METHOD AND DEVICE FOR NON-INVASIVELY CLASSIFYING A TUMOROUS MODIFICATION OF A TISSUE

Abstract

A method for non-invasively classifying a tumorous modification of a tissue according to different stages of the tumorous modification comprises the steps of: a) receiving raw magnetic resonance imaging (MRI) data that has been recorded by applying at least one diffusion weighted imaging (DWI) sequence using three to nine different b-values to a tissue being suspicious to a tumorous modification without application of a contrast agent; b) extracting at least two quantification scheme parameters from the raw MRI data by using at least one quantification scheme, wherein each of the quantification scheme parameters is related to a microstructural property of the tissue; c) applying a weight to each quantification scheme parameter, wherein the weight is dependent on a kind of the tissue and on the quantification scheme, whereby a set of weighted quantification scheme parameters is obtained; d) determining a scoring value by combining the weighted quantification scheme parameters within the set, wherein each of the weighted quantification scheme parameters is used only once for determining the scoring value; and e) classifying the tumorous modification of the tissue into one of at least two classes according to the scoring value. The method and a corresponding classification device are capable of performing non-invasive tissue characterization without contrast agent administration in a highly accurate manner while supplementary information related to conventional imaging properties and clinical information can further increase the high diagnostic accuracy. They are used in their entirety for classifying the tumorous modification of the tissue.

Claims

1. A computer-implemented method for non-invasively classifying a tumorous modification of a tissue into one of at least two classes, wherein each class refers to a different stage of the tumorous modification, wherein the method comprises the steps of: a) receiving raw magnetic resonance imaging (MRI) data that has been recorded by applying at least one diffusion weighted imaging (DWI) sequence using three to nine different b-values to a tissue being suspicious to a tumorous modification without application of a contrast agent; b) extracting at least two quantification scheme parameters from the raw Mill data by using at least one quantification scheme, wherein each of the quantification scheme parameters is related to a microstructural property of the tissue; c) applying a weight to each quantification scheme parameter, wherein the weight is dependent on a kind of the tissue and on the quantification scheme, whereby a set of weighted quantification scheme parameters is obtained; d) determining a scoring value by combining the weighted quantification scheme parameters within the set, wherein each of the weighted quantification scheme parameters is used only once for determining the scoring value; and e) classifying the tumorous modification of the tissue into one of at least two classes according to the scoring value.

2. The method of claim 1, wherein the tissue is a human tissue in vivo and wherein the tumorous modification is selected from the group consisting of breast cancer, cervix cancer, and prostate cancer.

3. The method of claim 1, wherein the stage of the tumorous modification is selected from one of benign or malignant; or benign, clinically insignificant, or clinically significant.

4. The method of claim 1, wherein the at least one diffusion weighted imaging (DWI) sequence uses three to nine different b-values, wherein the b-value is correlated with a magnetic field gradient as used for generating the DWI sequence.

5. The method of claim 4, wherein the b-value is selected from a range of 0 to 4000 s/mm.sup.2, wherein two adjacent b-values are separated from each other by at least 50 s/mm.sup.2.

6. The method of claim 1, wherein the quantification scheme is selected from diffusional kurtosis imaging (DKI), traditional monoexponential model, intravoxel incoherent motions (IVIM), or fractional order calculus (FROC), and wherein the quantification scheme parameter is selected from ADC; AKC; D-IVIM, or f-IVIM.

7. The method of claim 1, wherein the weight to each quantification scheme parameter is obtained by analyzing at least one training data set, wherein the training data set refers to data comprising a confirmed histopathological analysis.

8. The method of claim 1, wherein the scoring value Q is determined by combining the weighted quantification scheme parameters within the set {k.sub.i, p.sub.i; i=1 . . . n, n2} in accordance with Equation (3) as
Q=k.sub.0+.sub.i=1.sup.n2k.sub.i*p.sub.i, (3).

9. The method of claim 1, wherein a set of m weighted additional data {.sub.j, q.sub.j, j=1 . . . m} is, additionally, used for determining the scoring value in accordance with Equation (4) by
Q=k.sub.0+.sub.i=.sup.n2k.sub.i*p.sub.i+.sub.j=1.sup.m.sub.j*q.sub.j, (4).

10. The method of claim 1, wherein the additional data is obtained from at least one of: a non-invasive imaging modality and clinical data.

11. The method of claim 1, wherein the non-invasive imaging modality comprises at least one of: ultrasound, x-ray imaging, computer tomography, positron emission tomography (PET), or conventional MR sequencing.

12. The method of claim 1, wherein the scoring value is compared with at least one score cut-off value, by which the tumorous modification of the tissue is classified into one of the at least two classes.

13. At least one non-transitory machine-readable storage medium comprising a plurality of instructions stored thereon that, in response to execution by at least one processor, causes the at least one processor to perform the method of claim 1.

14. A classification device for non-invasively classifying a tumorous modification of a tissue into one of at least two classes, wherein each class refers to a different stage of the tumorous modification, comprising a receiving unit for receiving raw magnetic resonance imaging (MRI) data being recorded by applying at least one diffusion weighted imaging (DWI) sequence using three to nine different b-values to a tissue being suspicious to a tumorous modification without application of a contrast agent; and an evaluation unit comprising a DWI parameter generator, a DWI parameter engine, and a scoring engine, wherein, the DWI parameter generator is configured for providing at least one quantification scheme for further processing of the raw MRI data, wherein the DWI parameter engine is configured for extracting at least two quantification scheme parameters from the raw MRI data by using the quantification scheme, wherein each of the quantification scheme parameters is related to a microstructural property of the tissue, and wherein the scoring engine is configured for providing a set of weighted quantification scheme parameters, for determining a scoring value by combining the weighted quantification scheme parameters and, by using the scoring value, for classifying the tumorous modification of the tissue into one of at least two classes.

15. The classification device of claim 15, further comprising a adjacent context evaluation engine being adapted for providing additional data, wherein the additional data is obtained from at least one of: a tissue morphology engine and clinical information engine.

16. The classification device of claim 16, wherein the tissue morphology engine is adapted for measuring and post-processing tissue-related data by a non-invasive imaging modality, in particular, by ultrasound, x-ray imaging, computer tomography, positron emission tomography (PET), and/or conventional MM, especially conventional MR sequencing.

17. The classification device of claim 17, wherein the non-invasive imaging modality comprises at least one of: ultrasound, x-ray imaging, computer tomography, positron emission tomography (PET), or conventional MR sequencing.

18. The classification device of claim 16, wherein the clinical information engine is configured for providing clinical data.

19. The classification device of claim 19, wherein the clinical data comprises at least one of: patient age, patient weight, patient origin, history of cancer in patient, history of cancer in family, a risk scoring model, an exposure to at least one risk factors potentially increasing the risk of having a malignancy, an infectious disease, a region of a lesion, at least one blood parameter, or a genetic analysis.

20. The method of claim 10, wherein the clinical data comprises at least one of: patient age, patient weight, patient origin, history of cancer in patient, history of cancer in family, a risk scoring model, an exposure to at least one risk factors potentially increasing the risk of having a malignancy, an infectious disease, a region of a lesion, at least one blood parameter, or a genetic analysis

Description

SHORT DESCRIPTION OF THE FIGURES

[0063] Further optional details and features of the present invention may be derived from the subsequent description of preferred embodiments, preferably in combination with the dependent claims. Therein, the respective features may be realized in an isolated way or in arbitrary combinations. The invention is not restricted to the preferred embodiments. Identical reference numbers in the figures refer to identical elements or to elements having identical or similar functions or to elements corresponding to each other with regard to their functionality.

[0064] FIG. 1 illustrates microstructural tissue correlates in benign and malignant breast lesions for three different solitary quantification schemes as extracted from diffusion weighted imaging (DWI) for comparison purposes (prior art);

[0065] FIG. 2 illustrates microstructural tissue correlates in benign and malignant prostate lesions for three different solitary quantification schemes as extracted from DWI for comparison purposes (prior art);

[0066] FIG. 3 illustrates the correlation between patient age and cancer risk for breast lesions for comparison purposes (prior art);

[0067] FIG. 4 illustrates a classification system according to the present invention which comprises a magnetic resonance imaging device and a classification device;

[0068] FIG. 5 illustrates microstructural tissue correlates in benign and malignant breast lesions as determined according to the present invention;

[0069] FIG. 6 illustrates microstructural tissue correlates in clinically insignificant and clinically significant prostate lesions as determined according to the present invention; and

[0070] FIG. 7 illustrates microstructural tissue correlates in clinically insignificant and clinically significant cervix lesions as determined according to the present invention.

DESCRIPTION OF PREFERRED EMBODIMENTS

[0071] For comparison purposes, FIGS. 1 to 3 demonstrate that currently available solitary quantitative parameters based on diffusion weighted imaging (DWI) sequences for differentiating between benign and malignant tissue in magnetic resonance imaging (MRI) only exhibit of a rather limited diagnostic value.

[0072] In particular, FIG. 1 illustrates a scheme 110 which comprises three different receiver operating curves (ROC) 112, 114, 116 for three different types of solitary quantification schemes, wherein each of the solitary quantification schemes was applied for determining microstructural tissue correlates in benign and malignant breast lesions as extracted from diffusion weighted imaging (DWI). Hereby, each of the curves 112, 114, 116 renders values of a sensitivity as depicted versus values of 1-specify, wherein each of the curves 112, 114, 116 represents one of the following different solitary quantification schemes, wherein [0073] the curve 112 refers to diffusional kurtosis imaging (DKI) [0074] the curves 114, 116 refer to fractional order calculus (FROC).

[0075] As generally used, the term sensitivity refers to a true positive rate, in particular, to a percentage of patients actually having a malignancy and who were correctly identified as having the malignancy. Further, the term specificity refers to a true negative rate, in particular, to a percentage of patients actually having a benign lesion and who were correctly identified as not having a malignancy. In addition, quantitative values for respective areas (AUC) 118 as determined under each of the curves 112, 114, 116 are indicated in the bottom right of FIG. 1 and amount to values of 0.81, 0.82, and 0.68, respectively. Consequently, the correlation as derived from DWI in benign and malignant breast lesions by using solitary quantification schemes was found not to exceed the value of 82%.

[0076] Similarly, FIG. 2 illustrates a further scheme 120 comprising three different ROC curves 122, 124, 126, each representing a particular, different solitary quantification scheme, i.e. diffusional kurtosis imaging (DKI), fractional order calculus (FROC) and intravoxel incoherent motions (IVIM)), wherein each of the solitary quantification schemes was applied for determining microstructural tissue correlates in benign and malignant prostate lesions extracted from DWI. Again, quantitative values for respective areas (AUC) 128 as determined under each of the ROC curves 122, 124, 126 are indicated in the bottom right of FIG. 2 and amount to values of 0.86, 0.76, and 0.81, respectively. Consequently, the correlation as derived from DWI in benign and malignant prostate lesions by using solitary quantification schemes was found not to exceed the value of 86%.

[0077] Further, FIG. 3 illustrates a further ROC curve 130 with a further AUC area 132 under the further ROC curve 130, wherein patent age as a particular example for a clinical indicator may be employed as a potential addendum for the characterization of suspicious breast lesions. Again, a quantitative value for the AUC area 132 as determined under the further ROC curve 130 is indicated in the bottom right of FIG. 3 and amounts to a value of 0.74. Consequently, taking only into account the patient age, the correlation between patient age and cancer risk for breast lesions was found already to assume the value of 74%.

[0078] FIG. 4 illustrates a classification system 150 according to the present invention which at least comprises a magnetic resonance imaging device 152 and a classification device 154.

[0079] As schematically depicted in FIG. 4, the magnetic resonance imaging device 152 is adapted for providing raw magnetic resonance imaging (MRI) data as described in step a). For this purpose, the magnetic resonance imaging device 152 may comprise at least one magnetic resonance unit 156 which is configured for performing a magnetic resonance examination of a tissue being suspicious to a tumorous modification. Further, the magnetic resonance imaging device 152 may comprise a sequencing unit 158 which is configured for providing MRI sequences in form of at least one diffusion weighted imaging (DWI) sequence to be used for the magnetic resonance examination of the tissue, preferably in a sequential manner. Further, the magnetic resonance imaging device 152 may comprise a diffusion weighted imaging unit 160 which is adapted for providing the raw MRI data as recorded by magnetic resonance imaging device 152 by applying the at least one DWI sequence as generated by the sequencing unit 158 to the tissue being suspicious to the tumorous modification to the classification device 154.

[0080] The classification device 154 in the particularly preferred embodiment of FIG. 4 comprises a DWI parameter generator 162, which is configured for providing at least one quantification schemes, preferably one to five different quantification schemes, more preferred two to four different quantification schemes, for further processing of the raw MRI data according to step b), wherein each of the quantification schemes and of quantification scheme parameters, preferably of two to twenty different quantification scheme parameters, more preferred of two to ten different quantification scheme parameters, as extracted from the raw MRI data by using the at least one selected quantification scheme, is related to a microstructural property of the tissue under investigation. For the purpose of receiving the raw MRI data as recorded by applying the at least one DWI sequence to the tissue according to step a), the classification device 154 further comprises a DWI parameter engine 164 which is, in addition, configured for extracting at least two quantification scheme parameters from the raw MRI data as provided by the diffusion weighted imaging unit 160 by using at least one quantification scheme as provided by the DWI parameter generator 162 in accordance with step b).

[0081] Further, the classification device 154 in the particularly preferred embodiment of FIG. 4 comprises a scoring engine 166 which is configured for providing a set of weighted quantification scheme parameters {k.sub.i, p.sub.i; i=1 . . . n, n2} according to step c), which is obtained by applying a weight to each quantification scheme parameter as provided by the DWI parameter engine 164, wherein the weight is dependent on a kind of the tissue and on the quantification scheme as provided by the DWI parameter generator 162. Further, the scoring engine 166 is configured for determining a scoring value by combining the weighted quantification scheme parameters {k.sub.i, p.sub.i; i=1 . . . n, n2} within the set in accordance with step d). Further, the scoring engine 166 is configured for providing an assessment result by classifying the tumorous modification of the tissue into one of at least two classes according to the scoring value according to step e). As mentioned above, the tumorous modification, such as the modification of a breast tissue, may be sorted into one of the two classes benign or malignant. Alternatively, the tumorous modification, such as the modification of a prostate tissue, may be sorted into one of the three classes benign, clinically insignificant, or clinically significant. Further, the assessment result may be outputted to a decision output 168, wherein the decision output 168 may comprise any form that may be suitable for providing and/or presenting the assessment result.

[0082] In the embodiment as schematically depicted in FIG. 4, the DWI parameter generator 162, the DWI parameter engine 164, and the scoring engine 166 may, jointly, be configured as an evaluation unit of the classification device 154 while, in this embodiment, the DWI parameter engine 164 may, additionally, be configured as a receiving unit being designated for receiving the raw MRI data provided by the resonance imaging device 152 to the classification device 154.

[0083] In the preferred embodiment of FIG. 4, the classification device 154 may, further, comprise an adjacent context evaluation engine 170. Herein, the adjacent context evaluation engine 170 may be adapted for providing additional data {q.sub.j, j=1 . . . m} apart from the quantification scheme parameters {p.sub.i; i=1 . . . n} as provided by the DWI parameter engine 164 on the basis of the raw MRI data to the scoring engine 166. In this preferred embodiment, the scoring value Q may, preferably, be determined according to Equation (4) as presented above. In particular, the additional data may be acquired by further modalities which may be used for tissue characterization. For this purpose, the additional data may be provided to the adjacent context evaluation engine 170 by using a tissue morphology engine 172 which may be adapted for measuring and post-processing tissue-related data by a non-invasive imaging modality, in particular, by ultrasound, x-ray imaging, computer tomography, positron emission tomography (PET), and/or conventional MRI, especially conventional MR sequencing.

[0084] Alternatively or in addition, at least one further commonly used magnetic resonance imaging sequence 174 may, additionally, be applied for sequencing within the sequencing unit 158. In this embodiment, the further magnetic resonance imaging sequence 174 may, further, be used in the assessment by the scoring engine 166 to which it may be provided via the adjacent context evaluation engine 170 as optional supplementary information to be included into the analysis of the tissue.

[0085] Alternatively or in addition, the additional data may comprise clinical data, in particular patient age, patient weight, patient origin, history of cancer in patient and/or family, a risk scoring model (such as a GAIL-model for breast cancer), exposure to at least one risk factor potentially increasing the risk of having a malignancy (such as smoking, irradiation, exposure to chemical or biological substances), an infectious disease, a region of a lesion, at least one blood parameter, or a genetic analysis, which may be provided to the adjacent context evaluation engine 170 by using a clinical information engine 176 which may, particularly, be adapted for this purpose.

[0086] By using the adjacent context evaluation engine 170, the additional data can be provided to the scoring engine 166, where the additional data can be processed by combining them with the weighted quantification scheme parameters in order to determine the scoring value to be outputted to the decision output 168 as described above.

[0087] FIG. 5 illustrates a ROC curve 190 and an AUC area 192 under the ROC curve 190, wherein the method and the device according the present invention was applied for a tissue suspicious of breast lesions. Again, a quantitative value for the AUC area 192 as determined under the further ROC curve 190 is indicated in the bottom right of FIG. 5 and amounts to a value of 0.93. Consequently, by using the method and the device according the present invention, the AUC area 192 assumes the value of 93% which is significantly higher compared to known procedures, such as illustrated in FIG. 1.

[0088] Similarly, FIG. 6 illustrates a ROC curve 194 and an AUC area 196 under the ROC curve 194, wherein the method and the device according the present invention was applied for a tissue suspicious of prostate lesions. Again, a quantitative value for the AUC area 196 as determined under the further ROC curve 194 is indicated in the bottom right of FIG. 6 and amounts to a value of 0.93. Consequently, by using the method and the device according the present invention, the AUC area 196 assumes the value of 93% which is, again, significantly higher compared to known procedures, such as illustrated in FIG. 2.

[0089] In addition, FIG. 7 illustrates a ROC curve 198 and an AUC area 200 under the ROC curve 198, wherein the method and the device according the present invention was applied for a tissue suspicious of cervix lesions. Again, a quantitative value for the AUC area 200 as determined under the further ROC curve 198 is indicated in the bottom right of FIG. 7 and amounts to a value of 0.95. Consequently, by using the method and the device according the present invention, the AUC area 200 assumes the value of 95% which is, also, significantly higher compared to known procedures, such as illustrated in FIG. 2.

[0090] Experimental Results

[0091] Using the method and the device according to the present invention, a clear differentiation between malignant and benign tissue with regard to possible breast, cervix, and prostate cancer could be provided for approximately 400 patients.

[0092] Tablet provides an example of six patients with suspicious breast lesions which were classified by using the method according to the present invention, wherein each of 8 quantification scheme parameters p.sub.i, i=1 . . . 8 were weighted with individual weights k.sub.i, i=1 . . . 8, from which a scoring value Q was determined according to Equation (3) as

Q=k.sub.0+.sub.i=1.sup.8k.sub.i.Math.p.sub.i, (3)

[0093] i.e. by summing up the 8 selected weighted quantification scheme parameters k.sub.i.Math.p.sub.i. The scoring value Q was used to classify the tissue modification into one of the two classes benign and malignant by applying a score cut-off value of 0.

TABLE-US-00001 TABLE 1 BI- Pat. p.sub.1 p.sub.2 p.sub.3 p.sub.4 p.sub.5 p.sub.6 p.sub.7 p.sub.8 Q Hist. RADS 1 0.90 2.04 1.08 0.99 0.60 7.45 0.66 0.16 9.86 malignant 4 2 0.67 1.85 0.94 0.70 0.60 7.44 0.58 0.10 8.91 malignant 4 3 0.42 0.93 0.80 0.52 0.74 6.74 0.42 0.11 6.56 malignant 4 4 1.80 0.60 2.56 1.83 0.79 9.90 1.68 0.09 5.76 benign 4 5 2.24 0.68 2.97 2.91 0.82 14.80 2.02 0.20 5.34 benign 4 6 2.42 0.45 2.66 2.44 0.86 10.75 2.16 0.21 8.21 benign 4

[0094] In a similar manner, Table 2 provides an example of six patients with suspicious prostate lesions which were, again, classified by using the method according to the present invention, wherein each of 9 quantification scheme parameters p.sub.i, i=1 . . . 9 were weighted with individual weights k.sub.i, i=1 . . . 9, from which a scoring value Q was determined according to Equation (3) as

Q=k.sub.0+.sub.i=1.sup.9k.sub.i.Math.p.sub.i, (3)

[0095] i.e. by summing up the 9 selected weighted quantification scheme parameters k.sub.i.Math.p.sub.i. The scoring value Q was, again, used to classify the tissue modification into one of the two classes benign and malignant by applying a score cut-off value of 0.

TABLE-US-00002 TABLE 2 PI- Pat. p.sub.1 p.sub.2 p.sub.3 p.sub.4 p.sub.5 p.sub.6 p.sub.7 p.sub.8 p.sub.9 Q Hist. RADS 1 0.65 1.19 1.00 0.69 0.80 6.92 0.52 0.10 5.38 4.36 malign. 4 2 0.61 1.14 1.27 0.72 0.66 7.37 0.46 0.17 5.35 4.68 malign. 4 3 0.69 1.17 1.32 0.69 0.75 6.46 0.52 0.14 5.40 4.52 malign. 4 4 2.10 0.70 2.44 1.75 0.67 7.63 1.36 0.36 6.88 9.31 benign 4 5 2.07 0.81 2.29 1.91 0.55 8.55 1.21 0.39 6.82 9.22 benign 4 6 2.19 0.64 2.60 2.02 0.68 8.60 1.54 0.31 7.00 8.88 benign 4

[0096] In the following, three further examples are presented for providing further insight into the method according to the present invention.

[0097] Example 1 refers to a patient for prostate cancer by check-up by MRI who was examined after a digital rectal examination with an unclear palpable finding. The patient underwent a prostate MRI examination in a 3T MRI device including a routine protocol as suggested by the PI-RADS ACR protocol. The protocol consisted of morphological sequences without contrast enhancement (T2-weighted), contrast enhanced sequences (T1-weighted) and a diffusion weighted imaging (DWI) sequence with multiple b-values according to the present invention.

[0098] After MRI examination the prostate was evaluated for a suspicious lesion as described in the PI-RADS V2 guidelines using the T2-weighted and DWI weighted imaging sequences. A lesion was detected and classified as a lesion coded with a PI-RADS class 3 (intermediate). The lesion was marked using a segmentation. The raw image data of the lesion was then processed using three different quantification schemes, i.e. Kurtosis, IVIM and the traditional monoexponential model, and three different quantification scheme parameters, i.e. K.sub.Kurtosis, f.sub.IVIM, and D.sub.ADC, were extracted as described elsewhere in this document. The quantification scheme parameters were, subsequently, processed to finally receive a weighted scoring of each quantification scheme parameter resulting in a classifier to predict clinically significant or clinically insignificant data. Herein, the weighting for each parameter was obtained of a raw data training set of approximately 200 patients with histopathologically confirmed lesions.

[0099] In particular, 3 quantification scheme parameters p.sub.1=0.75, p.sub.2=1.29, and p.sub.3=0.15 were combined with 3 corresponding individual weights k.sub.i=15.10, k.sub.2=5.37, k.sub.3=4.61 and k.sub.0=25.09. Based on these data, the scoring value Q could be determined according to Equation (4) as Q=25.09+15.10*0.75+5.37*1.29+(4.61)*0.15=7.53. Applying a score cut-off value of 0, the suspicious prostate lesion was classified into the class clinically significant, which corresponded with the result provided by the histopathology.

[0100] Example 2 refers to a further patient for prostate cancer checkup by MRI who was examined after an elevated blood level of PSA had been found. The patient underwent a prostate MRI examination in the 3T MRI device including a routine protocol as suggested by the PI-RADS ACR protocol. The protocol consisted of morphological sequences without contrast enhancement (T2-weighted), contrast enhanced sequences (T1-weighted) and a diffusion weighted imaging (DWI) sequence with multiple b-values according to the present invention.

[0101] After MRI examination the prostate was evaluated for a suspicious lesion as described in the PI-RADS V2 guidelines using the T2-weighted and DWI weighted imaging sequences. A lesion was detected and classified as a lesion coded with a PI-RADS class 3 (intermediate). The lesion was marked using segmentation. The raw image data of the lesion was then processed using two different quantification schemes, i.e. Kurtosis, IVIM and the traditional monoexponential model, and three different quantification scheme parameters, i.e. K.sub.Kurtosis, f.sub.IVIM, and D.sub.ADC, were extracted as described in the patent application. The quantification scheme parameters were, subsequently, processed to finally receive a weighted scoring of each quantification scheme parameter resulting in a classifier to predict clinically significant or clinically insignificant data. Herein the weighting for each parameter was obtained of a raw data training set of approximately 200 patients with histopathologically confirmed lesions.

[0102] In particular, the 3 quantification scheme parameters p.sub.1=2.06, p.sub.2=0.62, and p.sub.3=0.24 were, again, combined with the same 3 corresponding individual weights k.sub.1=15.10, k.sub.2=5.37, k.sub.3=4.61 and k.sub.0=25.09 as in Example 1 above. Based on these data, the scoring value Q could be determined according to Equation (4) as Q=25.09+15.10*2.06+5.37*0.62+(4.61)*0.24=8.24. Applying a score cut-off value of 0, the suspicious prostate lesion was classified into the class clinically insignificant, which corresponded with the result provided by the histopathology.

[0103] Example 3 refers to patient with a suspicious breast lesion who was examined due to a suspicious finding on X-ray mammography screening. The patient underwent a regular breast MRI scan using a 1.5 T MRI device. Herein, the acquired image sequences consisted of a regular breast imaging protocol with unenhanced T1-weighted and T2-weighted sequences, dynamic contrast enhanced T1-weighted imaging sequences, and a diffusion weighted imaging (DWI) sequence with multiple b-values according to the present invention.

[0104] A lesion was detected that provided an unclear rating according to the BI-RADS classification scheme (BI-RADS 3). Applying the invention procedure out of the detected and segmented lesion eight different quantification scheme parameters were extracted using different quantification schemes, i.e. the traditional monoexponential model, Kurtosis, IVIM, and FROC. With further allowing the scoring engine to not only use a weighted scoring of the quantification scheme parameters but also clinical information in terms of the patient age it was possible to classify this lesion as malignant which, being in contrast to the suggestion of the conventional BI-RADS criteria, was confirmed by histopathology. The weighting for the scoring engine was based on training of approximately 200 cases.

[0105] In particular, 8 quantification scheme parameters p.sub.1=0.57, p.sub.2=1.22, p.sub.3=2.27, p.sub.4=0.69, p.sub.5=0.52, p.sub.6=11.40, p.sub.7=0.62, and p.sub.a=0.38 were combined with 8 corresponding individual weights k.sub.1=5.12, k.sub.2=5.94, and k.sub.3=0.8, k.sub.4=3.38, k.sub.5=9.95, k.sub.6=0.23, k.sub.7=4.55, k.sub.8=10.69 and k.sub.0=3.03. Based on these data, the scoring value Q could be determined according to Equation (4) as Q=3.03+5.12*0.57+(5.94)*1.22+0.80*2.27+(3.38)*0.69+(9.95)*0.52+0.23*11.40+4.55*0.62+(10.69)*0.38=5.61. Applying a score cut-off value of 0, the suspicious prostate breast was classified into the class malignant, which corresponded with the result provided by the histopathology.

[0106] As a result, the method according to the present invention allows determining different scoring values in malignant lesions compared to scoring values in benign lesions with a significant difference. The separation between malignant tissue and benign tissue for both prostate and breast lesions was found to be comparable or superior to conventional classification schemes of imaging using BI-RADS (Breast Imaging Reading and Documentation System) and PI-RADS (Prostate Imaging Reading and Documentation System) with conventional imaging protocols that include a combination of T1-weighted imaging before and after intravenous contrast administration, conventional DWI with ADC maps, and T2-weighted imaging.

[0107] Although limited to the description of using the method and device according to the present invention in breast and prostate tissue imaging, both the method and device are expected to be of added value for further kinds of tumorous modification in various tissues.

LIST OF CITED REFERENCES

[0108] [1] Freitag M T, Bickelhaupt S, Ziener C, Meier-Hein K, Radtke J P, Mosebach J, et al. Selected clinically established and scientific techniques of diffusion-weighted MRI: In the context of imaging in oncology. Der Radiologe 2016; 56(2):137-47. [0109] [2] Radbruch A, Weberling L D, Kieslich P J, Eidel O, Burth S, Kickingereder P, et al. Gadolinium retention in the dentate nucleus and globus pallidus is dependent on the class of contrast agent. Radiology. 2015; 275(3):783-91. [0110] [3] Errante Y, Cirimele V, Malllo C A, Di Lazzaro V, Zobel B B, Quattrocchi C C. Progressive increase of T1 signal intensity of the dentate nucleus on unenhanced magnetic resonance images is associated with cumulative doses of intravenously administered gadodiamide in patients with normal renal function, suggesting dechelation. Invest Radiol. 2014; 49(10):685-90. [0111] [4] Kanda T, Ishii K, Kawaguchi H, Kitajima K, Takenaka D. High signal intensity in the dentate nucleus and globus pallidus on unenhanced T1-weighted MR images: relationship with increasing cumulative dose of a gadolinium-based contrast material. Radiology. 2014; 270(3):834-41. [0112] [5] McDonald R J, McDonald J S, Kallmes D F, Jentoft M E, Murray D L, Thielen K R, et al. Intracranial Gadolinium Deposition after Contrast-enhanced MR Imaging. Radiology. 2015; 275(3):772-82.

[0113] [6] Hirano M et al. Diffusion-Weighted Imaging of Breast Masses: Comparison of Diagnostic Performance Using Various Apparent Diffusion Coefficient Parameters, Am. J. of Roentgenology 2012; 198(3):717-722. [0114] [7] Junker Det al. Evaluation of the PI-RADS Scoring System for Classifying mpMRI Findings in Men with Suspicion of Prostate Cancer, BioMed Research International 2013:1-9 [0115] [8] Othman E, Wang J, Sprague B L, Rounds T, Ji Y, Herschorn S D, et al. Comparison of false positive rates for screening breast magnetic resonance imaging (MRI) in high risk women performed on stacked versus alternating schedules. SpringerPlus. 2015; 4:77. [0116] [9] Baltzer P A T, Benndorf M, Dietzel M, Gajda M, Runnebaum I B, Kaiser W A. False-Positive Findings at Contrast-Enhanced Breast MRI: A BI-RADS Descriptor Study. American Journal of Roentgenology. 2010; 194(6):1658-63. [0117] [10] Quon J S, Moosavi B, Khanna M, Flood T A, Urn C S, Schieda N. False positive and false negative diagnoses of prostate cancer at multi-parametric prostate MRI in active surveillance. Insights into Imaging. 2015; 6(4):449-63. [0118] [11] Bammer R. Basic principles of diffusion-weighted imaging. Eur J Radiol. 2003; 45(3): 169-84. [0119] [12] Koh D M, Collind D J, Orton M R. Intravoxel incoherent motion in body diffusion-weighted MRI: reality and challenges. AJR Am J Roentgenol. 2011; 196(6):1351-61. [0120] [13] Yablonskiy D A, Bretthorst G L, Ackerman J J. Statistical model for diffusion attenuated MR signal. Magnetic resonance in medicine. 2003; 50(4):664-9. [0121] [14] Jensen J H, Helpern J A, Ramani A, Lu H, Kaczynski K. Diffusional kurtosis imaging: the quantification of non-gaussian water diffusion by means of magnetic resonance imaging. Magnetic resonance in medicine. 2005; 53(6)1432-40. [0122] [15] lima M, Yano K, Kataoka M, Umehana M, Murata K, Kanao S, et al. Quantitative non-Gaussian diffusion and intravoxel incoherent motion magnetic resonance imaging: differentiation of malignant and benign breast lesions. Invest Radiol. 2015; 50(4):205-11. [0123] [16] Sui Y, Wang H, Liu G, Damen F W, Wanamaker C, Li Y, et al. Differentiation of Low- and High-Grade Pediatric Brain Tumors with High b-Value Diffusion-weighted MR Imaging and a Fractional Order Calculus Model. Radiology. 2015; 277(2):489-96. [0124] [17] Panagiotaki E, Walker-Samuel S, Siow B, Johnson S P, Rajkumar V, Pedley R B, Lythgoe M F, and Alexander D C; Noninvasive Quantification of Solid Tumor Microstructure Using VERDICT MRI; Cancer Res; 74(7) Apr. 1, 2014, 1902-12. [0125] [18] Barbieri, S., Bronnimann, M., Boxier, S. et al. Differentiation of prostate cancer lesions with high and with low Gleason score by diffusion-weighted MRI. Eur Radiol 2016, pp 1-9. [0126] [19] Sui, Y., H. Wang, et al. (2015). Differentiation of Low- and High-Grade Pediatric Brain Tumors with High b-Value Diffusion-weighted MR Imaging and a Fractional Order Calculus Model. Radiology 277(2):489-496 [0127] [20] Montoya, P. I. et al. Diffusion weighted imaging of prostate cancer: Prediction of cancer using texture features from parametric maps of the monoexponential and kurtosis functions, 2016 Sixth International Conference on Image Processing Theory, Tools and Applications (IPTA), IEEE, 2016: 1-6 [0128] [21] Eberl M M, Fox C H, Edge S B, Carter C A, Mahoney M C. BI-RADS classification for management of abnormal mammograms. J Am Board Fam Med. 2006; 19(2):161-64. [0129] [22] Weinreb J C, Barentsz J O, Choyke P L, Cornud F, Haider M A, Macura K J, Margolis D, Schnall M D, Shtern F, Tempany C M, Thoeny H C, Verma S. Weinreb J C. PI-RADS Prostate Imaging-Reporting and Data System: 2015,Version 2. Eur Urol 2016; 69(1):16-40

LIST OF REFERENCE NUMBERS

[0130] 110 scheme

[0131] 112 ROC curve

[0132] 114 ROC curve

[0133] 116 ROC curve

[0134] 118 AUC area

[0135] 120 scheme

[0136] 122 ROC curve

[0137] 124 ROC curve

[0138] 126 ROC curve

[0139] 128 AUC area

[0140] 130 ROC curve

[0141] 132 AUC area

[0142] 150 classification system

[0143] 152 magnetic resonance imaging device

[0144] 154 classification device

[0145] 156 magnetic resonance unit

[0146] 158 sequencing unit

[0147] 160 diffusion weighted imaging unit

[0148] 162 DWI parameter generator

[0149] 164 DWI parameter engine

[0150] 166 scoring engine

[0151] 168 decision output

[0152] 170 adjacent context evaluation engine

[0153] 172 tissue morphology engine

[0154] 174 magnetic resonance imaging sequence

[0155] 176 clinical information engine

[0156] 190 ROC curve

[0157] 192 AUC area

[0158] 194 ROC curve

[0159] 196 AUC area

[0160] 198 ROC curve

[0161] 200 AUC area