CN113077417A - Method and system for predicting hepatocellular carcinoma microvascular invasion - Google Patents

Abstract

The application provides a prediction method and a system for hepatocellular carcinoma microvascular invasion, which take a received image of a patient to be detected as input to a pre-trained deep neural network for focus segmentation so as to obtain a region related to a focus on the image; and extracting features based on the region related to the focus and inputting the features into a pre-trained microvascular invasion prediction model to predict whether the patient has liver cancer MVI. The scheme realizes a full-automatic process for predicting the liver cancer MVI, is not influenced by the subjective experience of doctors any more, and improves the efficiency and the accuracy of predicting the liver cancer MVI before operation.

Description

Method and system for predicting hepatocellular carcinoma microvascular invasion

Technical Field

The invention relates to the field of medical informatization, in particular to a system and a method for predicting hepatocellular carcinoma microvascular invasion (also called liver cancer MVI for short) before an operation by utilizing a computer.

Background

Hepatocellular carcinoma (HCC) is one of the most common primary hepatic malignancies, with increasing incidence worldwide. With the development of medical technology, the therapeutic effect of liver cancer has been greatly improved, and surgical operations such as surgical resection and liver transplantation are still the most effective treatment methods. However, the recurrence rate after liver cancer surgery is high, the recurrence rate 5 years after surgical resection is about 70%, and the recurrence rate 5 years after liver transplantation is about 35%. There are studies that demonstrate microvascular invasion (MVI) is an independent risk factor affecting recurrence and metastasis after HCC treatment. Preoperative prediction of MVI is of great significance in assessing HCC patient prognosis, selecting surgical protocols, and formulating treatment protocols for anti-relapsing metastasis. For example, if a patient with a high risk of MVI is undergoing a hepatectomy therapy, an anatomical hepatectomy based on the Couinaud segment may be selected. Compared with non-anatomical resection, the anatomical resection can completely remove portal vein support with tumors, more effectively eradicate the intrahepatic MVI and reduce the recurrence rate. Patients with HCC who receive liver transplants who are not at risk for MVI have been shown to have better prognostic outcome. Therefore, accurate prediction of MVI is an important breakthrough in improving the prognosis of HCC patients.

The use of preoperative biopsy to detect MVI is not feasible due to sampling errors and the possibility of implant metastases resulting from intratumoral heterogeneity. The current preoperative research for predicting the MVI of liver cancer is generally based on the imaging omics (Radiomics) method, which extracts imaging omics characteristics from a large number of medical image images by means of a computer, and uses a statistical and/or machine learning method to classify or predict the liver cancer for disease characterization, tumor grading and staging, curative effect evaluation, prognosis prediction and the like. And the extraction of the image omics features often depends on a focus region (such as a cancer or tumor region) marked and segmented on an image by a professional doctor. The manual labeling and dividing method not only consumes much labor and time, is difficult to process a large sample amount, but also has strong subjectivity, because doctors of different levels may make different judgments due to different personal experiences, and thus, the regions labeled by different doctors have subjective deviations. These factors will undoubtedly have an impact on the efficiency and accuracy of preoperative prediction of liver cancer MVI.

Disclosure of Invention

Therefore, an object of the embodiments of the present invention is to provide a fully automatic method and system for predicting liver cancer MVI. The above purpose is realized by the following technical scheme:

according to a first aspect of the embodiments of the present invention, a method for predicting hepatocellular carcinoma microvascular invasion is provided, which includes providing a received image of a patient to be detected as an input to a pre-trained deep neural network for lesion segmentation to obtain a region related to a lesion on the image; and extracting features based on the region related to the focus on the image and inputting the features into a pre-trained microvascular invasion prediction model to predict whether the hepatocellular carcinoma microvascular invasion exists in the patient.

In some embodiments of the present invention, the microvascular invasion prediction model may employ a deep neural network model.

In some embodiments of the invention, the image may be an MRI image. In some embodiments of the invention, the imagery images may be MRI images of different phases.

In some embodiments of the present invention, the features extracted based on the region on the image associated with the lesion may be an iconomics feature.

In some embodiments of the present invention, the method may further comprise extracting clinical features of the patient to be examined and providing the clinical features as input to the microvascular invasion prediction model together with features extracted based on a region on the image associated with the lesion.

In some embodiments of the invention, the clinical features may be derived from one or more of the following: blood test results related to the patient to be tested, basic disease information related to the patient to be tested, symptom information related to the patient to be tested, and/or personal basic information related to the patient to be tested.

In some embodiments of the present invention, the deep neural network for lesion segmentation may be pre-trained based on a sample set containing video images labeled with regions associated with lesions segmented by an expert.

According to a second aspect of the embodiments of the present invention, there is also provided a system for predicting hepatocellular carcinoma microvascular invasion, which includes an automatic segmentation module, a feature extraction module, and a prediction module. The automatic segmentation module is used for providing the received image of the patient to be detected as input to a pre-trained deep neural network for lesion segmentation so as to obtain a region related to a lesion on the image. The feature extraction module is used for extracting features based on a region related to the focus on the image. The prediction module is used for inputting the features extracted by the feature extraction module into a pre-trained microvascular invasion prediction model to predict whether the hepatocellular carcinoma microvascular invasion exists in the patient.

According to a third aspect of embodiments of the present invention, there is also provided a computer-readable storage medium having stored thereon a computer program which, when executed, implements the method as described in the first aspect of the embodiments above.

The technical scheme provided by the embodiment of the invention has the following beneficial effects:

according to the scheme, the lesion region on the image can be automatically segmented, a full-automatic process for predicting the liver cancer MVI is realized, manual marking is not relied on, the influence of subjective experience of doctors is avoided, and the efficiency and the accuracy of predicting the liver cancer MVI before an operation are improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention. It is obvious that the drawings in the following description are only some embodiments of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort. In the drawings:

fig. 1 is a flowchart illustrating a method for predicting the MVI of liver cancer according to an embodiment of the present invention.

Fig. 2 is a schematic diagram illustrating a model structure of a neural network for predicting the MVI of liver cancer according to an embodiment of the present invention.

Fig. 3 is a schematic structural diagram of a system for predicting the MVI of liver cancer 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 is further described in detail by embodiments with reference to the accompanying drawings. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to provide a thorough understanding of embodiments of the invention. One skilled in the relevant art will recognize, however, that the invention may be practiced without one or more of the specific details, or with other methods, components, devices, steps, and so forth. In other instances, well-known methods, devices, implementations or operations have not been shown or described in detail to avoid obscuring aspects of the invention.

The block diagrams shown in the figures are functional entities only and do not necessarily correspond to physically separate entities. I.e. these functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor means and/or microcontroller means.

The flow charts shown in the drawings are merely illustrative and do not necessarily include all of the contents and operations/steps, nor do they necessarily have to be performed in the order described. For example, some operations/steps may be decomposed, and some operations/steps may be combined or partially combined, so that the actual execution sequence may be changed according to the actual situation.

Fig. 1 is a flowchart illustrating a method for predicting hepatocellular carcinoma microvascular invasion according to an embodiment of the present invention. As shown in fig. 1, the method 100 generally includes: step S101) providing the received image of the patient to be detected as input to a pre-trained deep neural network for lesion segmentation to obtain a region related to a lesion on the image; and step S102) extracting characteristics based on the area related to the focus on the image and inputting the characteristics to a pre-trained microvascular invasion prediction model to predict whether the hepatocellular carcinoma microvascular invasion exists in the patient.

More specifically, in step S101, the received image of the patient to be detected is provided as an input to a pre-trained deep neural network for lesion segmentation to automatically segment a region on the image related to the lesion. In some embodiments, the image of the patient may be an MRI image obtained by Magnetic Resonance Imaging (MRI) of the patient, and the deep neural network for lesion segmentation may be trained on MRI image of different phases to identify a region on the image associated with a lesion. In some embodiments, the method further comprises the steps of generating a sample set and pre-training a deep neural network for lesion segmentation based on the sample set. The following describes how to train a deep neural network for lesion segmentation, taking MRI images as an example. For MRI images, the corresponding sample set and test set for training contains a plurality of MRI images to which the region associated with the lesion segmented by the expert has been labeled. For the accuracy of the subsequent lesion region segmentation, the lesion region on the MRI image may be segmented and labeled by a medical expert with a very high experience in this respect when acquiring the sample set for training. For example, in order to more accurately label a region related to a lesion on an MRI image, 2 radiologists may respectively perform independent segmentation on an MRI image, a medical expert may check whether the two segmentations are consistent, and another radiologist medical expert with a high experience (for example, with an MRI reading experience of more than 15 years) may check the segmentation quality and improve the final segmentation effect. Through experimental analysis and comparison of the inventor, the deep neural network trained on the sample set and the test set marked in the way and used for lesion segmentation sufficiently considers the expert experience, and the recognition efficiency exceeds the level of a common doctor.

Since a lesion region (e.g., a tumor region) on an MRI image differs from other parts in terms of image texture, gray scale, color change, edge characteristics, and the like, each parameter of a deep neural network serving as a recognition model is continuously adjusted by using an already labeled lesion region as a target region to be recognized during training. The deep neural network for lesion segmentation takes the MRI image as input, and its output is the identified target region (i.e., lesion region) on the MRI image. In training, the features of the labeled lesion region extracted from the sample MRI image may be, for example, Histogram of Oriented Gradient (HOG) features, Local Binary Pattern (LBP) operators, Haar-like features, and so on. In some embodiments, the deep neural network model employed is a Res-uet network model, a uet network model, or a full convolutional network model (FCN). In other embodiments, the deep neural network model may be a Gated Recurrent Units (GRU), Long-short memory network (LSTM), Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), or the like. In the embodiment of the present invention, the features and network models used for training the deep neural network for lesion segmentation are not particularly limited, and any features and network models that can help to identify a target region on an MRI image may be applied to the solution of the present application.

In the actual working process, because the MRI images are relatively complex, even if doctors with corresponding experience mark the lesion area on each MRI image, the time is as much as tens of minutes, and various subjective errors exist. The trained deep neural network for lesion segmentation can rapidly and automatically mark and segment a lesion region on an image of a patient to be detected, which is received each time, so that the time and labor cost of manual marking are reduced, the influence of subjective experience of a doctor is avoided, and the efficiency and the accuracy of preoperative liver cancer MVI prediction are improved.

Continuing with fig. 1, in step S102), features are extracted based on the region on the image related to the lesion and input to a pre-trained microvascular invasion prediction model to predict whether the patient has hepatocellular carcinoma microvascular invasion. In some embodiments, the microvascular invasion predictive model may employ a deep neural network model such as RNN, CNN, LSTM, GRU. In still other embodiments, the microvascular invasion prediction model may also employ other types of machine learning models, such as support vector machines, random forests, decision trees, and the like.

In fact, the accuracy of the microvascular invasion prediction model depends largely on the extraction and selection of features. In some embodiments, the features extracted based on the region on the image associated with the lesion may be image omics features such as, for example, first order statistics, gray level co-occurrence matrix (GLCM), Gray Level Run Length Matrix (GLRLM), gray level size region matrix (GLSZM), Gray Level Dependent Matrix (GLDM), adjacent gray level hue difference matrix (NGTDM), and the like. In some embodiments, not only the proteomic features of the lesion region are extracted, but also the proteomic features of the lesion region and the lesion margin, for example, the proteomic features of the tumor and the tumor margin are extracted simultaneously. In still other embodiments, features of an enhanced image obtained by performing wavelet transform and laplacian of gaussian transform on an MRI image may also be extracted.

In some embodiments, the method may further include extracting clinical features of the patient to be tested and providing the extracted clinical features as input to the microvascular invasion prediction model along with features extracted based on a region on the image of the patient to be tested associated with the lesion. Such clinical characteristics may include one or more of the following categories: blood test results related to the patient to be tested, basic disease information related to the patient to be tested, symptom information related to the patient to be tested, and/or personal basic information related to the patient to be tested. That is, at least two broad classes of features are used in training the microvascular invasion prediction model: imaging omics features and clinical features. Therefore, the image state of the current focus area can be considered, personalized information such as the prior basic disease related to the patient and the personal physical condition of the patient is also considered, and the prediction result of the microvascular invasion prediction model can be more personalized and accurate.

Wherein the microvascular invasion prediction model is trained to predict whether a patient to be tested has a liver cancer MVI based on various medical test results of the patient. The training sample set and the test set are data from patients who have undergone liver cancer treatment, and the data comprise MRI detection results, laboratory detection results, pathological report results after operation and the like of the patients in a period of time before the operation. The sample set includes positive and negative samples, i.e., patient data samples labeled as present MVI and patient data samples not present MVI. In some embodiments, the features used to train the microvascular invasion prediction model may include one or more of the following: a cinematographic feature extracted from a region on the MRI image associated with the lesion, a maximum diameter of a region on the MRI image associated with the lesion, data and/or clinical features of a region on the MRI image associated with the lesion.

Existing preoperative MVI prediction is typically the analysis of CT images, but in reality MRI has more modalities that can present more abundant information. However, MRI images are complex, and a professional doctor is required to process and segment a lesion region for a long time, so that the existing imaging omics scheme relying on manual marking of a tumor region has very low processing efficiency for MRI images. By adopting the scheme of the embodiment of the invention, the lesion region on the MRI image can be rapidly and automatically segmented and labeled, and the efficiency and the accuracy of preoperative liver cancer MVI prediction can be effectively improved. In some embodiments, in order to more accurately predict the MVI of liver cancer, the method can simultaneously extract features of MRI images of a plurality of different time phases to predict the MVI of liver cancer. Examination of a patient with MRI may result in various physical characteristic parameters about the patient, such as proton density, spin-lattice relaxation time T1, spin-spin relaxation time T2, diffusion coefficient, magnetization coefficient, chemical shift, and the like. MRI imaging can also be affected by factors such as the length of relaxation time, the speed of flow of the liquid, diffusion of water molecules, and the like. Thus in an embodiment of the invention, there is further included acquiring a series of different phases of MRI images about the patient by adjusting MRI imaging parameters. In addition, enhanced MRI images may also be obtained by injecting MRI contrast agents. The adjustable MRI imaging parameters mainly include settings of relevant parameters such as radio frequency pulses, gradient fields, and signal acquisition times, and their arrangement in time sequence (also referred to as MRI pulse sequence). However, during the MRI examination of a patient to acquire MRI images of different phases, the patient may shift in position, which may cause a deviation in the position of a lesion region between the acquired MRI images of different phases of the same patient. Thus in yet another embodiment, the method further comprises aligning or registering between different phase MRI images acquired of the same patient at the same time. The multi-phase MRI images may be aligned, for example, using medical image registration software such as Simple Elastix software.

The MVI prediction method for liver cancer described above with reference to the steps shown in fig. 1 is further illustrated below by taking MRI image features of 6 different phases as an example.

(1) And acquiring an image and clinical characteristics of the patient to be detected. Wherein 6 different phase MRI images were obtained using a General Electric 3.0T MRI enhanced scan (Discovery 750W, General Electric Company, America), 6 sets of scan parameters were set as follows:

t2: TR 12000ms, TE 90ms, layer thickness 7mm, volume Size 1.76 × 1.32 × 7.00mm, FOV 38.0cm, matrix 256 × 256

TR of DWI is 5000ms, TE is 56.1ms, layer thickness is 7mm, volume Size is 2.97X 7.00mm, FOV is 38.0cm, matrix is 256X 256

T1 (pre-contract) TR 3.7ms, TE 1.1ms, slice thickness 7mm, volume Size 1.98 × 1.56 × 5.00mm, FOV 38.0cm matrix 256 × 256

TR of T1 artery period 2.8msec, TE 1.3msec, layer thickness 7mm, volume Size 1.98 × 1.56 × 5.00mm, FOV 40.0cm, matrix 256 × 256

TR (T1) gate period 2.8msec, TE (TE) 1.3msec, layer thickness 7mm, volume Size 1.98 × 1.56 × 5.00mm, FOV (FOV) 40.0cm, matrix 256 × 256

T1 delay period, F TR 2.8msec, TE 1.3msec, layer thickness 7mm, volume Size 1.98 × 1.56 × 5.00mm, FOV 40.0cm, matrix 256 × 256

Disodium gadoxetate was injected at a dose of 0.2mmol/kg at 1.5 mL/s.

The clinical features of the method are derived from liver function examination, blood routine examination, blood coagulation function and the like which are received within 7 days before the operation of the patient, and may include, for example, alpha-fetoprotein (AFP), carcinoembryonic antigen (CEA), CA125, CA15-3, CA724 and CA19-9, hepatitis B surface antibody (HbsAb), hepatitis B e antibody (HbeAb), hepatitis B e antigen (HbeAg), hepatitis B core antibody (HbcAb), hepatitis C antibody (HCVAb) and the like.

(2) And providing the image of the patient to be detected to a pre-trained deep neural network for lesion segmentation to automatically segment the region related to the lesion on the image. In this embodiment using MRI image features for 6 different phases, the deep neural network used for lesion segmentation uses the Res-Unet model, and the input module of the neural network is also modified accordingly to 6 channels to simultaneously accept and process MRI images from 6 different phases of the same patient.

(3) The image omics features such as first order statistics, gray level co-occurrence matrix (GLCM), Gray Level Run Length Matrix (GLRLM), gray level size area matrix (GLSZM), Gray Level Dependency Matrix (GLDM), adjacent gray level hue difference matrix (NGTDM), etc. are extracted from the automatically segmented region associated with the lesion on the MRI image obtained by the deep neural network used for lesion segmentation. In some embodiments, a region associated with a lesion may be first subjected to a gaussian or wavelet transform, and then corresponding features may be extracted from the transformed image. The inventor finds that the characteristics extracted from the obtained image after Gaussian Laplacian transformation or wavelet transformation are more beneficial to the accuracy of the MVI prediction of the late-stage liver cancer.

(4) And providing the extracted image omics characteristics and clinical characteristics as input to a pre-trained microvascular invasion prediction model for prediction. Fig. 2 shows a schematic diagram of a neural network structure as a microvascular invasion prediction model according to an embodiment of the invention. In this embodiment, a two-layer fully-connected network is used, with two types of inputs: clinical features and imaging omics features. These features are fused in the second layer and to prevent overfitting, a Dropout layer and l2 regularization rule loss function are used. Wherein the length of the element for the imaging omics feature is labeled h1 and the length of the element for the clinical feature is labeled h 2; the concatenated feature length for the second layer is labeled h 3. The network is implemented by a pytorech. The microvascular invasion prediction model predicts whether the patient has a liver cancer MVI prediction by combining the imaging group characteristics extracted from a plurality of different phase MRI images of the same patient with clinical characteristics. Through experimental analysis of the inventors, such a combination of features is more accurate than the predicted effect based on only the prediction of MRI images and only the combination of single-phase MRI images and clinical features.

Fig. 3 is a functional block diagram of a system 300 for predicting the MVI of liver cancer according to an embodiment of the present invention. Although the block diagrams depict components in a functionally separate manner, such depiction is for illustrative purposes only. The components shown in the figures may be arbitrarily combined or separated into separate software, firmware, and/or hardware components. Moreover, regardless of how such components are combined or divided, they may execute on the same host or multiple hosts, where multiple hosts may be connected by one or more networks.

As shown in fig. 3, the system 300 includes an automatic segmentation module 301, a feature extraction module 302, and a prediction module 303. The automatic segmentation module 301 is configured to provide the received image of the patient to be detected as an input to a pre-trained deep neural network for performing lesion segmentation, as described above with reference to step S101, so as to obtain a region on the image related to a lesion. The feature extraction module 302 extracts relevant features based on the region on the image that is relevant to the lesion as described above in connection with fig. 1. The prediction module inputs the extracted features from the feature extraction module to a pre-trained microvascular invasion prediction model to predict whether the patient has hepatocellular carcinoma microvascular invasion as described above in connection with step S102. In some embodiments, the feature extraction module 302 further includes obtaining clinical features associated with the patient from laboratory examination results of the patient. The microvascular invasion prediction model can also be trained to take as input the clinical features of the patient together with features extracted based on the region on the cine image that is relevant to the lesion for the liver cancer MVI prediction.

In another embodiment of the present invention, a computer-readable storage medium is further provided, on which a computer program or executable instructions are stored, and when the computer program or the executable instructions are executed by a processor or other computing units, the technical solutions described in the foregoing embodiments are implemented, which are similar in implementation principle and are not described herein again. In embodiments of the present invention, the computer readable storage medium may be any tangible medium that can store data and that can be read by a computing device. Examples of computer readable storage media include hard disk drives, Network Attached Storage (NAS), read-only memory, random-access memory, CD-ROMs, CD-R, CD-RWs, magnetic tapes, and other optical or non-optical data storage devices. The computer readable storage medium may also include computer readable media distributed over a network coupled computer system so that computer programs or instructions may be stored and executed in a distributed fashion.

Reference in the specification to "various embodiments," "some embodiments," "one embodiment," or "an embodiment," etc., means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Thus, appearances of the phrases "in various embodiments," "in some embodiments," "in one embodiment," or "in an embodiment," or the like, in various places throughout this specification are not necessarily referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. Thus, a particular feature, structure, or characteristic illustrated or described in connection with one embodiment may be combined, in whole or in part, with a feature, structure, or characteristic of one or more other embodiments without limitation, as long as the combination is not logical or operational.

The terms "comprises," "comprising," and "having," and similar referents in this specification, are intended to cover non-exclusive inclusions, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements but may alternatively include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. The word "a" or "an" does not exclude a plurality. Additionally, the various elements of the drawings of the present application are merely schematic illustrations and are not drawn to scale.

Although the present invention has been described by the above embodiments, the present invention is not limited to the embodiments described herein, and various changes and modifications may be made without departing from the scope of the present invention.

Claims (10)

1. A method for prediction of hepatocellular carcinoma microvascular invasion, comprising:

providing the received image of the patient to be detected as input to a pre-trained deep neural network for lesion segmentation to obtain a region on the image, which is related to a lesion;

features are extracted based on a region related to a focus on the image and input to a pre-trained microvascular invasion prediction model to predict whether the hepatocellular carcinoma microvascular invasion exists in the patient.

2. The method of claim 1, wherein the microvascular invasion prediction model employs a deep neural network model.

3. The method of claim 1, wherein the imagery images are MRI images.

4. The method of claim 1, wherein the imagery images are MRI images of different phases.

5. The method of claim 1, wherein the extracted features based on a region on the imagery image associated with a lesion are imagery omics features.

6. The method of claim 5, further comprising extracting clinical features of a patient to be examined and providing the clinical features as input to the microvascular invasion predictive model together with features extracted based on a region on the visual image associated with the lesion.

7. The method of claim 5, wherein the clinical features are from one or more of: blood test results related to the patient to be tested, basic disease information related to the patient to be tested, symptom information related to the patient to be tested, and/or personal basic information related to the patient to be tested.

8. The method of any one of claims 1-7, wherein the deep neural network for performing lesion segmentation is pre-trained based on a sample set comprising imagery images labeled with regions associated with lesions segmented by an expert.

9. A system for prediction of hepatocellular carcinoma microvascular invasion, comprising:

the automatic segmentation module is used for providing the received image of the patient to be detected as input to a pre-trained deep neural network for lesion segmentation so as to obtain a region related to a lesion on the image;

the characteristic extraction module is used for extracting characteristics based on a region related to a focus on the image;

and the prediction module is used for inputting the features extracted by the feature extraction module into a pre-trained microvascular invasion prediction model to predict whether the hepatocellular carcinoma microvascular invasion exists in the patient.

10. A computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, which program, when executed, carries out the method of any one of claims 1-8.

Images