The differentiation of the character of solid lesions in the breast in the compression sonoelastography. Part II: Diagnostic value of BIRADS-US classification, Tsukuba score and FLR ratio.

UNLABELLED: Sonoelastography is a dynamically developing method of ultrasound examination used to differentiate the character of focal lesions in the breasts. The aim of the Part II of the study is to determine the usefulness of sonoelastography in the differentiation diagnosis of focal breast lesions including the evaluation of the diagnostic value of Tsukuba score and FLR ratio in characterizing solid lesions in the breasts. Furthermore, the paper provides a comparison of classic B-mode imaging and sonoelastography.
MATERIAL AND METHODS: From January to July 2010 in the Ultrasound Department of the Cancer Centre, The Institute of Maria Skłodowska-Curie, 375 breast ultrasound examinations were conducted. The examined group included patients who in B-mode examinations presented indications for pathological verification. They were 80 women aged between 17 and 83 (mean age was 50) with 99 solid focal lesions in the breasts. All patients underwent: the interview, physical examination, B-mode ultrasound examination and elastography of the mammary glands and axillary fossae. The visualized lesions were evaluated according to BIRADS-US classification and Tsukuba score as well as FLR ratio was calculated. In all cases, the histopathological and/or cytological verification of the tested lesions was obtained.
RESULTS: In the group of 80 patients, the examination revealed 39 malignant neoplastic lesions and 60 benign ones. The mean age of women with malignant neoplasms was 55.07 (SD = 10.54), and with benign lesions - 46.9 (SD = 15.47). In order to identify threshold values that distinguish benign lesions from malignant ones, a comparative analysis of statistical models based on BIRADS-US classification and Tsukuba score was conducted and the cut-off value for FLR was assumed. The sensitivity and specificity values for BIRADS-US 4/5 were 76.92% and 96.67% and for Tsukuba 3/4 - 64.1% and 98.33% respectively. The assumed FLR threshold value to differentiate between benign and malignant lesions in the breasts equaled 3.13. The combined application of both classifications (with the threshold value of BIRADS-US 4/Tsukuba 3) improved the total value of sensitivity and specificity of character differentiation of focal lesions (87.2% and 95% respectively).
CONCLUSIONS: In the case of problematic focal lesions, i.e. BIRADS-US 3, the study revealed that obtaining Tsukuba score of 1 and 2 for lesions classified as BIRADS-US 3 confirms their benign character. This allows to avoid the cytological verification.

Introduction

Displacement (elasticity) of tissues constitutes a feature which may undergo changes during the ageing process, in the course of acute or chronic inflammations, in medical conditions concerning connective tissues as well as in neoplastic diseases. Sonoelastography is a method integrated with conventional B-mode examination and allows for the imaging of the elastic properties of tested tissues. Based on the findings, elastography enables to determine (differentiate) their character. Numerous studies conducted on the basis of experimental models and clinical material demonstrated that glandular, adipose and fibrous tissues as well as the majority of benign lesions undergo greater deformation than malignant neoplastic lesions(.

Currently, sonoelastography is a dynamically developing method used in order to differentiate between focal breast lesions(. The available ultrasound apparatuses with sonoelastography option are based on various methods of calculating the value of tissue deformations. One of such methods is the indication of longitudinal movement towards a given direction, which is used in static elastography applied by the author of this article. In this case, tissue compression is performed with the use of the ultrasound transducer. The acquisition of echoes from the tested tissues takes place in real-time prior to and after applying compression. Next, on the basis of tissue movement, the arrangement of tissue deformations is achieved, which is presented in the form of color maps that are combined with a B-mode real-time image. Especially for this technique, a calculation method, called combined autocorrelation method (CAM), was prepared. It enables to obtain images with high processing velocity and accuracy(.

The article presented below is the second part of the publication prepared on the basis of the author's doctoral thesis entitled: The usefulness of sonoelastography in the diagnostics of solid lesions in the breast (written under the supervision of: Prof. Iwona Sudoł-Szopińska, defended on 25 October 2012 in the Cancer Centre, The Institute of Maria Skłodowska-Curie in Warsaw).

The first part of the publication( discussed the features of ultrasound image, which differentiate focal breast lesions. Additionally, it was demonstrated that the highest total value of sensitivity and specificity (173.6%) in differentiating the character of lesions was obtained in the BIRADS-US 4 category.

The aim of the second part of the publication was to evaluate the usefulness of sonoelastography in the differential diagnosis of solid, focal lesions in the breasts including:

the evaluation of the diagnostic value of Tsukuba score and FLR ratio in characterizing solid lesions in the breasts;

a comparative assessment of classic B-mode imaging (BIRADS-US classification) and sonoelastography (Tsukuba score, FLR ratio).

Material and methods

From January to July 2010 in the Ultrasound Department of the Cancer Centre, The Institute of Maria Skłodowska-Curie in Warsaw, 375 ultrasound examinations (US) were conducted due to focal lesions in the breasts found by palpation as well as abnormalities detected in the imaging tests or breast pain.

Out of this group, 80 women were selected for a sonoelastography examination. They were 17–83 years old (mean age was 50) with 99 solid, focal lesions present in the breasts, which were qualified for histopathological and/or cytological verification on the basis of the classic B-mode breast ultrasound.

All patients underwent: the interview, physical examination and imaging scans: US examination and sonoelastography as well as histopathological and/or cytological verification.

US examinations were conducted with the use of EUB-7500 HV apparatus (Hitachi Medical, Japan) with a linear transducer of 7.5–13 MHz. Focal lesions in the breasts were initially evaluated in B-mode images whose features and usefulness in character differentiation were presented in detail in the first part of the publication(. The examination encompassed solid focal lesions in the breasts which were categorized as BIRADS-US 3, 4 and 5 according to the standards of the Polish Ultrasound Society(.

Sonoelastography was conducted by means of HI-RTE tool (Hitachi Real-time Tissue Elastography). The examination consisted in applying pressure with the probe directly above the lesion to the depth of 1–2 mm with the frequency of 1–2 compressions per second and constant control of force by means of the numerical indicator displayed on the screen of the US apparatus. Such a technique is consistent with the method presented by Itoh et al.( All scans were archived.

The obtained elastograms were evaluated on the basis of a five-point Tsukuba classification which presents the degree of tissue deformation on the basis of five patterns(.

Score 1 encompasses hypoechoic lesions (areas) in a B-mode scan in which strain during elastography appears in the entire area; the elastogram presents them in the identical color to the adjacent glandular tissues. This is a typical pattern for benign lesions.

Score 2 encompasses hypoechoic lesions (areas) in a B-mode image whose greater part undergoes deformation during elastography; the elastogram presents them as a mosaic of blue and green. This is also a typical pattern for benign lesions, for example fibroadenomas.

Score 3 encompasses hypoechoic lesions (areas) in a B-mode scan in which strain during elastography is present in their peripheral areas; the elastogram presents a central part in a blue color and a peripheral area in green. Such a pattern more frequently corresponds to benign lesions.

Score 4 encompasses hypoechoic lesions (areas) in a B-mode scan in which no strain appears in the entire area; the elastogram presents them in a blue color. This pattern indicates malignant lesions.

Score 5 occurs when hypoechoic lesions in a B-mode examination and adjacent tissues present no strain in the elastogram; the elastogram shows these areas in a blue color. This pattern indicates invasive carcinoma.

Apart from the evaluation of the character of lesions according to Tsukuba score, the value of FLR ratio was also obtained on the basis of the achieved elastograms(. It constitutes a ratio of the strain of adipose tissue surrounding the lesion, which is indicated by a cursor, and the strain of abnormal lesion (area). It is calculated by US software.

Method of statistical analysis

The obtained results were analyzed statistically by means of statistical software packages (Statistica and IDAMS). The assumed statistical significance level was α≤0.05.

As in the case of BIRADS-US classification, four statistical models based on Tsukuba score were prepared for the purposes of statistical calculations in order to identify threshold values differentiating benign and malignant lesions. The values of sensitivity, specificity, PPV and NPV as well as the sum of sensitivity and specificity were compared for individual models. The threshold value constituted the cut-off value obtained for the Tsukuba model that gave the highest total sum of sensitivity and specificity.

Schematic pattern of a lesion with Tsukuba score 1

Schematic pattern of a lesion with Tsukuba score 2

Schematic pattern of a lesion with Tsukuba score 3

Schematic pattern of a lesion with Tsukuba score 4

Schematic pattern of a lesion with Tsukuba score 5

Measurement of FLR ratio on the elastogram

Results

In the group of 80 patients, 99 focal, solid lesions in the breasts were visualized. Thirty-nine lesions were of malignant character (group I of the analyzed lesions) and 60 lesions were benign (group II of the analyzed lesions).

The average age of patients in group I was 55.07 and it was significantly higher than the average age of the women in group II, which constituted 46.9 (p = 0.0067).

The detailed results of pathomorphological verification of the examined lesions were presented in the first part of the article(.

The evaluation of lesions on elastograms

Evaluation according to Tsukuba score

Tab. 1 presents the arrangement of particular scores in groups I and II on the basis of a five-point Tsukuba scale.

Tab. 1

Strain of examined focal lesions in groups I and II on the basis of Tsukuba score

Tsukuba score		Score 1	Score 2	Score 3	Score 4	Score 5	Total	%
Number of lesions	Group I	0	6	8	17	8	39	39,39
% of the column		0,00	15,79	34,78	94,44	100,00
% of the verse		0,00	15,38	20,51	43,59	20,51
Number of lesions	Group II	12	32	15	1	0	60	60,61
% of the column		100,00	84,21	65,22	5,56	0,00
% of the verse		20,00	53,33	25,00	1,67	0,00
Total		12	38	23	18	8	99
%		12,12	38,38	23,23	18,18	8,08		100,00

Out of 39 neoplastic malignant lesions, 25 (64.1%) received Tsukuba score 4 and 5; eight lesions (20.5%) were classified as score 3 and 6 lesions (15.4%) as score 2. No lesion was marked as score 1.

Fig. 7 presents the arrangement of strain of focal breast lesions in group I in relation to the obtained pathological verification.

Fig. 7

Numerical arrangement of focal breast lesions assigned to individual Tsukuba scores in group I

Out of 60 benign lesions, 44 (73.3%) achieved Tsukuba scores 1 and 2; fifteen lesions (25%) obtained score 3; one lesion (1.7%) demonstrated the features of a lesion that does not undergo deformation and was classified to score 4. No lesion was assigned to score 5.

Fig. 8 presents the arrangement of strain of focal breast lesions in group II in relation to the obtained pathological verification.

Fig. 8

Numerical arrangement of focal breast lesions assigned to individual Tsukuba scores in group II

The statistical analysis revealed that malignant lesions obtain significantly higher values of Tsukuba score. The value of the chi-squared test of independence: χ2 = 52.029, df = 4, p = 0.0000.

Analysis of statistical models that use Tsukuba score

Four statistical models based on Tsukuba score were prepared for the purposes of statistical calculations in order to identify threshold values differentiating benign and malignant lesions (tab. 2).

Tab. 2

The values of sensitivity, specificity, PPV and NPV as well as the sum of sensitivity and specificity for individual Tsukuba score

Statistical models based on Tsukuba score	Tsukuba 1/2		Tsukuba 2/3		Tsukuba 3/4		Tsukuba 4/5
Evaluation	Malignant	Benign	Malignant	Benign	Malignant	Benign	Malignant	Benign
Malignant lesion	39	0	33	6	25	14	8	31
Benign lesion	48	12	16	44	1	59	0	60
Sensitivity	100,00		70,3–92,8	84,62	48,4–77,3	64,10	20,51
Specificity	20,00		61,0–82,9	73,33	91,1–99,7	98,33	100,00
PPV	44,83		53,4–78,8	67,35	81,1–99,3	96,15	100,00
NPV	100,00		76,2–94,4	88,00	70,3–88,2	80,82	65,93
Sensitivity + specificity	120,00			157,95		162,44	120,51

After analyzing the individual statistical indicators, i.e. sensitivity, specificity, PPV, NPV and the sum of sensitivity and specificity, the highest threshold value was obtained for the Tsukuba 3/4 model (64.1%, 98.3%, 96.2%, 80.8% and 162.4% respectively). In the analyzed clinical material, this model proved to be the most accurate in differentiating between benign and malignant lesions based on Tsukuba score.

Evaluation of FLR ratio

The next stage of the ultrasound examination of focal breast lesions constituted the calculation of FLR ratio in elastograms. The calculations for 94 out of 99 lesions were obtained. Due to the fact that the arrangement of FLR values was highly asymmetrical, Box-Cox transformation technique was applied for the purposes of statistical calculations: y = (xλ-1)/λ where λ = 0.037982.

In the statistical analysis, the median and average values of FLR were compared in both groups. In group I, the FLR median for malignant lesions was 5.075 and the average value constituted 6.372 (SD = 2.958). In group II, the FLR median for benign lesions was 1.555 and the average value constituted 1.606 (SD = 2.467) (tab. 3).

Tab. 3

Values and median for FLR ratio in groups I and II

Group	N	Average	SD	Min.	Q25	Median	Q75	Max.
Group I	38	6,372	2,958	0,890	3,520	5,075	7,600	65,780
Group II	56	1,606	2,467	0,010	1,120	1,555	2,800	11,300
Total	94	2,828	3,300	0,010	1,380	2,700	4,900	65,780

The statistical analysis of this parameter revealed that malignancies receive significantly higher values of FLR ratio. The value of the Mann-Whitney U test: Z = 6.349, p = 0.0000.

Based on the obtained FLR values for lesions in groups I and II, the threshold value to differentiate the character of focal lesions was determined. The highest values of sensitivity, specificity, PPV, NPV were obtained for FLR = 3.13 (82.05%, 85%, 78.05% and 87.93% respectively).

Correlation of BIRADS-US classification, Tsukuba score and FLR ratio

BIRADS-US classification versus Tsukuba score

In the subsequent stage of the analysis, the overall assessment of focal breast lesions in accordance with BIRADS-US classification, Tsukuba score and in relation to a pathological verification was compared in groups I and II (tab. 4).

Tab. 4

Relation between BIRADS-US classification and lesion assessment on the basis of Tsukuba score for both groups

Group	BIRADS-usg	Tsukuba 1	Tsukuba 2	Tsukuba 3	Tsukuba 4	Tsukuba 5	Total
Group I	3	0	0	0	0	0	0
	4	0	5	0	3	1	9
	5	0	1	8	14	7	30
Total (in group I)		0	6	8	17	8	39
Group II	3	7	10	1	0	0	18
	4	5	21	13	1	0	40
	5	0	1	1	0	0	2
Total (in group II)		12	32	15	1	0	60
Total		12	38	23	18	8	99

The information included in table 4 reveal that in the group of malignancies, 39 lesions were assigned to BIRADS-US 4 and 5, including 9 lesions classified as BIRADS-US 4 and 30 to BIRADS-US 5. Nine lesions classified as BIRADS-US 4 comprised: 5 lesions with Tsukuba score 2 (an example is presented in fig. 9), 3 lesions with Tsukuba score 4 and one lesion with Tsukuba score 5. From among 30 lesions classified abs BIRADS-US 5, 1 received Tsukuba score 2 (fig. 10), 8 gained Tsukuba score 3, 14 were classified as Tsukuba score 4 and 7 – as Tsukuba score 5.

Fig. 9

Hypoechoic lesion with the dimensions of 6×7×11 mm classified as Tsukuba score 2 (A) and BIRADS-US 4 (B). On histopathology, ductal carcinoma in situ (CDIS) was diagnosed

Fig. 10

Hypoechoic lesion with the dimensions of 12×9×8 mm classified as Tsukuba score 2 (A) and BIRADS-US 5 (B). On histopathology, ductal invasive carcinoma was diagnosed

Benign lesions, on the other hand, were classified to categories: BIRADS-US 3, 4 and 5. Eighteen lesions classified as BIRADS-US 3 comprised: 7 lesions with Tsukuba score 1, 10 lesions with Tsukuba score 2 and 1 lesion with Tsukuba score 3. Fourty lesions classified as BIRADS-US 4 comprised: 5 lesions with Tsukuba score 1, 21 lesions with The results of the statistical analysis demonstrated that the model that strongly indicates a benign lesion has the following parameters: BIRADS-US 3, 4 and Tsukuba 1, 2 and 3. When higher parameters are obtained, the lesions should be treated as malignant. The aforementioned cut-off values were characterized by the sensitivity of 87.18%, specificity of 95%, PPV of 91.89%, NPV of 91.94% and the sum of specificity and sensitivity equaled 182.18% (tab. 5).

Tab. 5

Assessment of models based on BIRADS-US classification and Tsukuba score

	Tsukuba 1	Tsukuba 1	Tsukuba 2	Tsukuba 2	Tsukuba 3	Tsukuba 3	Tsukuba 4	Tsukuba 4	Tsukuba 5	Tsukuba 5
Evaluation	Malignant	Benign	Malignant	Benign	Malignant	Benign	Malignant	Benign	Malignant	Benign
BIRADS-US 3
Malignant lesion	39	0	39	0	39	0	39	0	39	0
Benign lesion	53	7	43	17	42	18	42	18	42	18
Sensitivity	100,00		100,00		100,00		100,00		100,00
Specificity	11,67		28,33		30,00		30,00		30,00
PPV	42,39		47,56		48,15		48,15		48,15
NPV	100,00		100,00		100,00		100,00		100,00
Sensitivity + specificity	111,67		128,33		130,00		130,00		130,00
BIRADS-US 4
Malignant lesion	39	0	34	5	34	5	31	8	30	9
Benign lesion	48	12	17	43	3	57	2	58	2	58
Sensitivity	100,00		87,18		73,3–94,4	87,18	79,49		76,92
Specificity	20,00		71,67		86,3–98,3	95,00	96,67		96,67
PPV	44,83		66,67		78,7–97,2	91,89	93,94		93,75
NPV	100,00		89,58		82,5–96,5	91,94	87,88		86,57
Sensitivity + specificity	120,00		158,85			182,18	176,15		173,59

The statistical analysis of the selected models based on BIRADS-US classification, Tsukuba score and FLR ratio were juxtaposed in order to determine the most useful method in determining the character of examined lesions.

The comparative analysis of Tsukuba 3/4 and FLR model (threshold value of 3.13) for focal lesions revealed that FLR model shows a greater value of the sum of sensitivity and specificity. Moreover, in differentiating between benign and malignant lesions, FLR model was characterized by a statistically higher sensitivity in comparison with Tsukuba model (82.05% versus 64.1%; p = 0.0370). On the other hand, the model Tsukuba 3/4 showed significantly higher specificity as compared to FLR (98.33% versus 85%; p = 0.042).

To conclude, the analysis showed that the model that combines BIRADS-US classification (BIRADS-US 4/5) and Tsukuba score (Tsukuba 3/4) demonstrated significantly higher sensitivity (p = 0.0088) and negative predictive value (p = 0.0321) as compared to Tsukuba score 3/4. Moreover, in relation to FLR indicator, a statistically significant improvement of specificity (p = 0.0340) and positive predictive value (p = 0.0454) was noted. In the case of BIRADS-US (the cutoff value of 4), the authors also obtained higher values of sensitivity and NPV but without statistical significance.

Discussion

Sonoelastography is currently more and more frequently used during breast US examinations particularly in the cases of problematic focal lesions classified to BIRADS-US 3 and 4. Numerous publications concerning the application of this method confirm its usefulness especially in these cases(. The multicenter research and metaanalyses demonstrated that this method improves the specificity of breast US examination (up to 99%) significantly more frequently(. Moreover, some studies also reveal an improvement of sensitivity (up to 97%)(. Conventional B-mode examinations are characterized by a high sensitivity (up to 98%), but lower specificity (up to 96%)(. The combination of both ultrasound techniques could enable a more accurate character differentiation of focal lesions in the breasts and thus, a more accurate selection of patients for lesion biopsy(.

In the author's own research nearly all benign focal lesions were marked with Tsukuba score 1–3. Only one lesion was classified to Tsukuba score 4 (a false positive case). This lesion was diagnosed in an 83-year old patient. On histopathology, it was verified as a fibroadenoma with the features of hyalinization and fibrosis. In a conventional examination, the tumor showed lower echogenicity with hyperechoic foci and had indistinct margin. It fulfilled ultrasound criteria of a complex fibroadenoma of BIRADS-US 4 in which no strain appeared in the elastogram. It is commonly known that complex fibroadenomas are burdened with higher relative risk of developing a breast neoplasm (RR – a ratio of the probability that a given end point will occur in an experimental group, which is subject to the tested measure, versus a control group). This was confirmed in the study Dupont et at.( which revealed 3.1 times greater relative risk of developing breast carcinoma in persons with negative family history and 3.9 times greater risk in the case of positive family history.

When assessing benign lesions in elastograms, Schaefer et al.( found a considerably greater number of false positive cases than the author of this paper. In their study, 1 lesion, which on histopathology occurred to be fibrocystic breast disease, was marked with Tsukuba score 5. Furthermore, 30 benign lesions were marked as Tsukuba 4 (however, the authors do not describe histopathological findings in this group). In the quoted study, as compared to the author's own research, fibroadenomas were more frequently interpreted as lesions that do not undergo deformations (false positive results) in the elastograms (34.5% versus 20%). In the author's own study, however, fibrocystic breast lesions which showed strain on the elastograms occurred more frequently as compared to the study quoted above (56.7% versus 37.2%) (true negative results). In the author's own study, malignant lesions, in the majority of cases (64.1%), obtained Tsukuba score 4 and 5 since they did not show strain. Tsukuba score 3 encompassed 8 such lesions (20.5%) and Tsukuba score 2 – 6 lesions (15.4%). According to the references(, Tsukuba 2 is characteristic of benign changes. In the material of the author, during conventional imaging, 5 lesions obtained BIRADS-US 4 (12.8%) and 1 lesion – BIRADS-US 5 (2.6%). Histopathology revealed that one of these changes was CDIS (presented in fig. 9); in 3 cases, CDIS was accompanied by slight foci of invasive carcinoma and the remaining 2 cases constituted invasive ductal carcinomas (presented in fig. 10). Itoh et al.( analyzed 111 focal lesions in the breast, 52 of which were malignant. Five of the malignancies (9.6%) were marked with Tsukuba score 2. Two of them, after pathological examination, occurred to be CDIS. These results are confirmed in the author's own observations as well as in the study of Krouskop et al.(, which revealed that in situ neoplasms are characterized by lower values of Young's modulus and thus, undergo greater deformations than invasive carcinomas. This also results from different biological features of these neoplasms. They infiltrate only the epithelium and do not cross the basement membrane of the ducts. Therefore, they do not cause the desmoplasia of the stroma, which in the case of invasive carcinomas is responsible for their deformability.

Another interesting fact in the results of Schaefer et al.( is that except for 2 malignant lesions marked as Tsukuba 1, the remaining ones showed no strain and were classified as Tsukuba 4 and 5. These two false negative cases, revealed by histopathological examination, were marked as invasive ductal carcinomas of intermediate malignancy grade. The authors explain that in a conventional US examination, the lesions were assigned to BIRADS-US 4, which constituted an indication for histopathological examination.

In the author's own study, the lesions which were classified as Tsukuba 3 were both benign and malignant (malignant ones constituted 35%). Similar results were obtained by other authors( who reported 13–30% of malignant lesions in this Tsukuba category. On the other hand, Schaefer et al. obtained different results(. In the group with Tsukuba score 3, no malignancies were found. Therefore, it seems that the elasticity pattern of Tsukuba 3 proposed by Itoh et al.( is as controversial and diagnostically problematic as category 4 in BIRADS-US classification. In their studies, Locatelli et al.( and Regini et al.( applied a classification prepared by the Italian Multicentre Study Group. It encompasses solid and cystic lesions and Tsukuba score in the way described above except for the problematic score 3 which was presented in a different manner. It characterized lesions with the majority of the green color and single areas of blue. With such a criterion, the authors managed to obtain a considerably lower percentage of malignant lesions in this group (circa 6%) in comparison with the aforementioned results based on the classification proposed by Itoh et al.( They also obtained high sensitivity (88.5%) and specificity (92.7%) of sonoelastography in determining the character of focal lesions in the breast(. Itoh et al.(, the creators of Tsukuba score and the pioneers in the clinical application of breast elastography, obtained different results for benign lesions. Out of 111 focal breast lesions presented in the published papers, merely 3.8% of benign changes (2/52) received Tsukuba score 3. In the group of malignancies, however, they obtained 22% (13/59). The percentage was then similar to the one obtained in the author's own research. The aforementioned results are significant in terms of further diagnosis. In the case of lesions marked with Tsukuba 3 and higher, the scholars suggest cytological and/or histopathological verification. The author's own results that include such a high percentage of malignancies in Tsukuba 3, confirm that lesions with this score should undergo pathological verification. The choice of the procedure, either a fineneedle aspiration biopsy (FNAB) or a core-needle biopsy (CNB), should be made in relation to BIRADS-US category.

In the author's own study, based on pathological verification of lesions assessed in the elastograms, the types of malignancies marked with individual Tsukuba scores were determined. The most common neoplasms in the group of malignancies were ductal carcinomas (carcinomas in situ, mixed and cribriform occurred more rarely). They showed various elastic properties ranging from lesions in which strain did not appear to those in which it appeared in a considerable area (i.e. Tsukuba 2–5). The creators of Tsukuba score – Itoh et al.( as well as other scholars made similar observations(.

When analyzing benign lesions in relation to Tsukuba classification, no significant differences in their deformation were observed. In the elastograms, the most common fibrocystic dysplasia was classified to Tsukuba scores 1–3. The second most common benign lesion was fibroadenoma which was characterized by diverse level of deformability. Only one fibroadenoma, which has been discussed above, did not undergo deformation probably due to the presence of hyalinization and fibrosis (false positive result). Itoh et al. presented similar observations(. In their study lesions corresponding to fibrocystic disease ANDI (aberrations of normal development and involution) and fibroadenomas were classified as Tsukuba 1–4.

Similarly to BIRADS-US classification, the threshold value for Tsukuba scale was determined taking into consideration the results of pathological examinations. The model with the threshold value Tsukuba 3/4 proved the most accurate in differentiating between benign and malignant lesions: sensitivity – 64.1%, high specificity – 98.33% (and only one false positive result). Zhi et al.( obtained similar results in their study of 559 solid focal lesions which comprised 415 benign and 144 malignant changes (sensitivity 70.1% and specificity 93%). Other authors( also present high values of specificity (nearly 90%) in elastography with the threshold value of Tsukuba 3/4.

On the other hand, Schaefer et al.( reported different outcomes. Using the same threshold value of Tsukuba 3/4, the authors obtained sensitivity of 96.9% and specificity of 76.0%. Such a high sensitivity probably resulted from a high percentage of true positive outcomes (among 62/64 neoplastic lesions in Tsukuba score 4 and 5) and slight number of false negative cases (only 2 lesions marked with Tsukuba 1). A relatively low specificity, on the other hand, was caused by a high number of false positive results (32/129).

US examinations with elastography option not only allow for a qualitative evaluation of focal lesions in the breast with the use of Tsukuba score, but also for a quantitative assessment with the help of FLR ratio. This indicator is calculated by the software of the US apparatus and expresses the ratio of the strain of the adjacent adipose tissue in the breast to the strain of the focal lesion. The adipose tissue is considered a reference area due to its stable deformability(. In the author's own research, average values of FLR for malignant neoplastic lesions were significantly higher than for benign ones and constituted 6.37 and 1.61 respectively. With the threshold value of FLR 3.13, lesions were adequately differentiated with the sensitivity of 82.05% and specificity of 85%. The author's own research confirmed the usefulness of this quantitative parameter in determining the character of focal lesions in the breasts. Other scholars obtained similar results. For instance, Thomas et al.( analyzed the usefulness of FLR indicator on the material of 113 benign and 114 malignant lesions. The FLR mean value for benign lesions was 1.6±1.0 and for the malignant ones 5.1±4.2. Similarly to the author's own material, such a difference was statistically significant. For the assumed threshold value of 2.455, the authors obtained high levels of sensitivity and specificity (90% and 89% respectively). They also drew attention to a high value of the indicator in the case of scars – it was higher than the threshold value and constituted 3.3±1.1. In the author's own study that encompassed 99 focal breast lesions, postoperative scars were not analyzed. Moreover, in the study of Zhi et al.(, the mean value of FLR for benign lesions was 1.83±1.22, and 8.38±7.65 for the malignant ones. The assumed threshold value was 3.05. High values of sensitivity and specificity obtained in this study should be emphasized: sensitivity constituted 92.4% and specificity – 91.1%. The authors explained that they applied a different technique to calculate the ratio. Instead of using the subcutaneous adipose tissue as a reference area, they used glandular tissue located at the same depth as the tested focal lesion.

The author's own research also analyzed the correlations between BIRADS-US categories and Tsukuba scores. According to the assumed guidelines, the category BIRADS-US 3 probably indicates a benign character of lesions, but the risk of malignancy constitutes ≤2% and therefore, it is recommended to perform a US control examination after the lapse of 6 months or a cytological verification(. In the analyzed material, all 18 lesions classified as BIRADS-US 3 were of benign character. Only one lesion among them obtained Tsukuba score 3, the remaining ones were marked with Tsukuba 1 or 2. Therefore, it may be stated that when a focal breast lesion is assigned to BIRADS-US 3 category and to Tsukuba 1 or 2 in the elastograms, it might be diagnosed as benign and thus, it is not necessary to conduct FNAB, but a US control examination should be recommended after the lapse of 12 months. Wojcinski et al.( arrived at the same conclusions in their paper presenting the results of the German multicenter study which encompassed 779 focal lesions in the breast. In the case of a lesion classified to BIRADS-US 3 and Tsukuba 1, 2 and even 3, they recommended a US control examination performed in a short period of time (the authors, however, did not specify this period) and did not render the cytological verification essential. In this group of patients, the highest accuracy of excluding a disease was noted (NPV of 96.9%). In the case of lesions classified to BIRADS-US 3 and Tsukuba 4 and 5, the German authors recommend histological verification due to a high risk of malignancy (45.5%). Such cases, however, were not observed in the author's own research.

Cho et al.( did not detect any malignant changes in the group of 27 lesions classified to BIRADS-US 3 and most of which obtained Tsukuba scores 1 and 2 in elastograms (two lesions were marked with Tsukuba scores 3 and 4).

The analysis of focal lesions classified as BIRADS-US 4 (49 cases in the author's own material), 40 of which were benign and 9 were malignant, demonstrated a diversity in terms of their elasticity. The majority of cases, i.e. 26 lesions, were marked with Tsukuba score 2 and 13 changes with Tsukuba score 3. The remaining ones were assigned to Tsukuba 1, 4 and 5. Assuming that the category BIRADS-US 4 and Tsukuba 2 encompass benign lesions according to the threshold values established in this paper, 5 false negative results were obtained both in BIRADS-US classification and in Tsukuba score. Histopathological verification demonstrated that in 4/5 cases the lesions constituted one carcinoma in situ and 3 carcinomas in situ with a slight component of invasive carcinomas. Such diseases pose diagnostic problems both in conventional US examination and in the assessment of their elastic features in elastograms. Such a considerable percentage of lesions in this category, including preinvasive ones, necessitates their pathological verification. Similar conclusions were drawn by Wojcinski et al.( If upon elastography, changes are classified as Tsukuba 1–3, the authors recommend histopathological verification even though the risk of malignancy is low (24.2%). In their opinion, the recognition of a benign character does not necessitate further diagnosis. In the case of lesions classified as BIRADS-US 4 and Tsukuba 4 or 5, the authors suggest that the second histopathological verification should be performed even if the lesion has been verified as benign. Wojcinski et al. do not recommend elastography in BIRADS-US 1, 2 and 5 since clinical trials have not demonstrated the usefulness of such procedures. The author's own study brings similar conclusions. The category BIRADS-US 5 encompassed 32 lesions, comprising 30 malignancies. Two benign lesions, which showed deformations, obtained Tsukuba score 2 (pathological verification confirmed inflammation – fig. 11) and 3 (verified as papilloma with hyalinization – fig. 12). In the group of malignant lesions classified as BIRADS-US 5, as many as 8 lesions obtained Tsukuba score 3 and 1 lesion was marked with Tsukuba score 2.

Fig. 11

Hypoechoic lesion with the dimensions of 15×14×13 mm classified as Tsukuba score 2 (A) and BIRADS-US 5 (B). On histopathology, inflammatory infiltrations were diagnosed

Fig. 12

Hypoechoic lesion with the dimensions of 13×15×16 mm classified as Tsukuba score 3 (A) and BIRADS-US 5 (B). On histopathology, intraductal papilloma was diagnosed

The comparative analysis of qualitative assessment in the elastograms according to Tsukuba score and quantitative assessment by means of FLR ratio, expressed higher sum of sensitivity and specificity in the quantitative assessment. In the case of qualitative evaluation, however, a very high specificity level was obtained for lesions classified as Tsukuba 3. It constituted 98.33%.

In the available literature concerning the usefulness of elastography in determining the character of solid focal lesions in the breast, the authors mainly focus on seeking the answer to a question of whether or not the method improves the character differentiation of focal breast lesions and facilitates the detectability of neoplasms. Due to the fact that elastography constitutes a supplementation of a classic ultrasound scan, the author attempted to answer the following question: to what extent will the combined assessment of focal lesions, made by means of both techniques, i.e. classic B-mode and elastography, influence the sensitivity and specificity in their character differentiation? The statistical analysis with reference to the pathological verification demonstrated that categories BIRADS-US 3 and 4 as well as the scores Tsukuba 1, 2 and 3 indicate benign lesions with the sensitivity of 87.2% and specificity of 95% as well as PPV and NPV of above 90%. In comparison with Tsukuba scale, a combined assessment significantly improved the sensitivity and NPV of the examination. However, a statistically significant improvement of specificity and PPV was noted in terms of FLR ratio.

To conclude, the conducted own research indicate that breast elastography is a promising technique of ultrasound imaging. The study confirms the validity of the technique as part of a diagnostic algorithm of focal breast lesions. The author demonstrated a high level of specificity of this examination in the diagnosis of the most problematic solid focal lesions in the breast classified to BIRADS-US 3 and 4 in a B-mode examination. The majority of such cases constituted benign lesions in the author's material. Similarly to the studies of other authors(, it was confirmed that in the case of lesions classified to BIRADS-US 3 and Tsukuba 1 and 2, only a control US scan should be performed and that cytological verification is not necessary. On the other hand, in the case of changes classified to BIRADS-US 4, pathological verification is essential. Additionally, all lesions classified as Tsukuba 1 were benign. In classic B-mode imaging they were characterized by low or medium risk of malignancy (BIRADS-US 3 and 4). Therefore, it needs to be stated that this group of lesions, whose strain is identical to the surrounding tissues, are of benign character and do not require pathological verification. Moreover, the study of Yi et al., conducted on the material of 1786 nonpalpable focal breast lesions, confirm the lack of necessity to perform cytological verifications of lesions classified as Tsukuba1/BIRADS-US 4a(.

So far, in Poland, there have been few studies concerning the evaluation of the usefulness of sonoelastography in differential diagnosis of focal breast lesions(. The research presented herein was innovative both in terms of the number of subjects and its methods, which apart from the evaluation of elastograms in Tsukuba scale and the calculation of FLR indicators, also encompassed a comparative analysis of combined BIRADS-US and Tsukuba models. It has been demonstrated that focal lesions in the breast constitute a diversified group in terms of their elastic properties, which enables to exclude their suspicious character in elastography. This particularly concerns the lesions assessed as BIRADS-US 3 in a classic B-mode examination. When both techniques are used, the evaluation of focal breast lesions classified as BIRADS-US 3 and 4 allows for a determination of their character with greater accuracy.

Conclusions

The research carried out by the author revealed that when a B-mode examination is enriched with elastographic assessment, it becomes a valuable tool in the diagnosis of solid focal lesions in the breast.

In the assessment of the diagnostic value of BIRADS-US classification and Tsukuba score, it occurred that the models which differentiate the character of solid focal lesions in the breast with the highest accuracy were BIRADS-US 4/5 (sensitivity of 76.92%, specificity of 96.67%) and Tsukuba 3/4 (sensitivity of 64.1%, specificity of 98.33%). The average values of FLR for malignant neoplastic lesions were significantly higher than for benign ones (p = 0.0000). The threshold value obtained in the examined clinical material, which differentiated between benign and malignant lesions in the breast, equaled 3.13.

The conducted comparative analysis of the classic B-mode imaging and sonoelastography demonstrated a significant improvement in the diagnosis of benign lesions by means of sonoelastography. Furthermore, the combined application of both classifications (with the threshold value BIRADS-US 4/Tsukuba 3) improved the total value of sensitivity and specificity of character differentiation of focal lesions (87.2% and 95% respectively). In the case of problematic focal lesions classified as BIRADS-US 3, the study revealed that when such changes obtain Tsukuba score of 1 and 2, their benign character is confirmed. Thus, it renders cytological verification unnecessary.

Tab. 6

Analysis of statistical indicators in the selected models

	Tsukuba 3		FLR model		BIRADS-US 4		BIRADS-US 4/Tsukuba 3
Evaluation	Malignant	Benign	Malignant	Benign	Malignant	Benign	Malignant	Benign
Malignant	25	14	32	7	30	9	Benign	5
Benign	1	59	9	51	2	58	5	57
Sensitivity	64,10		82,05		76,92		87,18
Specificity	98,33		85,00		96,67		95,00
PPV	96,15		78,05		93,75		91,89
NPV	80,82		87,93		86,57		91,94
Sensitivity + specificity	162,44		167,05		173,59		182,18