The invention relates to a method for classifying collective cancer cell clusters as an individual prognostic cancer marker.

State-of-the-art devices are known to collect elastographic or viscoelastic data. In addition, procedures are disclosed to derive distinctions of the measured tissue types from the measurement data.

US 8,281 ,663 B2 describes a device for the active coupling of mechanical waves into a patient. Hereto, the data containing elastographic data is collected by using MRT as an imaging method.

The result is the complex shear modulus for at least one space point and is identified with the term viscoelastic properties of this space point.

Preference is given to at least one spatial direction of mechanical waves, whereby data are collected along this spatial direction with highest possible spatial resolution in order to generate quantitative maps of viscoelastic parameters.

DE 10 2010 041 912 B3 describes a procedure for the diagnosis and/or prognosis of cancer diseases, for the diagnosis of the origin of tumour cells, for the therapy optimization of cancer patients and for the screening of active substances for oncology. For this purpose, the mechanical properties of tumour cells and reference cells under mechanical stress are analysed, which lead to linear or non-linear deformation of the respective stressed cell. The suitability of this method for the prognosis of structural and growth properties is only possible on the cellular level, due to the sample quality.

DE 102012 211 735 B4 contains a device and a method for characterizing cells on the basis of the temperature dependence of their mechanical properties and their use. This device, known as an "optical engraver", performs the task of reproducibly characterizing and differentiating cells in such a way as to enable the diagnosis of cancer and the optimization of methods for screening substances as potential medical agents. The method requires the use of cellular samples, in particular samples in aqueous solution. A suitable interpretation for a cell group as it is present in organs or tumours is difficult.

DE 10 2015 204 868 refers, inter alia, to an elastography device with at least one excitation unit for generating mechanical tissue movements in human or animal tissue and an image recording device. Compressed air actuators are used, as they can be operated with the 5 bar compressed air, which is the technical standard for hospitals. The technique described here serves to collect elastographic data, preferably the shear modulus, with the aid of MRT.

The investigation and interpretation of such elastography data sets is discussed in the scientific literature. The article [Magnetic resonance elastography of the human brain: technique, findings and clinical applications; Hiscox et al.] reviews MRE (Magnetic Resonance Elastography) studies in the human brain for the detection of neurodegenerative diseases, neuroinflammatory processes and tumours. Within this study, the different tissue types were examined with regard to the behaviour of their viscoelastic properties. In order to make the measured tissue types comparable, different methods for the normalization of the measured variables were selected. From this, it could be concluded that the different tissue types also differ in their viscoelastic properties.

In the article [High Resolution Imaging of Viscoelastic Properties of Intracranial Tumours by Multi-Frequency Magnetic Resonance Elastography, M. Reiss-Zimmermann et al.] the multi frequency MRE (Magnetic Resonance Elastography) was performed on different patients to characterize brain tumours based on the complex shear modulus G*. Representation of the parameters |G*| (magnitude of the complex shear modulus) and the phase angle of G* as quantitative maps of viscoelastic parameters are used to determine the tumour’s mechanical properties in different patients. In particular, the calculation of the parameters |G*| and the phase angle based on image data is described in detail. The results, description of the brain tumours by their viscoelastic properties, are not generalized to all tissue types.

The article [Tomoelastography of the Prostate Using Multifrequency MR Elastography and Externally Placed Pressurized-Air Drivers; F. Dittmann et al.] describes the use of compressed air actuators for quantification of viscoelastic parameters of in-vivo prostate tissue. These actuators are used to couple mechanical waves into patients. This technique uses preferably multiple surface-placed compressed air drivers and significantly improves the measurement of elastography data in the prostate. The viscoelasticity data are mapped in terms of the shear wave speed as a surrogate of stiffness in order to improve the imaging-based multiparameteric diagnosis of prostate cancer. As the state of research, a statement can be derived from the stiffness map with regard to the structural properties of only one tumour class.

The known methods for the interpretation of quantitative elastography data of tumours have in common a lack of objectified diagnostic scoring systems. Medical imaging modalities today can acquire morphological information with high spatial resolution; however, in the current radiological practice decisions are made based on visual inspection of images. Diseases are rated in qualitative terms like “enlarged,” “small,” or “enhanced.” Such qualitative information often suffers from the limited comparability among readers, modalities, platforms, and scan protocols. Today, radiologists need many years of training to master the interpretation of subtle morphometric variations. Common scoring systems such as PI-RADS (Use of the Prostate Imaging Reporting and Data System (PI-RADS) for Prostate Cancer Detection with Multiparametric Magnetic Resonance Imaging: A Diagnostic Meta-analysis, Esther H J Hamoen et al..) or BI-RADS (Breast, Imaging, Reporting and Data System (Bl RADS), Samuel J. Magny et al.) are focused on single organs and do not include viscoelastic tissue properties. Beyond that, none of the currently used diagnostic information is related to the biophysical properties of the affected tissue. This means that no quantitative biophysical imaging markers can be compared across modalities and tumour entities even though tumour entities in different organs can behave very similar in their invasive potential. Viscoelastic properties are quantitative and intrinsic to the material but are still ignored in the current clinical scores for staging tumours. As a result, the prognosis of aggressiveness and metastatic potential of tumours based on quantitative viscoelastic imaging markers is still a great challenge and unmet need in clinical oncology. Over- and under treatment of patients with solid tumours is a central problem in oncology and decisive scoring does not exist.

Depending on the size of a tumour and on the resolution of the measuring system used, it is possible to identify areas of different viscoelastic properties within the investigated tumours. However, it is currently not possible to make diagnostic decisions and prognostic statements based on these mechanical tissue properties. Today, no scoring system for tumour aggressiveness exists which accounts for quantitative viscoelastic material properties.

The task of the present invention is to overcome the disadvantages of the state of the art and to propose a therapy-supporting procedure for the classification of solid tumour’s viscoelastic data. The main problem with the classification of solid tumours is that currently more than 180 different entities of solid tumours are known in the human or animal body. These tumours originate from a multitude of different tissue types and their prognosis covers the whole spectrum from completely harmless to extremely malignant. Tumours develop from genetic mutations of their originating tissues and can therefore, not be described as a single disease. While most of the current tumour classification come from an outdated pre-genetic era, solely genetic phenotyping is also inadequate, since tumours are known to be highly heterogeneous tissues including many genetically different subpopulations of cells. On the other hand, the spreading of a tumour is a mechanic process. Therefore, the ability of a solid tumour to grow and eventually metastasize should also be reflected/rooted in its mechanical properties and especially in their interplay with the mechanical properties of their surroundings. Based on these mechanical parameters the proposed novel classification method provides a unifying scoring system for solid tumours. It aims beyond the solely detection of potentially malignant tumors (diagnostics) by including predictive statements on how the disease will progress (prognosis).

To establish a viscoelastic scoring system, elastographic data are collected with a measurement uncertainty below 10% and are accessible to a data analysis method carried out by a computer to assist in the identification of altered structural tissue properties and the prognosis of aggressiveness and metastatic potential of a tumour in human or animal tissue.

The prediction of growth properties is divided into two cases:

IV. the prediction is infiltrative cell spreading and displacing growth

V. the results are inconclusive and the sample needs further investigation

The terms "infiltrative cell spreading" and "displacing growth" introduced in case IV are explained in more detail below.

All spreading characteristics are preceded by “displacing growth”, since all tumour entities must first show initial and thus displacing growth in order to develop at all.

In this context, "displacing growth" is caused by the increase in cell mass by uncontrolled cell proliferation, which results in harmful displacement of healthy tissue by excluded volume effects.

In contrast to displacing growth, infiltrative cell spreading is whether a cell moves individually or collectively and whether it eventually is enabled to individually or collectively leave the sample in processes that can be described as local metastasis.

The method carried out by a computer of a therapy-supporting procedure for the classification of cell clusters using measured viscoelasticity parameters provides individualized predictive statements about the growth behaviour or the growth properties of the classified suspect or pathologic tissue. This means that the prognostic statements depend on the probability that the predicted growth behaviour will occur. This makes it advantageous to estimate tumour aggressiveness including infiltrative cell spreading. The method carried out by a computer comprises classification of space-consuming human or animal tissue on the basis of viscoelastic data. To obtain viscoelastic data, elastography is performed and the measured variable recorded for this purpose is the complex shear modulus G* or equivalent measures of the material properties, such as complex wave number k*, shear wave speed c or attenuation rate a. The complex shear modulus G* is a complex number which can be expressed by the following equivalent representations: G* = G’ + iG” and G ^* = |G ^* | exp{i <p}.

The term complex wave number k* is a complex number which can be expressed by the following equivalent representation k* = k’ + ik”. The shear wave speed c is related to the real part of the complex wave number k* and the associated shear wave frequency f. c = 2nf/k’.

The attenuation rate a in the following is the phenomenon that in a system that is in principle capable of oscillation, the amplitude of an oscillation decreases with time or, depending on the circumstances, no oscillation at all can occur. The oscillation is based after once supplied energy on the interrelation of several energy forms. Especially when viscoeleatic properties are measured, kinetic energy and potential energy are mutually exchanged. If energy in the form of heat is added to the total energy balance of the oscillation process, this is the cause of the attenuation. The degree of attenuation is described by the attenuation rate a. The attenuation rate a is related to the imaginary part of the complex wave number k* and the associated shear wave frequency f. a = -f/k”.

Hence, the viscoelastic data contains magnitude shear modulus |G*| or equivalent measures of the material properties, such as complex wave number k*, shear wave speed c or attenuation rate a as measure for the resistance to deformation and phase angle cp.

At least one set of elastographic data is collected from a reference tissue, further referred to as reference data set or “Ref.”, and at least one set of elastographic data is collected from a suspect tissue, further referred to as sample data or “Spl.”.

The reference tissue is preferably the tissue surrounding the suspect tissue, e.g. the tumour, - further also referred to as surrounding tissue. The surrounding tissue is preferably, but not exclusively, the tissue the suspect tissue originates from, e.g. in case the suspect tissue is metastasis - thus a tumour that has spread from its initial site to a spatial distinct site - the reference tissue is not the tissue the suspect tissue originates from the initial site but the tissue surrounding the suspect tissue at the spatially distinct site. In the brain, this spatially distinct site could be a similar region in the contralateral hemisphere. Also in case of thin-layered epithelial tissues, it is often difficult to obtain data of the surrounding tissue of origin as a reference in the patient. As an alternative (in particular for the case that no data of the tissue of origin in the patient is available), the averaged material properties of tissue of origin of other patients serve as a reference.

To classify at least the two cases IV or V, the following steps are performed with the datasets: a) Averaging the angle f from the reference dataset “Ref.” and classification of results into solid “S”, liquid “L” or intermediate “M”, b) Averaging the angle f from the sample dataset “Spl.” and classification of results into solid “S”, liquid “L” or intermediate “M” and

Steps a) and b) refer to the global fluidity index of the sample and the global fluidity index of the reference. The global fluidity index is the averaged value of the fluidity index over the entire data set.

In the following, the term "averaging" is understood as a generalization of the underlying mathematical procedures of numerical mathematics, which have to be executed one after the other in a well-defined way in order to extract the claimed statement from the raw data. These are far more than the known procedures for averaging data points. The understood averaging includes numerical procedures of noise reduction by Gaussian smoothing, phase unwrapping, temporal frequency selection, directional filtering and multicomponent- multifrequency wave inversion as further detailed in figures 2 and 3. This procedure is applied for each single data point and provides maps of c, a, |G*| or cp. The outcome of this whole “averaging” allows the classification of "stiffness" and "fluidity".

In the scientific literature [Creep and relaxation of nonlinear viscoelastic materials: with an introduction to linear viscoelasticity, Findley WN et al], the phase angle f is also named loss tangent as it can be derived by f = arctan(G’VG’). The phase angle f represents a fluidity index of the tissue.

The fluidity index is classified into the three classes "S" for "solid", "L" for "liquid" or "M" for "intermediate". The assignment of the class of the respective fluidity index from a given measured data set point is made by the following relationships:

The quotient of G” and G’ becomes the loss factor. The loss factor becomes 0 for an ideally elastic body and for an ideally viscous body, the loss factor tends towards infinity. Thus all angles f between 0 ° and approximately however smaller than 90 ° are possible. The angle f of 90 ° is here explicitly excluded, since thereby a division by 0 would be demanded and this is not executable.

The following ranges are used to make a classification into "S", "M" or "L". It is true that the closer the intermediate state "M" can be narrowed down, the better a clear prediction of the growth properties can be given.

The fluidity index is based on f so that small values which are clearly below 45 ° are assigned to “S”, larger values which are clearly above 45 ° are assigned to “L” and intermediate cases of approximately 45 ° are assigned to “M”.

The fluidity index "intermediate" or "M" lies between the ranges of the classes "solid" and "liquid". It is preferred in the range from 31° to 59°, especially preferred from 41° to 49° and in particular preferred form 44° to 46°.

The fluidity index "solid" or "S" is preferred in the range from 0° to 30°, especially preferred from 0° to 40° and in particular preferred from 0° to 43°. Similarly, for the fluidity index "liquid" or "L", the range from 60° to approximately but less than 90° is preferred. The range from 50° to approximately less than 90° is particularly preferred and in particular, the range from 47° to approximately less than 90° is preferred.

The degree of tissue fluidity is a key determinant for cell migration.

As physical quantities, the specified angular ranges are naturally subject to the unavoidable statistical fluctuations of the measurement. Especially in the area of the boundaries from "L" to "M" or from "M" to "S", the uncertainty intervals can overlap in such a way that further measurements are necessary to specify the classification. For estimating the level of uncertainty, the Gaussian Failure Mode propagation standard is used in accordance with the recommendations of the competent national metrology institutes.

Steps a) and b) are followed by step c). This is:

Comparison of the classification of results of the reference data set and the classification result of the sample data set in respect to the magnitude of the complex shear modulus |G*| or equivalent measures of the material properties, such as shear wave speed c or attenuation rate a of the reference data set with the complex shear modulus |G*| or equivalent measures of the material properties, such as shear wave speed c or attenuation rate a of the sample data set, and distinction of the classes "sample stiffer than reference" and "sample not stiffer than reference":

The so-called magnitude ratio Q is used for the direct comparison of both data sets. To calculate the magnitude ratio, the quotient of |G*| or its equivalent measures of the material properties, such as complex wave number k*, shear wave speed c or attenuation rate a of the sample and |G*| or its equivalent measures of the material properties, such as complex wave number k*, shear wave speed c or attenuation rate a of the reference is formed. The result is formally a number which is greater than zero and smaller than infinity. Depending on the value of the number, either the sample is stiffer than the reference or the sample is not stiffer than the reference.

The two classes "stiffer than reference" and "not stiffer than reference" are easy to distinguish for values close to extremes. If the magnitude ratio is well above 1 (e.g. 5), the sample is stiffer than the reference. Similarly, if the value of the magnitude ratio is close to zero (e.g. 0.2), the sample is clearly not stiffer than the reference.

In embodiments of the invention, the distinction between the classes "stiffer than reference" and "softer than reference" is carried out by comparing more than one pair of the following material properties, such as complex wave number k*, shear wave speed c or attenuation rate a. The pair of parameters is understood to be the respective measure and measured on the reference tissue and measured on the sample tissue.

This is advantageous, because in this way independently measured parameters can be used for the distinction between the classes "stiffer than reference" and "softer than reference" and thus the degree of uncertainty is reduced.

For values close to 1 , the measurement uncertainty of the incoming magnitudes is taken into account for classification purposes.

Like all physical measured values, |G*|, f or equivalent measures of the material properties, such as complex wave number k*, shear wave speed c or attenuation rate a and consequently the magnitude ratio derived from |G*| have a method-dependent but calculable measurement uncertainty based on systematic and statistical errors. For estimating the level of uncertainty, the Gaussian Failure Mode propagation standard is used in accordance with the recommendations of the competent national metrology institutes. Standard error analysis to calculate the measurement uncertainty is employed.

These calculations for example are ISO / BIPM standard "ISO / 1 EC Guide 98-4: 2012" based. This results in the fact that the resolution of the used measurement technique and apparatus as well as the number of the measured data recorded with it apply as influencing variables of the uncertainty.

Thus a reduction of the uncertainty is possible by improvement of the data, for example by improving the spatial and temporal resolution.

A magnitude ratio close to 1 is then referred to as such if the value 1 is included in the range of the calculated measurement uncertainty. If the measurement uncertainty of the magnitude ratio is such that the magnitude ratio cannot be determined unambiguously (e. g. 1 +/- 0.4), the data shall be recorded in this case with a device of higher resolution and/or with a larger number of measurements. If, despite these measures, the result remains ambiguous, the sample shall be classified as case V and further criteria for prognostic classification may be used.

For a sample stiffer than the reference - represented with respect to the measurement uncertainty by a magnitude ratio larger than 1:

For a sample not stiffer than the reference - represented with respect to the measurement uncertainty by a magnitude ratio smaller than or equal to 1 :

In addition to the magnitude ratio and the global fluidity of the sample, both introduced above, there are further criteria for more precise differentiation. These criteria have a cumulative amplifying effect. This means that the more criteria are fulfilled, the more reliable the prediction and the more likely it is that the prediction will occur in the future. If, after the classification, the sample falls into case V, the growth characteristics of the sample may be further specified by adding an additional criterion.

In a further preferred step, the homogeneity of the data set of the sample is examined.

The "homogeneity", analogous to the magnitude ratio and fluidity index, is coupled to the measurement uncertainty of the measurement data.

Within a data set, a data point is averaged from several measured values.

The uncertainty of a data point within the data set can be reduced by performing several measurements. This can be done for each data point of the data set and thus extended to the entire data set.

For the definition of the term "homogeneity" now local measures of fluidity are considered.

If the local measure of fluidity in the entire data set of the sample is statistically stable, i.e. the recording of further measured values confirms the existing result and/or reduces the uncertainty of the existing results, the data points of the data set can be compared with each other.

If there is no difference between the values of the fluidity (S, L and M) of the individual data points in this comparison, a homogeneous sample is present. This means that in this case all data points have a uniform fluidity measure within the scope of the measurement uncertainty.

In the case of a composite sample, i.e. the data set at hand is only a subset of a data set preferably covering the sample in its entirety, such as biopsies, the individual subquantities should also have a uniform fluidity measure in the case of a homogeneous sample.

In other words, the condition is if the entire sample has the fluidity measure solid, liquid or intermediate within the scope of the measurement uncertainty, it is homogeneous. If at least two areas, each with different fluidity measures in the sense of solid, liquid or intermediate, can be identified taking the measurement accuracy into account, the sample is inhomogeneous.

Thus, the sample is ultimately classified as "homogeneous" or "inhomogeneous" according to the above definition.

The homogeneity classification is assigned to step d) of the classification scheme. d) Examination of the sample dataset with respect to their homogeneity and classification of the results into the states homogeneous “Horn.” and inhomogeneous “Inhom.”.

In step d), the data set of the suspect or pathologic tissue, the sample, is available with preferably a measurement uncertainty of less than 10%.

For a sample stiffer than the reference and a homogeneous distribution of the dominant fluidity property:

For a sample not stiffer than the reference and a homogeneous distribution of the dominant fluidity property:

For a sample stiffer than the reference and an inhomogeneous distribution of the dominant fluidity property:

For a sample not stiffer than the reference and an inhomogeneous distribution of the dominant fluidity property:

If, after further classification, the sample is placed in class V and further investigations are suggested, the growth and spreading characteristics can be further specified by adding a further criterion.

This criterion, which specifies the prognosis in more detail, is the border region of the sample.

There are the following intensifying factors for displacing growth. For example, the increase in mass is particularly simple if the magnitude ratio is greater than 1, taking into account the measurement uncertainty introduced above. A particularly high ratio, with values of the magnitude high above 1, indicates particularly aggressive growth. This applies in particular to fluid samples in fluid or intermediate reference.

In contrast to displacing growth, the estimation of the potential for infiltrative cell spreading of a sample is closely linked to processes at the cellular level. Relevant here is whether a cell moves individually or collectively and whether it can finally individually or collectively leave the sample in processes that can be described as local metastasis.

This property of cells individually or collectively leaving the sample makes it necessary to take a close look at the border region.

The border region is the area between the sample tissue "Spl." and its immediate surroundings. Preferably, but not necessarily, the surrounding tissue is understood as reference tissue "Ref.

The structure of the border region of the sample is determined in relation to the surrounding tissue. For this purpose, a spatial resolution is used in the representation of the viscoelastic measured variables which is preferably available in the range from 2 mm to 5 mm. In order to make a more detailed statement, a resolution from 0.5 mm to 2 mm is particularly preferred, since it can be used advantageously to make statements regarding any existing substructures of the boundary region. Substructures are defined as the distinctive shaping of the border region. The probability that the sample region will show infiltrative cell spreading properties in the future increases with a diffuse morphology of its border region.

A diffuse morphology is understood to mean the loss of any distinct interfaces between the tumour and surrounding tissue beyond the limits of resolution. Thus, the border region appears as blurred region between the tumour mass and surrounding tissue.

In a preferred classification of the border region, a distinction is made between a sharp, defined and clearly delimited border region and diffuse, blurred border region.

This criterion has predictive significance with regard to the infiltrative cell spreading of the investigated sample tissue.

Thus the classification of the border region between the suspect tissue and its surrounding tissue into a sharply focused border region “foe. edge” or diffuse border region “diff. edge” is performed.

The classification with regard to the border region is assigned to step e) of the classification scheme. e) Examination of the border region between the suspect tissue and its surrounding tissue and classification of the results into sharply focused “foe. edge” or diffuse “diff. edge”.

In step e), the data set of suspect tissue, surrounding tissue and in particular the border region between suspect tissue and surrounding tissue shall be available with an uncertainty of less than 10%.

The prediction of growth properties is preferably divided into five cases:

I. the prediction is neither displacing growth nor infiltrative cell spreading,

II. the prediction is only displacing growth,

III. the prediction is only infiltrative cell spreading,

IV. the prediction is infiltrative cell spreading and displacing growth

V. the results are inconclusive and the sample needs further investigation If the results for the global fluidity index, the magnitude ratio, the homogeneity and the boundary region are presented, an accurate prediction of the growth and spreading properties of the sample tissue can be obtained. The growth and spreading properties of space-occupying masses in human or animal tissue are classified into cases I - V by using the results of steps a) to e) listed in the standing table.

In addition to cases I to V, numbers in front of closing brackets are also entered in the table. These serve as a label for later identification in the subclaim.

For a sample stiffer than the reference and a homogeneous distribution of the dominant fluidity property:

For a sample not stiffer than the reference and a homogeneous distribution of the dominant fluidity property:

For a sample stiffer than the reference and an inhomogeneous distribution of the dominant fluidity property:

For a sample not stiffer than the reference and an inhomogeneous distribution of the dominant fluidity property:

According to the definition above the resulting case I is neither displacing nor infiltrative cell spreading, the resulting case II is only displacing growth, the resulting case III is only infiltrative cell spreading, the resulting case IV is displacing growth and infiltrative cell spreading.

Use of a data processing apparatus for carrying out the method for at least the steps a) to c), the partial classification with the steps a) to d) or the full classification with the steps a) to e) is preferred.

While using a data processing apparatus, a computer program product that carries out the method for at least the steps a) to c), the partial classification with the steps a) to d) or the full classification writhe the steps a) to e) is also preferred.

Hence, a data processing system or storage medium which can be read by a data processing apparatus and on which a computer program for executing the method for at least the steps a) to c), the partial classification with the steps a) to d) or the full classification with the steps a) to e) is especially preferred.

The use of a computer program product may be extended to process additionally complementary information, for machine based learning, in particular for the processing of patient-based data, and/or to perform medical imaging techniques for image support. For the realisation of the invention, it is also possible to combine the above-described inventive designs, embodiments and features of the claims in an appropriate way.

The invention is described in the following in several execution examples and represented in the corresponding figure.

Figure 1 shows a diagram. On the abscissa axis the phase angles ( f ) are plotted in radians. The ordinate shows the magnitude of the associated complex shear modulus (|G*|). With the filled triangles data points for the reference tissue of glioblastomas (104) are indicated, corresponding hollow triangles show the respective measured values of the glioblastomas (103). The data of some references of meningioma (102) are also indicated in the diagram by filled squares. Hollow squares indicate measured values of meningioma (101).

Figure 2 shows a flow chart for performing shear waves speed (c) analysis based on raw MRI data comprising magnitude and phase images. Crucial steps include noise reduction by Gaussian smoothing, phase unwrapping, temporal frequency selection, directional filtering, wave inversion and weighted averaging of wave numbers. The outcome of this numerical procedure are c-maps, one for each image slice, depicting the distribution of stiffness in a pixel- resolved fashion for further analysis of tumour aggressiveness as explained herein.

Figure 3 shows a flow chart for analysing the complex shear modulus (G*) based on raw MRI data comprising magnitude and phase images. Crucial steps include noise reduction by Gaussian smoothing, phase unwrapping, temporal frequency selection, low-pass filtering, and multifrequency wave inversion. The outcome of this numerical procedure are maps of the magnitude of G* (|G*|) and phase angle of G* ( f ), one |G*|-map and one f-map for each image slice, depicting the distributions of stiffness and fluidity in a pixel-resolved fashion for further analysis of tumour aggressiveness as explained herein.

For the data collection of the viscoelastic data sets, MRE techniques are preferably used as imaging methods. Here, the image data is acquired by means of an MRI scanner while mechanical waves for electrography are externally introduced into the body in order to quantify viscoelastic parameters. Particularly preferred are mechanical waves in the frequency range from 10 Hz to 100 Hz for clinical applications and from 100 Hz to 8000 Hz for preclinical investigations. An example for the collection of the viscoelastic data of brain tumours by MRE is given in [High Resolution Imaging of Viscoelastic Properties of Intracranial Tumours by Multi-Frequency Magnetic Resonance Elastography, M Reiss-Zimmermann et al.]. Here a clinical 3 Tesla-MRI scanner equipped with a standard 12-channel head coil was used for MRE. Harmonic vibrations of frequency f= 30 Hz, 35 Hz, 40 Hz, 45 Hz, 50 Hz, 55 Hz and 60 Hz were induced into the brain by an acoustic subwoofer connected via carbon fiber rod to a head cradle. Shear waves are captured by a single-shot spin-echo sequence with motion-encoding gradients consecutively applied along all three Cartesian axes. The temporal resolution is 8 dynamics per wave cycle and obtained by shifting a trigger pulse in consecutive scans by increments of 1/8*f. MDEV inversion is applied for parameter recovery based on the unweighted average of the multicomponent and multi-frequency wave fields as explained in detail in Streitberger et al. PLoS One 2012;7(1): e29888 providing maps of |G*| and cp. Figure 1 displays the distribution of these parameters in meningioma and glioblastoma as well as reference tissue demonstrating the separation of both entities based on cp.

An overview of elastography data obtained from various tumour entities is listed in the table below. The data is taken from published and unpublished in vivo studies. For each tissue type (reference or sample) one stiffness related parameter (shear wave speed c or the magnitude of the complex shear modulus |G*|) and the phase angle cp is given with the respective measurement uncertainties.

HCC: Hepatocellular carcinoma, MEN: Meningioma, GBM: Glioblastoma, PCa: prostate cancer PZ: peripheral zone, TZ: transition zone, HC: healthy volunteers, DTT: distal tumor- adjacent tissue, *so far unpublished

Published studies are listed below:

Shahryari - [Tomoelastography Distinguishes Noninvasively between Benign and Malignant Liver Lesions, M. Shahryari et al.]

Streitberger - [How tissue fluidity influences brain tumor progression, K-J Streitberger et al.]

Asbach - [In Vivo Quantification of Water Diffusion, Stiffness, and Tissue Fluidity in Benign Prostatic Hyperplasia and Prostate Cancer, P. Asbach et al ]

A summary of the proposed scoring system applied to different tumour entities of different organs is given in the table below. Unless otherwise indicated the summary is based on the data from the previous table and the classification scheme described above.

HCA: Hepatocellular adenoma, FNH: Focal nodular hyperplasia, CCA: Cholangiocarcinoma, HCC: Hepatocellular carcinoma, MEN: Meningioma, GBM: Glioblastoma, *no data so far

I. the prediction is neither displacing growth nor infiltrative cell spreading,

II. the prediction is only displacing growth,

III. the prediction is only infiltrative cell spreading,

IV. the prediction is infiltrative cell spreading and displacing growth V. the results are inconclusive and the sample needs further investigation

Reference Numbers

101 - Meningioma Data Points

102 - Reference Meningioma Data Points

103 - Glioblastoma Data Points

104 - Reference Glioblastoma Data Points