A Comparative Study of Calcied Epithelioma of Malherbe And Epidermoid Cyst With Ultrasound

Objective: To explore the ultrasonic(US)features of calcied epithelioma Malherbe (CEM) and epidermoid cyst(EC) in differential diagnosis and improve the accuracy of US diagnosis of CEM. Materials and Methods: We retrospectively analyzed 309 subjects, who underwent US examination before operation and pathological diagnosis for CEM or EC post-operation. The subjects were categorized into the training and validation sets according to different inspection times. The US and clinical features of CEM were evaluated through univariate analysis. A multivariate logistic regression model was used to establish the diagnostic model. Results: Statistically signicant variables in univariate analysis(P<0.05)were included in the multivariate logistic regression model. The results demonstrated that the diagnostic model for CEM was statistically signicant (P<0.001). The risk factors included posterior echo attenuation and hypoechoic halos (OR=9.277, 10.254) and the protective factors included age, thickness diameter, and posterior echo enhancement (OR=0.936, 0.302, 0.156). The diagnostic models tested in the training set (AUC=0.974, 95% CI=0.955-0.994) and the validation set(AUC=0.967, 95% CI=0.926-1.000) demonstrated good discriminant ability. Conclusions: The diagnostic accuracy for CEM is higher compared to EC when the nodule is characterized by posterior echo attenuation, hypoechoic halos, smaller thickness, and younger age. US diagnostic model can guide the diagnosis of CEM.


Introduction
Calcifying epithelioma of Malherbe (CEM), also known as pilomatrixoma (PM), is a benign tumor that stems from hair stromal cells, mostly found between the dermis and the subcutaneous fat layer 1,2 . In most cases, CEM grows gradually but may also grow rapidly with local aggression. The incidence of CEM is low, about 1/800-1000; also, the coincidence rate of ultrasonic diagnosis is low 3,4 . An epidermoid cyst (EC) is among the most common tumors that grow in the skin or super cial subcutaneous tissue. Being a cystic lesion, EC is easily misdiagnosed owing to the variation and complexity of the ultrasonic manifestations [5][6][7] . Of note, CEM and EC are both derived from epidermal tissue; however, CEM requires extensive excision to ensure a thorough radical cure, whereas EC requires that the whole lesion be removed 1,8,9 . Therefore, accurate diagnosis of the two diseases is highly imperative for the choice of correct surgical methods. Reports have demonstrated the high accuracy of high-frequency ultrasound in the diagnosis of super cial lesions [1][2][3] . The authors found that the misdiagnosis rate of CEM was high in clinical practice. By analyzing ultrasound diagnostic results of the cases, which the pathological ndings were CEM over the past 5 years, we found that the correct rate of diagnosis was 38.2% with ultrasound, whereas the misdiagnosis rate was 61.8% (EC 52.4% and other lesions 47.6%). Nearly half of CEM cases were misdiagnosed as EC. Despite suggestions from previous studies that intramural calci cation is a crucial ultrasonic feature of CEM, the actual role it plays in the differentiation of CEM from EC is unclear.
Moreover, the contribution of the characteristic changes of EC, such as onionskin sign, ssure sign, apple sign, dotted and striped hyperecho to the differentiation is elusive 8,10−12 . Some scholars found that the length, width, and thickness of EC were larger than CEM 13,14 ; however, the value of these three diameters in the differential diagnosis is yet to be explored. Herein, we aimed to construct a diagnostic model of CEM through a comparison of the clinical and ultrasonic characteristics between CEM and EC and improve the accuracy of ultrasonic diagnosis of CEM.

Results
Basic demographic characteristics of the training set and validation set Clinical characteristics of the training set We explored the clinical characteristics of CEM and EC in the training set and found that the differences in age, onset time, and location were statistically signi cant (P<0.05, Table 2). However, differences in gender, skin color, skin damage, and skin adhesion were not statistically signi cant (P 0.05, Table 2).   Nomogram of ultrasound diagnostic model for CEM A nomogram of the ultrasound diagnostic model for CEM was constructed according to the multivariate analysis of the training set ( Figure 2). Then, we calculated the sum of the variable points through the nomogram to estimate the probability to diagnose CEM (Figure 3-4).

ROC curve of the training and validation sets
Through the calculation of the maximum Youden index and the ROC curve analysis of the training set, we con rmed 0.632 as the cut-off value of the diagnostic model, with 94.3% sensitivity and 91.8% speci city.

Discussion
The present study has revealed the risk factors for the diagnostic model of CEM, including posterior echo Although CEM and EC can occur at any age, this work revealed that CEM was more common in children and adolescents, whereas EC was prevalent in young and middle-aged adults. The median age of onset was 9 years and 43 years. The median time from discovery to the removal of the lesion was 5.5 months and 24 months for CEM and EC, respectively. The course of CEM disease was much less compared to that of EC, which is potentially related to the younger age of onset of CEM and much attention from family members. We also revealed that age is a protective factor, that is, older patients are less likely to be diagnosed with CEM, which is consistent with previous ndings 15,16 .
Furthermore, this study demonstrated that the length, width, thickness of the nodules of CEM were smaller than EC (P<0.001), which concur with previous studies 13,14 . The multivariate logistic regression model implicated thickness as a risk factor for CEM diagnosis. As the thickness reduce, it is more likely to diagnose CEM. However, there are no existing literature reports on whether the thickness is more valuable for the diagnosis of CEM.
Both CEM and EC belong to subcutaneous and super cial masses but their pathological basis is different. Notably, ultrasonic characteristics are closely related to the pathological basis. Pathologically, CEM comprises mainly the basophil cells and shadow cells. The composition ratio of the two varies in different stages. For instance, the shadow cells increase with time, which induces characteristic calci cation and osteogenesis [17][18][19][20][21] . In most cases, the pathological manifestations of EC are substantial lesions with increased keratides and squamous changes in lining keratides 22,23 . Herein, we found that 64.4% CEM was characterized by spotty and/or patchy strong echo, which was consistent with the pathological ndings. Previous studies had reported that calci cation inside the nodule, including, internal strong echo, was a crucial ultrasonic feature of CEM 24,25 . However, we did not include this feature in the diagnostic model because it was less predictive than other ultrasonic features in the AUC analysis. Calci cation does not play a key role in the differential diagnosis between CEM and EC. The nding was different from previous evidence.
In the present study, 55.6% of the nodules exhibited reduced posterior echo. Some experts believe that posterior echo attenuation is potentially associated with calci cation within the nodules or the increase of shadow cells within the non-calci ed nodules 4,19 . EC is referred to as cysts, though their ultrasound appearances are not typical of anechoic nodules and may differ based on the number of keratides 15,26 . We could observe hyperechoic, hypoechoic, isoechoic, mixed echoic, and anechoic nodules, however, 78.7% of EC showed posterior echo enhancement. This could be related to the low attenuation of the acoustic beam passing through the EC. Previously, it was revealed that the degree of enhancement was associated with the water content in the cyst and the distribution of keratinic substances 27 . Our study has implicated posterior echo attenuation as a risk factor for CEM, whereas posterior echo enhancement is a protective factor. These ndings concur with their pathological basis and are among the important characteristics of CEM and EC differentiation. Compared to calci cation, posterior echo attenuation is more important in the diagnosis of CEM.
In addition, we revealed that the hypoechoic halo is an important factor in the diagnosis of CEM, and it is closely related to its pathological basis. .05) and that the blood ow signals and hypoechoic halos were not correlated.
The present study has limitations. First, there were several types of subcutaneous nodules at the super cial tissue yet we only compared CEM to EC but did not differentiate between CEM and other super cial soft tissue masses. Second, this being a single-center study with a limited number of selected patients, there may exist potential bias.

Materials And Methods
The demographic characteristics of the patients were reviewed from the outpatient/inpatient electronic medical records system. Ultrasonic imaging data were obtained from image workstations. Con rms that all experiments were performed in accordance with relevant named guidelines and regulations. The requirement for informed consent from patients was waived.

Recruitment of subjects
There were 309 cases enrolled in this study from December 2015 to December 2020. Inclusion criteria included: Patients whose lesions were surgically removed and the pathological results were CEM or EC; ultrasound was performed before the surgery in our hospital; complete clinical data were available.  (Figure 1). Maximum Youden index values were applied to establish the optimal threshold cut-off values for prognostic factors as continuous variables based on the receiver operating characteristic (ROC). A diagnostic model was constructed as a nomogram. To assess the discriminatory capability of the diagnostic model, we applied the area under the curve (AUC) of the receiver operating characteristic (ROC) curve. P-values< 0.05 was considered statistically signi cant.
Declarations Figure 1 Statistical analysis scheme.

Figure 2
The nomogram for discriminating CEM nodules.

Figure 3
The case of CEM identi ed using the constructed nomogram.  The case of EC identi ed using the constructed nomogram.
(4A) The patient was 9 year-old and had hypoechoic nodules on the subcutaneous tissue of the right abdominal wall with posterior echo enhancement (arrows) and without hypoechoic halos via ultrasound. The thickness of the nodules was 0.7cm.
(4B) Total point which was taken from nomogram of patient was 117 (22 + 96 + 0+ 0 = 118). This point was converted to a predicted probability of 0.1. A similar numerical value was obtained by visual assessment of the nomogram (arrow). The pathologic result was EC.