comprises the following steps: acquiring N of target object 1 A plurality of clinical parameters; acquiring N based on CT image of target object 2 Image parameters; at said N 1 Clinical parameters and N 2 N parameters are determined in the image parameters; an immune-related pneumonia prediction model is trained with the N parameters of a plurality of target subjects. The scheme provided by the invention is beneficial to training an immune-related pneumonia prediction model with better performance at a higher speed, so that whether a subject suffers from immune-related pneumonia or not can be accurately predicted, and diagnosis of doctors and later treatment of the subject are facilitated to a certain extent.
Description
Technical Field
The invention relates to the technical field of medical treatment, in particular to a training method of an immune-related pneumonia prediction model.
Background
Lung cancer is one of the most frequently occurring malignant tumors. The incidence rate of the advanced lung cancer is 60-70%, and the main treatment means of the advanced lung cancer at present comprise chemotherapy, targeting and immunotherapy. For patients with EGFR, ALK, ROS isogenic mutations, targeted drug therapy can bring significant survival benefits to the patient. For patients without genetic mutations, immunotherapy is currently an important therapeutic approach, and several studies confirm the benefit of immunotherapy to the survival of patients.
The immunotherapy for clinical use is mainly based on PD1/PD-L1 immune checkpoint inhibitors, including NIVOLUMAB, PEBLIZUMAB and other medicaments, and researches show that the medicaments not only remarkably improve the survival time of patients, but also can remarkably reduce adverse reactions compared with chemotherapy.
Currently, immune checkpoint inhibitors have been widely used in clinic for the treatment of advanced non-small cell lung cancer and for consolidation treatment after radiotherapy and chemotherapy of locally advanced non-small cell lung cancer. Although the incidence of adverse reactions is low, the patient's immune adverse reactions are characterized by being difficult to predict, and part of severe immune-related pneumonia is often fatal, however, there is currently no effective prediction model for predicting the occurrence of immune-related pneumonia.
Therefore, how to provide a training method for an immune-related pneumonia prediction model, to train an effective immune-related pneumonia prediction model, so as to more accurately predict whether a subject has immune-related pneumonia, becomes one of the problems to be solved in the present day.
Disclosure of Invention
The invention provides a training method of an immune-related pneumonia prediction model, which is used for training an effective immune-related pneumonia prediction model so as to accurately predict whether a tested person has immune-related pneumonia or not, and is favorable for diagnosis of doctors and later treatment of the tested person to a certain extent.
The invention provides a training method of an immune-related pneumonia prediction model, which comprises the following steps:
acquiring N of target object 1 A plurality of clinical parameters;
acquiring N based on CT image of target object 2 Image parameters;
at said N 1 Clinical parameters and N 2 N parameters are determined in the image parameters;
an immune-related pneumonia prediction model is trained with the N parameters of a plurality of target subjects.
Optionally, the clinical parameters include: sex, age, smoking history, pathology type, immunotherapy pattern, past history of chest radiotherapy.
Optionally, the acquiring is based on N of CT images of the target object 2 A plurality of image parameters, comprising:
segmenting a lung region in a CT image of a target object;
generating a CT value histogram based on the lung region;
obtaining N based at least on the CT value histogram 2 Parameters among the image parameters.
Optionally, the generating a CT value histogram based on the lung region includes:
acquiring a CT value of each voxel in the lung region;
counting the frequency of voxels corresponding to CT values in a preset CT value interval;
and generating a CT value histogram based on the CT values in the preset CT value interval and the frequency of the voxels corresponding to the CT values.
Optionally, the preset CT value interval is (-1000 hu,0 hu), and generating the CT value histogram based on the CT values in the preset CT value interval and the frequency of the voxels corresponding to the CT values includes:
dividing the preset CT value interval into K subintervals equally, wherein K is a natural number greater than or equal to 1;
obtaining the frequency number corresponding to each subinterval, wherein the frequency number corresponding to each subinterval refers to the sum of the frequency numbers of voxels corresponding to the CT values of the subinterval;
and generating a CT value histogram by taking the frequency corresponding to each subinterval as the frequency of the voxels corresponding to each CT value of the subinterval.
Optionally, the obtaining N based at least on the CT value histogram 2 Parameters among the image parameters include:
and taking entropy, energy, kurtosis, skewness and peak value of the distribution curve of the CT value histogram as parameters in image parameters.
Optionally, the obtaining N based at least on the CT value histogram 2 The parameters in the image parameters further comprise:
the volume of the lung region is calculated,
calculating the ratio of the volume occupied by the voxels of M CT value intervals to the volume of the lung region respectively, wherein M is a natural number which is more than or equal to 1;
the M ratios are used as parameters in the image parameters.
Optionally, the obtaining N based at least on the CT value histogram 2 The parameters in the image parameters further comprise:
obtaining a standard curve;
acquiring a Hailingcase distance parameter and an IOU parameter based on a standard curve and a distribution curve of a CT value histogram;
and taking the sea-ringer distance parameter and the IOU parameter as parameters in the image parameters.
Optionally, N based on CT image of target object is acquired 2 The image parameters further include:
acquiring a CT value of each voxel in the lung region;
taking the average value, the median and the standard deviation of CT values of all voxels in the lung region as parameters in the individual image parameters.
Optionally, the method comprises the steps of 1 Clinical parameters and N 2 N parameters are determined in the image parameters, including:
normalizing the image parameter matrix;
determining importance of clinical parameters and image parameters and importance thereof based on a Borata method;
n parameters are selected from the clinical parameters and the image parameters for which importance and importance are determined by recursive feature elimination.
The invention also provides a training device of the immune-related pneumonia prediction model, which comprises the following steps:
a first acquisition unit for acquiring N of the target object 1 A plurality of clinical parameters;
a second acquisition unit for acquiring N based on the CT image of the target object 2 Image parameters;
a determining unit for determining the number of the N 1 Clinical parameters and N 2 N parameters are determined in the image parameters;
and the training unit is used for training the immune-related pneumonia prediction model with the N parameters of the target objects.
The invention also provides a computer device comprising at least one processor and at least one memory, wherein the memory stores a computer program which, when executed by the processor, enables the processor to perform the training method of the immune-related pneumonia prediction model.
The invention also provides a computer readable storage medium, which when executed by a processor within a device, causes the device to perform the above-described method of training an immune-related pneumonia prediction model.
Compared with the prior art, the technical scheme of the invention has the following beneficial effects:
first, after N of the target object is acquired 1 N based on target object CT image and clinical parameters 2 After the image parameters, the image parameters are recorded in the N 1 Clinical parameters and N 2 N parameters are determined from the image parameters. Finally, training an immune-related pneumonia prediction model by using N parameters of a plurality of target objects. The parameters are selected from clinical parameters of the target object and image parameters of the CT image of the target object when the immune-related pneumonia prediction model is trained, and final training parameters are determined based on the clinical parameters and the image parameters, so that the immune-related pneumonia prediction model with better performance can be trained, and the immune-related pneumonia prediction model with better performance can be trained at a higher speed because N parameters of each target object are finally determined when the immune-related pneumonia prediction model is trained, so that whether the testee suffers from immune-related pneumonia can be accurately predicted, and diagnosis of doctors and later treatment of the testee can be facilitated to a certain extent.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention. In the drawings:
FIG. 1 is a flow chart of a training method for an immune-related pneumonia prediction model according to an embodiment of the invention;
FIG. 2 is a schematic diagram of a CT histogram according to an embodiment of the present invention;
FIG. 3 is a flow chart of determining N parameters according to an embodiment of the present invention;
FIG. 4 is a schematic diagram showing the importance ranking of clinical parameters and image parameters according to an embodiment of the present invention;
fig. 5 is a schematic diagram of RMSE values corresponding to different parameter sets according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail below with reference to the accompanying drawings, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Fig. 1 is a flow chart of an immune-related pneumonia prediction model training method according to an embodiment of the invention. As shown in fig. 1, the training method of the immune-related pneumonia prediction model includes the following steps:
s101, obtaining N of target object 1 And clinical parameters.
S102, acquiring N based on target object CT image 2 And image parameters.
S103, at the N 1 Clinical parameters and N 2 N parameters are determined from the image parameters.
S104, training an immune-related pneumonia prediction model according to the N parameters of a plurality of target objects.
S101 is executed to obtain N of the target object 1 A clinical parameter, wherein N 1 Is a natural number of 1 or more. In this embodiment, the target subject may also be referred to as a target patient, which refers herein to a patient undergoing an immunological treatment. The clinical parameters may include, but are not limited to, one or more of gender, age, smoking history, pathology type, immunotherapy pattern, past history of chest radiotherapy. Pathological types, including squamous cell carcinoma, adenocarcinoma, small cell undifferentiated carcinoma, and large cell undifferentiated carcinoma. Immunotherapeutic modalities include treatment with immunotherapeutic agents alone and in combination with other agents. The past history of chest radiotherapy refers to whether the patient has undergone chest radiotherapy prior to the immunization therapy. Those skilled in the art will recognize that the clinical parameters are merely illustrative, and in other embodiments, the clinical parameters may include other clinical parameters, and in practical applications, different clinical parameters may be selected according to practical clinical needs.
S102 is executed to obtain N based on CT image of the target object 2 Image parameters, wherein N 2 Is a natural number of 1 or more. In this embodiment, N may be obtained by segmenting a lung region in a CT image of a target object, generating a CT value histogram based on the lung region, and obtaining N based at least on the CT value histogram 2 Parameters among the image parameters.
Specifically, in this embodiment, the lung region in the CT image of the segmented target object may be segmented into a lung (double lung) region by a thresholding method, a region growing method, a mean iterative segmentation method, a maximum inter-class variance segmentation method, a mode classification-based method, an image registration-based method, a shape model-based method, an AI algorithm (a three-dimensional convolutional neural network segmentation model, etc.). In the present embodiment, this is not limited as long as the lung region can be segmented from the CT image of the target object.
In this embodiment, the CT value histogram of each voxel in the lung region is first acquired, then the frequency of the voxels corresponding to the CT values in the preset CT value interval is counted, and finally the CT value histogram is generated based on the CT values in the preset CT value interval and the frequency of the voxels corresponding to the CT values. Those skilled in the art will recognize that the CT values corresponding to different voxels may be the same, or that the same CT value corresponds to a plurality of voxels, and that the frequency of the voxels corresponding to the CT values is the number of voxels corresponding to a certain CT value, for example, when the CT value is 100HU, the number of voxels is 12, and the frequency of the voxels corresponding to 100HU is 12. In this embodiment, the preset CT value interval may be (-1000 hu,0 hu), and the CT value histogram is generated based on the CT values in the preset CT value interval and the frequency of the voxels corresponding to the CT values, which may be specifically obtained by:
firstly, the preset CT value interval is equally divided into K subintervals, where K is a natural number greater than or equal to 1 in this embodiment. Then, the frequency number corresponding to each subinterval is obtained, wherein the frequency number corresponding to each subinterval is the sum of the frequency numbers of voxels corresponding to the CT values of the subinterval. And finally, generating a CT value histogram by taking the frequency of the voxels corresponding to the CT values of each subinterval as the frequency corresponding to the subinterval. In this embodiment, taking a preset CT value interval (-1000 HU,0 HU) as an example, the preset CT value interval is equally divided into 100 subintervals, the number of CT values included in each subinterval is 10, the number of voxels corresponding to the subinterval is (-10 HU,0 HU) as an example, if the frequency of voxels corresponding to the subinterval is 12 when the CT value is-10 HU, the frequency of voxels corresponding to the CT value is-9 HU is 15, the frequency of voxels corresponding to the CT value is-8 HU is 11, the frequency of voxels corresponding to the CT value is-7 HU is 13, the frequency of voxels corresponding to the CT value is-6 HU is 14, the frequency of voxels corresponding to the CT value is-5 HU is 12, the frequency of voxels corresponding to the CT value is-4 HU is 16, the frequency of voxels corresponding to the CT value is-3 HU is 17, the frequency of voxels corresponding to the CT value is-2 HU is 15, the frequency of voxels corresponding to the CT value is-1 HU is 12, the frequency of voxels corresponding to the CT value is-0 HU is 13, the frequency of voxels corresponding to the CT value is-10 HU, and the frequency of voxels corresponding to the CT value is 150 HU (-10 HU) can be generated as a subinterval of the corresponding subintervals of the subintervals of 150. Fig. 2 is a schematic diagram of a CT value histogram of an embodiment of the present invention, where the current curve and the black region covered by the current curve in fig. 2 are CT value histograms of lung regions. In fig. 2, the abscissa represents a CT value, and the ordinate represents the frequency of voxels corresponding to the CT value. In fig. 2, the CT value interval (-1000 HU,0 HU) is divided into 10 sub-intervals (-1000 HU, -900 HU), (-900 HU, -800 HU), (-800 HU, -700 HU), … …, (-100 HU,0 HU), and then each of the 10 sub-intervals is divided into 10 sub-intervals, that is, 100 sub-intervals are divided into (-1000 HU,0 HU), and the frequency of voxels corresponding to each CT value in each sub-interval is the same, that is, the frequency corresponding to the sub-interval. In this embodiment, the preset CT value interval is (-1000 hu,0 hu), and the preset CT value interval is equally divided into 100 sub-intervals, and in other embodiments, the number of the preset CT value interval and the number of the sub-intervals may be determined according to the actual clinical requirement, so that the preset CT value interval is (-1000 hu,0 hu), and the number of the sub-intervals is 100 should not be taken as a limitation of the technical solution of the present invention.
The CT value histogram of the target object lung region is generated through the above-mentioned process. In this embodiment, N may be obtained based at least on the CT value histogram 2 Parameters among the image parameters. Specifically, entropy, energy, kurtosis, skewness and peak of the distribution curve of the CT value histogram may be used as parameters in the image parameters.
The entropy of the distribution curve of the CT value histogram is obtained by the following formula:
the skewness of the distribution curve of the CT value histogram is obtained by the following formula:
kurtosis of the distribution curve of the CT value histogram is obtained by the following formula:
the energy of the distribution curve of the CT value histogram is obtained by the following formula:
the peak value of the distribution curve of the CT value histogram is obtained by the following formula:
peak＝l max
where l is the CT value of each voxel, l min For the minimum value of all CT values, l max For the maximum value of all CT values, N (l) is the number of voxels corresponding to each CT value, m is the average value of all CT values, s is the standard deviation of all CT values, N is the number of all voxels, and c is a common constant.
The volume of the lung region may be calculated, and the ratio of the volume occupied by the voxels of the M CT value intervals to the volume of the lung region may be calculated, where M is a natural number greater than or equal to 1. The M ratios are used as parameters in the image parameters. Specifically, the volume of the lung region may be determined by the number of voxels in the lung region and a preset scale. The preset scale may be empirically set. Similarly, the volume occupied by each voxel in each CT value interval is the product of the number of voxels in the CT value interval and the preset scale. The number of voxels in the CT interval can be directly obtained from the frequency number of the corresponding interval in the CT histogram. In this embodiment, M may be 100, that is, the volume occupied by the voxels of 100 CT value intervals is calculated, for example, 100 intervals shown in fig. 2, and the number of voxels of each interval is the frequency number corresponding to the interval. In other embodiments, M may take different values depending on the actual clinical need.
Furthermore, a standard curve may be obtained. And acquiring the Hailingcase distance parameter and the IOU parameter based on the standard curve and the distribution curve of the CT value histogram. And taking the sea-ringer distance parameter and the IOU parameter as parameters in the image parameters. With continued reference to fig. 2, the baseline curve in fig. 2 is a standard curve, where the standard curve is generally obtained based on CT values of a normal human lung region, and specifically may be obtained by averaging the frequency of voxels corresponding to the same CT value among a large number of CT values of a normal human lung region. For example: when the CT value of the lung area of the target object 1 is-700, the frequency of the corresponding voxel is 50, when the CT value of the lung area of the target object 2 is-700, the frequency of the corresponding voxel is 60, when the CT value of the lung area of the target object 3 is-700, the frequency of the corresponding voxel is 50, when the CT value of the lung area of the target object 4 is-700, the frequency of the corresponding voxel is 80 and … …, when the CT value of the lung area of the target object Y is-700, the frequency of the corresponding voxel on the reference curve is the sum of the frequency of the corresponding voxels and the ratio of Y when the CT value of the target object Y is-700, and for the same CT value-700, the frequency of the corresponding voxel on the reference curve is the target object 1 and … …. After determining the average frequency of the voxels corresponding to each CT value, the standard curve can be obtained by taking the CT value as the abscissa and the average frequency of the voxels corresponding to the CT value as the ordinate.
In this embodiment, the sea-ringer distance parameter is obtained by the following formula:
wherein p is i For points on the CT value histogram curve, q i Is a point on the standard curve.
The closer the Hailin value, H (P, Q), is to 0, the higher the similarity of the CT histogram of the target subject's lung region to the standard curve distribution is indicated.
In this embodiment, the IOU parameter is obtained by the following formula:
here, area (a) is the Area under the curve of the CT value histogram, and Area (B) is the Area under the standard curve. When the IOU is closer to 1, the similarity between the CT value histogram curve and the standard curve of the lung region of the target object is higher. In this embodiment, the similarity between the CT value histogram distribution of the lung region of the target object and the standard curve can be determined by the size of the IOU.
In this embodiment, except that N can be obtained based on at least the CT value histogram 2 In addition to the parameters in the image parameters, N based on the CT image of the target object can be obtained by 2 Parameters of the image parameters, in particular, CT values of each voxel in the lung region are acquired. Taking the average value, the median and the standard deviation of CT values of all voxels in the lung region as N 2 Parameters among the image parameters.
The average of the CT values for all voxels in the lung region is obtained by the following formula:
the standard deviation of CT values for all voxels in the lung region is obtained by the following formula:
where l is the CT value of each voxel, l min For the minimum value of all CT values, l max For the maximum value of all CT values, m is the average value of all CT values, and N is the number of all voxels.
The median of the CT values of all voxels in the lung region means that the CT values of all voxels are arranged in a predetermined order (e.g., from small to large or from large to small), and the CT value of the voxel located in the middle position is taken as the median of the CT values of all voxels in the ordered queue. If the ordered queue is the 1 st to 2n+1 st voxels, the CT value of the n+1 st voxel is the median of the CT values of all voxels, and if the ordered queue is the 1 st to 2n nd voxels, the CT value of the n or n+1 th voxel is the median of the CT values of all voxels. For example, if n=4, the CT values of the 1 st to 9 th voxels are arranged in order from the smallest to the largest, and the CT value of the 5 th voxel is the median of the CT values of the 9 th voxels. If n=3, the CT values of the 3 rd or 4 th voxels are the median of the CT values of the 6 th voxels, if the 1 st to 6 th voxels are sequentially arranged in order of the CT values from the smallest.
In the present embodiment, after N of the target object is acquired 1 Clinical parameters and N 2 After the image parameters, at N 1 Clinical parameters and N 2 N parameters are determined from the image parameters, wherein N is<N 1 +N 2 . Referring to fig. 3, fig. 3 is a flow chart illustrating a determination of N parameters according to an embodiment of the present invention, specifically including:
s301: and normalizing the image parameter matrix.
S302: the importance of clinical parameters and image parameters and the importance thereof are determined based on the Boruta method.
S303: n parameters are selected from the clinical parameters and the image parameters for which importance and importance are determined by recursive feature elimination.
S301 is specifically performed, the image parameter matrix may be normalized by a minimum-maximum normalization method, a Z-Score normalization method, an L2 normalization method, or the like.
S302 is performed, specifically,
1. generating a feature matrix by taking clinical parameters and image parameters as feature values, randomly sequencing (shuffle) the feature values in the feature matrix, and splicing the feature (shuffle features) values after the shuffle with the original feature values (real features) to form a new feature matrix.
2. The new feature matrix is used as an input of a feature importance model to output the importance and importance of each feature. The feature importance model may be RandomForest, lightGBM, XGBoost, etc.
3. Calculating Z_score of real features and shadow features, wherein Z_score is obtained by the following formula:
z_score=average_feature/feature_importance standard deviation
Wherein feature_importance= oob _acc-oob _acc_after_permeation of a single feature over a single number.
Where oob _acc_after_persistence refers to the accuracy of the sample of out_of_bag on a single tree after the dimension feature is shuffled.
4. The largest Z_score found in the shadow features is denoted as Z_max.
5. Real features with Z_score greater than Z_max are marked as "important", real features with Z_score significantly less than Z_max are marked as "unimportant", and permanently deleted from the feature set.
6. All shadow features are deleted.
7. 1-6 are repeated until all real features are marked as "important" or "unimportant".
For real features labeled "important", its Z_score can be taken as the importance of the real features.
Fig. 4 is a schematic diagram of importance ranking of clinical parameters and image parameters according to an embodiment of the present invention, as shown in fig. 4, the abscissa in fig. 4 is the clinical parameters and the image parameters, and the ordinate is the importance. In practical applications, for the same clinical parameters or image parameters, the importance of the selected samples is not a constant value (the importance of the same clinical parameters or image parameters is different for different target objects), but is distributed in a predetermined range. In this embodiment, the median of the importance of the distribution range is taken as the importance of the parameter, and in fig. 4, the black horizontal line in the columnar mark represents the importance of the parameter. As can be seen from fig. 4, the importance of the clinical parameters and the image parameters is mainly distributed between 0 and 8, and it is apparent that 6 parameters of the clinical parameters such as age, pathology type, past chest radiotherapy history, smoking history, past therapy history, sex, and double lung volume parameters are relatively low in importance in fig. 4, and the importance of 11 th parameter, 12 th parameter, 13 th parameter … … 30 th parameter is significantly higher than the importance of 2 nd parameter, 3 rd parameter, 4 th parameter, 6 th parameter … … 9 th parameter from left to right in fig. 4.
After determining the importance of clinical parameters and image parameters and their importance by the Boruta method, in order to find the best parameters that can train the immune-related pneumonia prediction model, S303 is performed, and N parameters are selected from the clinical parameters and image parameters having importance and importance by recursive feature elimination.
In order to make the result output by the immune-related pneumonia prediction model most accurate when selecting parameters in clinical parameters and image parameters with importance and importance through recursive feature elimination, the parameters are selected through recursive feature elimination, and meanwhile, the deviation between a predicted value and a true value can be measured through a RMSE value to determine a finally selected parameter set.
In this embodiment, N parameters may be selected from a plurality of parameters with higher importance according to the importance, and in practical application, a parameter with importance greater than a predetermined threshold may be used as the parameter with higher importance, and the predetermined threshold may be determined empirically. In this embodiment, N parameters may be selected from the 11 th parameter … … 30 th parameter in fig. 4.
Specifically, a preset weight may be given to each parameter, and then training is performed using an immune-related pneumonia prediction model, and the weight of each parameter is adjusted so that the loss function is smaller than a preset value. After the weights of the adjusted parameters are obtained, absolute values of the weights are taken, and the parameters corresponding to the minimum absolute value weights are removed. The foregoing process is then repeated among the remaining parameters to select a plurality of parameters, in this embodiment less than 20 parameters are selected by recursive feature elimination. In order to make the result output by the immune-related pneumonia prediction model most accurate, any combination is carried out on the parameters selected by the recursive feature elimination to obtain each group of parameters, and a plurality of samples corresponding to each group of parameters are used as the input of the immune-related pneumonia prediction model to obtain the RMSE value corresponding to the group of parameters. Those skilled in the art will appreciate that RMSE values reflect how close a predicted value is to a real value, the smaller the RMSE value, the closer the predicted value is to the real value. Therefore, in this embodiment, a set of parameters corresponding to the minimum RMSE value is used as the N parameters to be finally selected.
Referring to fig. 5, fig. 5 is a schematic diagram of RMSE values corresponding to different parameter sets in the embodiment of the present invention, and it should be noted that, the abscissa in fig. 5 only illustrates the number of parameters included in the parameter set, and in practical application, the parameter set further includes which parameters are specifically included in the parameter set, in this embodiment, 8 parameters corresponding to the abscissa 8 in fig. 5 are a haligex distance parameter, energy of a distribution curve of a CT value histogram, kurtosis of a distribution curve of a CT value histogram, entropy of a distribution curve of a CT value histogram, a ratio of a volume occupied by a voxel with a CT value of (-1000 HU, -900 HU), a ratio of a volume occupied by a voxel with a CT value of (-500 HU, -400 HU), a ratio of a volume occupied by a voxel with a CT value of (-300 HU, -200 HU), and a ratio of a volume occupied by a voxel with a CT value of (-200 HU, -100 HU). As can be seen from fig. 5, for the immune-related pneumonia prediction model, when the input parameters are the aforementioned 8 parameters, the RMSE value is the smallest, that is, the predicted value and the true value are the closest, or when the input parameters are the aforementioned 8 parameters, the prediction result of the immune-related pneumonia prediction model is the most accurate, and the aforementioned 8 input parameters can fully benefit the immune-related pneumonia prediction model, and when the input parameters are not the aforementioned 8 input parameters, the immune-related pneumonia prediction model may be unstable.
Up to this point, 8 parameters for training the immune-related pneumonia prediction model were determined through the above-mentioned procedure, and in this embodiment, the 8 parameters may be a hain lattice distance parameter, energy of a distribution curve of a CT value histogram, kurtosis of a distribution curve of a CT value histogram, entropy of a distribution curve of a CT value histogram, a ratio of a volume occupied by a voxel with a CT value of (-1000 HU, -900 HU) to a lung volume, a ratio of a volume occupied by a voxel with a CT value of (-500 HU, -400 HU) to a lung volume, a ratio of a volume occupied by a voxel with a CT value of (-300 HU, -200 HU) to a lung volume, and a ratio of a volume occupied by a voxel with a CT value of (-200 HU, -100 HU) to a lung volume. The four parameters, i.e. the sea-ringer distance parameter, the energy of the distribution curve of the CT value histogram, the kurtosis of the distribution curve of the CT value histogram and the entropy of the distribution curve of the CT value histogram, can be used for comparing the differences of the parameters of different target objects on the distribution curve of the CT value histogram, such as the form of the distribution curve, the first-order characteristics of the distribution curve and the like. The ratio of the volume occupied by the voxels in the four CT value intervals to the lung volume can reflect the difference of the lung of the target object in tissue density distribution, and in practical application, the target object can be distinguished to belong to a pneumonia group or a non-pneumonia group by comparing the 8 parameters of different target objects.
S104 is performed to train an immune-related pneumonia prediction model with N parameters of the plurality of finally obtained target subjects. The immune-related pneumonia prediction model can be trained by using a large number of 8 parameters of the target object, so that an immune-related pneumonia prediction result is obtained. In this embodiment, the immune-related pneumonia prediction model may be a U-NET neural network model using VGG as a backup, or may be a full convolution neural network model.
The embodiment of the invention also provides a training device of the immune-related pneumonia prediction model, which comprises the following steps:
a first acquisition unit for acquiring N of the target object 1 A plurality of clinical parameters;
a second acquisition unit for acquiring N based on the CT image of the target object 2 Image parameters;
a determining unit for determining the number of the N 1 Clinical parameters and N 2 N parameters are determined in the image parameters;
and the training unit is used for training the immune-related pneumonia prediction model with the N parameters of the target objects.
The implementation of the training device of the immune-related pneumonia prediction model in this embodiment may refer to the implementation of the training method of the immune-related pneumonia prediction model, and will not be described herein.
Based on the same technical concept, the embodiment of the invention provides a computer device, which comprises at least one processor and at least one memory, wherein the memory stores a computer program, and when the program is executed by the processor, the processor is enabled to execute the training method of the immune-related pneumonia prediction model.
Based on the same technical concept, embodiments of the present invention provide a computer-readable storage medium, which when executed by a processor within a device, enables the device to perform the above-described training method of an immune-related pneumonia prediction model.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, or as a computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (10)
1. A training method of an immune-related pneumonia prediction model is characterized in that,
acquiring N of target object 1 A plurality of clinical parameters;
acquiring N based on CT image of target object 2 Image parameters;
at said N 1 Clinical parameters and N 2 N parameters are determined in the image parameters;
training an immune-related pneumonia prediction model with the N parameters of a plurality of target subjects;
the acquisition is based on N of CT images of the target object 2 A plurality of image parameters, comprising:
segmenting a lung region in a CT image of a target object;
generating a CT value histogram based on the lung region;
obtaining N based at least on the CT value histogram 2 Parameters in the image parameters;
said obtaining N based at least on said CT value histogram 2 Of individual image parametersParameters, further comprising:
obtaining a standard curve;
acquiring a Hailingcase distance parameter and an IOU parameter based on a standard curve and a distribution curve of a CT value histogram;
taking the sea-ringer distance parameter and the IOU parameter as parameters in the image parameters;
the standard curve is obtained based on CT values of the lung regions of normal people, specifically, the frequency of voxels corresponding to the same CT value in a large number of CT values of the lung regions of normal people is averaged, after the average frequency of the voxels corresponding to each CT value is determined, the CT value is taken as an abscissa, and the average frequency of the voxels corresponding to the CT value is taken as an ordinate to obtain the standard curve;
the sea-ringer distance parameter H (P, Q) is obtained by the following formula:
wherein p is i For points on the CT value histogram curve, q i Points on the standard curve;
the IOU parameter is obtained by the following formula:
here, area (a) is the Area under the curve of the CT value histogram, and Area (B) is the Area under the standard curve.
2. The method of claim 1, wherein,
the clinical parameters include: sex, age, smoking history, pathology type, immunotherapy pattern, past history of chest radiotherapy.
3. The method of claim 1, wherein the generating a CT value histogram based on the lung region comprises:
acquiring a CT value of each voxel in the lung region;
counting the frequency of voxels corresponding to CT values in a preset CT value interval;
and generating a CT value histogram based on the CT values in the preset CT value interval and the frequency of the voxels corresponding to the CT values.
4. The method of claim 3, wherein,
the preset CT value interval is (-1000 hu,0 hu), and generating the CT value histogram based on the CT values in the preset CT value interval and the frequency of the voxels corresponding to the CT values includes:
dividing the preset CT value interval into K subintervals equally, wherein K is a natural number greater than or equal to 1;
obtaining the frequency number corresponding to each subinterval, wherein the frequency number corresponding to each subinterval refers to the sum of the frequency numbers of voxels corresponding to the CT values of the subinterval;
and generating a CT value histogram by taking the frequency corresponding to each subinterval as the frequency of the voxels corresponding to each CT value of the subinterval.
5. The method of claim 1, wherein the obtaining N based at least on the CT value histogram 2 Parameters among the image parameters include:
and taking entropy, energy, kurtosis, skewness and peak value of the distribution curve of the CT value histogram as parameters in image parameters.
6. The method of claim 1, wherein the obtaining N based at least on the CT value histogram 2 The parameters in the image parameters further comprise:
the volume of the lung region is calculated,
calculating the ratio of the volume occupied by the voxels of M CT value intervals to the volume of the lung region respectively, wherein M is a natural number which is more than or equal to 1;
the M ratios are used as parameters in the image parameters.
7.The method of claim 1, wherein N based on CT images of the target object are acquired 2 The image parameters further include:
acquiring a CT value of each voxel in the lung region;
taking the average value, the median and the standard deviation of CT values of all voxels in the lung region as parameters in the individual image parameters.
8. The method of claim 1, wherein said at said N 1 Clinical parameters and N 2 N parameters are determined in the image parameters, including:
normalizing the image parameter matrix;
determining importance of clinical parameters and image parameters and importance thereof based on a Borata method;
n parameters are selected from the clinical parameters and the image parameters for which importance and importance are determined by recursive feature elimination.
9. The method of claim 1, wherein the closer the hain lattice distance parameter value, H (P, Q), is to 0, the higher the similarity of the CT value histogram of the target subject's lung region to the standard curve distribution.
10. The method of claim 1, wherein the closer the IOU parameter value is to 1, the higher the similarity of the CT value histogram curve to the standard curve for the lung region of the target subject;
and judging the similarity between the distribution of the CT value histogram of the lung region of the target object and the standard curve according to the magnitude of the IOU parameter value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010407606.2A CN111739645B (en) | 2020-05-14 | 2020-05-14 | Training method of immune-related pneumonia prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010407606.2A CN111739645B (en) | 2020-05-14 | 2020-05-14 | Training method of immune-related pneumonia prediction model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111739645A CN111739645A (en) | 2020-10-02 |
| CN111739645B true CN111739645B (en) | 2024-01-30 |
Family
ID=72647200
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010407606.2A Active CN111739645B (en) | 2020-05-14 | 2020-05-14 | Training method of immune-related pneumonia prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111739645B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102429679A (en) * | 2011-09-09 | 2012-05-02 | 华南理工大学 | Computer-assisted emphysema analysis system based on chest CT (Computerized Tomography) image |
| CN103793711A (en) * | 2014-01-17 | 2014-05-14 | 首都医科大学 | Multidimensional vein extracting method based on brain nuclear magnetic resonance image |
| CN104933711A (en) * | 2015-06-10 | 2015-09-23 | 南通大学 | Automatic fast segmenting method of tumor pathological image |
| CN106981068A (en) * | 2017-04-05 | 2017-07-25 | 重庆理工大学 | A kind of interactive image segmentation method of joint pixel pait and super-pixel |
| CN109658411A (en) * | 2019-01-21 | 2019-04-19 | 杭州英库医疗科技有限公司 | A kind of correlation analysis based on CT images feature Yu Patients with Non-small-cell Lung prognosis situation |
| CN109919973A (en) * | 2019-02-19 | 2019-06-21 | 上海交通大学 | Multi-angle of view target association method, system and medium based on multiple features combining |
| CN110163891A (en) * | 2019-04-30 | 2019-08-23 | 长沙师范学院 | Long term object tracking based on sensor model |
| CN110391015A (en) * | 2019-06-14 | 2019-10-29 | 广东省人民医院(广东省医学科学院) | A method of tumor immunity is quantified based on image group |
| CN110766723A (en) * | 2019-10-22 | 2020-02-07 | 湖南大学 | Unmanned aerial vehicle target tracking method and system based on color histogram similarity |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101953626B1 (en) * | 2017-06-29 | 2019-03-06 | 서강대학교산학협력단 | Method of tracking an object based on multiple histograms and system using the method |
-
2020
- 2020-05-14 CN CN202010407606.2A patent/CN111739645B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102429679A (en) * | 2011-09-09 | 2012-05-02 | 华南理工大学 | Computer-assisted emphysema analysis system based on chest CT (Computerized Tomography) image |
| CN103793711A (en) * | 2014-01-17 | 2014-05-14 | 首都医科大学 | Multidimensional vein extracting method based on brain nuclear magnetic resonance image |
| CN104933711A (en) * | 2015-06-10 | 2015-09-23 | 南通大学 | Automatic fast segmenting method of tumor pathological image |
| CN106981068A (en) * | 2017-04-05 | 2017-07-25 | 重庆理工大学 | A kind of interactive image segmentation method of joint pixel pait and super-pixel |
| CN109658411A (en) * | 2019-01-21 | 2019-04-19 | 杭州英库医疗科技有限公司 | A kind of correlation analysis based on CT images feature Yu Patients with Non-small-cell Lung prognosis situation |
| CN109919973A (en) * | 2019-02-19 | 2019-06-21 | 上海交通大学 | Multi-angle of view target association method, system and medium based on multiple features combining |
| CN110163891A (en) * | 2019-04-30 | 2019-08-23 | 长沙师范学院 | Long term object tracking based on sensor model |
| CN110391015A (en) * | 2019-06-14 | 2019-10-29 | 广东省人民医院(广东省医学科学院) | A method of tumor immunity is quantified based on image group |
| CN110766723A (en) * | 2019-10-22 | 2020-02-07 | 湖南大学 | Unmanned aerial vehicle target tracking method and system based on color histogram similarity |
Non-Patent Citations (3)
| Title |
|---|
| 基于BS-HMM和巴式距离的手势识别技术研究;朱正伟;祝磊;饶鹏;;计算机应用与软件(06);全文 * |
| 基于海林格距离和SMOTE的多类不平衡学习算法;董明刚;姜振龙;敬超;;计算机科学(01);全文 * |
| 基于集成分类器的用户属性预测研究;王斯盾;琚生根;周刚;刘玉娇;;四川大学学报(自然科学版)(06);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111739645A (en) | 2020-10-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10939874B2 (en) | Deep-learning-based cancer classification using a hierarchical classification framework | |
| US11350901B2 (en) | Recurrence prognosis and prediction of added benefit of adjuvant chemotherapy in early stage non-small cell lung cancer with radiomic features on baseline computed tomography | |
| CN107077736A (en) | System and method according to the Image Segmentation Methods Based on Features medical image based on anatomic landmark | |
| CN110604550A (en) | Prediction method of normal tissue organ complications after tumor radiotherapy | |
| JP7034306B2 (en) | Region segmentation device, method and program, similarity determination device, method and program, and feature quantity derivation device, method and program | |
| JP2020032044A (en) | Similarity determination device, method, and program | |
| Irene et al. | Segmentation and approximation of blood volume in intracranial hemorrhage patients based on computed tomography scan images using deep learning method | |
| Yang et al. | Learning COVID-19 pneumonia lesion segmentation from imperfect annotations via divergence-aware selective training | |
| WO2018219818A1 (en) | Quantified aspects of lesions in medical images | |
| US11055844B2 (en) | Predicting response to immunotherapy using computer extracted features of cancer nuclei from hematoxylin and eosin (HandE) stained images of non-small cell lung cancer (NSCLC) | |
| Davamani et al. | Biomedical image segmentation by deep learning methods | |
| CN111739645B (en) | Training method of immune-related pneumonia prediction model | |
| US10839513B2 (en) | Distinguishing hyperprogression from other response patterns to PD1/PD-L1 inhibitors in non-small cell lung cancer with pre-therapy radiomic features | |
| CN115274119B (en) | Construction method of immunotherapy prediction model fusing multi-image mathematical characteristics | |
| Zhang et al. | Dial/Hybrid cascade 3DResUNet for liver and tumor segmentation | |
| CN116740386A (en) | Image processing method, apparatus, device and computer readable storage medium | |
| CN111738980B (en) | Medical image display method, computer equipment and storage medium | |
| US11615881B2 (en) | Liver cancer detection | |
| CN113706541B (en) | Image processing method and device | |
| D'Amico et al. | Early radiomics experiences in predicting CyberKnife response in acoustic neuroma | |
| CN115458160B (en) | Whole-course intelligent management system, method, equipment and medium for breast tumor patients | |
| Indumathi et al. | Machine Learning based Lung Cancer Detection & Analysis | |
| Le et al. | Initial Approach to the Identification Degree of Skin Damage and Classification of Acne by YOLOv8 | |
| EP4053853A1 (en) | Apparatus for determining a temporal breast cancer risk | |
| Khan et al. | Automatic Segmentation and Shape, Texture-based Analysis of Glioma Using Fully Convolutional Network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20220318 Address after: 100089 area a, 21 / F, block a, No. 8, Haidian Street, Haidian District, Beijing Applicant after: BEIJING SHENRUI BOLIAN TECHNOLOGY Co.,Ltd. Applicant after: SHENZHEN DEEPWISE BOLIAN TECHNOLOGY Co.,Ltd. Address before: 200051 units 06 and 07, 23 / F, 523 Loushanguan Road, Changning District, Shanghai Applicant before: SHANGHAI YIZHI MEDICAL TECHNOLOGY Co.,Ltd. |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |