Introduction
Among the three major salivary glands in the human body, the parotid gland, as the largest salivary gland, accounts for about 3% of head and neck tumors, most of which are benign tumors about 80%, and pleomorphic adenoma (PA) and adenolymphoma (AL) are the two most common benign tumors, accounting for 80% and 10% [
1‐
3], respectively. Although both PA and AL were benign tumors, there were large differences in their biological behavior, surgical methods, and postoperative outcome. About 2 -25% of PA is malignant, If the tumor is removed only along the capsule, about 15% of cases will relapse, so PA only needs to retain the facial nerve during the operation, remove the tumor and the involved parotid gland together. However, AL rarely has malignant changes, and most of the capsule is complete, the tumor only needs to be removed along the capsule. Unfortunately both PA and AL exhibit similar clinical characteristics, making it difficult to distinguish them. Therefore, accurate preoperative differential diagnosis is very important for the perioperative preparation, condition communication and the selection of surgical methods [
4‐
7].
At present, there are many methods for diagnosing and differentiating salivary gland tumors, for example, by radiological tests such as computed tomography (CT), ultrasonography (US), and magnetic resonance imaging (MR). Among them, the US examination for salivary gland tumors has many advantages, including low cost, ease of operation, and high safety, but for tumors located in the deep lobes, the test results are not very ideal, and the quality of the test results is closely related to the personal level of the physician and his medical experience, and different physicians may obtain different results [
8]. MR is indicated for non-invasive determination of different histological subtypes of parotid gland tumors. However, it is not indicated in patients with metal prostheses and pacemakers, which may affect the diagnosis, be expensive, and take longer [
9]. Currently, CT is not the preferred test for the evaluation of parotid gland tumors, however, CT has high diagnostic value for assessing the extent of tumors, especially those located in deep regions, and can help clinicians evaluate tumors more accurately, and CT can also detect lymphadenopathy to help identify benign or malignant tumors for appropriate treatment [
10]. Contrast contrast-enhanced CT can clearly observe the relationship between the lesion and surrounding tissues, organs, and blood vessels, and although there are reports that contrast may cause adverse effects, the overall prevalence of adverse reactions is about 0.7%, of which more than 80% are mild [
11]. In the past, there were few studies on enhanced CT imaging, and the results were not very satisfactory, and no diagnostic system has been formed, and CT imaging focuses on the morphological characteristics of tumor margin, size, location, and density [
12,
13]. In addition, physicians evaluate CT results based on their own skills and clinical experience, so that their results are more biased towards supervisor’s judgment than objective indicators.
The histogram analysis method provides more and more detailed quantitative information on image heterogeneity by quantifying the grayscale information of lesion images, calculating the characteristic parameters of ROI in the image, evaluating the relationship and distribution of ROI gray intensity, and providing more detailed quantitative information on image heterogeneity, which has the advantages of simple operation process, objective data quantification and strong reproducibility of results [
14,
15]. So far, many research results have shown that histogram analysis has high medical value for diagnosing benign and malignant tumors, evaluating treatment effects, and predicting prognosis [
16‐
18]. In this study, the imaging features and the histogram parameters of PA and AL were analyzed to explore the feasibility and value of enhanced CT lesion imaging features and histogram analysis in identifying PA and AL.
Discussion
The most prevalent benign parotid gland tumor, PA, is more prevalent in females than males, with a more painless growth rate and a prolonged illness. 15% of cases had a postoperative risk of recurrence, with uniform or uneven CT plain density, and well-demarcated [
23‐
25]. AL, also known as Voxinoma, is more common in middle-aged and elderly men with smoking. The risk of malignant change is less than 1% and rarely postoperative recurrence. Its CT plain scan is less uniform and slightly dense and with clear edge [
26‐
28]. In this study, the general characteristics of the two lesion images and the CT values of each period were compared, and we found that only the CT values of AP and the net enhancement value of AP were statistically different between the two groups, and both had certain differential diagnostic efficacy, which was basically consistent with the research conclusion of Seung Hoon Woo et al [
26].
At present, the differentiation of PA and AL is mainly through the above imaging features, which is the key diagnostic mode, however, based on the analysis of physicians is subjective and often affected by many factors, such as the size of the lesion, the clarity of the image, and the physician’s personal ability and experience, this study based on the average CT value and AP net enhancement value to distinguish the AUC of PA and AL is 0.79 and 0.78, respectively. The accuracy of traditional imaging features still needs to be improved to distinguish between the two.
Texture analysis is different from the competent image feature analysis, which quantifies the grayscale information of the image, reflects the microstructure and internal biological indicators of tumor tissue with objective indicators, provides more tumor heterogeneity information that cannot be observed by the naked eye, and distinguishes malignant tumors from normal tissues or benign lesions mainly through tumor heterogeneity [
14,
15]. Histogram analysis is a useful texture analysis method that the main parameters include the mean, variance, skewness, scale and the 1st, 10th, 50th, 90th and 99th percentiles [
19]. The parameters of the histogram can reflect the structure and heterogeneity of tumor tissue to a certain extent, and the average value can reflect the central trend and average level of the data. The variance is used to describe the degree of dispersion of gray values, that is, it can reflect the heterogeneity of the lesion components, the larger the variance, the more the data deviate from the mean, the stronger the heterogeneity; The skewness value is an indicator to measure the asymmetry of the gray value, the larger the absolute value of the skewness, the more asymmetric the distribution; The kurtosis value indicates that compared with the normal distribution, it can reflect the relative steepness of the distribution, the smaller the kurtosis value, the flatter the gray value distribution, and the more complex the tumor tissue components. The percentile reflects the structure of the grayscale histogram and quantifies the intra-tumor heterogeneity [
29,
30]. This study is based on CT texture analysis, and aims to investigate the value of the general features of enhanced CT images and the histogram parameters of each stage to identify PA and AL.
In the current study, the enhanced CT data of 20 PAs and 29 ALs were retrospectively analyzed, obtained the histogram characteristic parameters of the largest ROI of the tumor lesion, and found statistical differences in the mean, 10th, 50th, 90th, 99th percentile of AP and FSP and variance of AP and kurtosis of VP. We found that the mean, 10th, 50th, 90th, 99th percentile of AL was higher than the PA during FSP and the different parameters increase the variance during AP. We speculate that the histologic differences between the two kinds of tumors were reflected on these histogram parameters differences. As the name implies, PA is also known as mixed tumor, and the tumor composition is often composed of a mixture of two different germ layer tissues, with a rich and diverse tissue structure, the basic structure includes glandular epithelium, myoepithelium, mucus, mucus tissue [
31]. Low HU on early CT images of PA may be due to the abundance of mucus-like or fibrous matrix elements and low microvascular density, which is largely consistent with earlier findings [
20,
23,
25]. In contrast, high HU on early WT CT images may be due to histological features consisting of epithelial, simple, predominantly cancer cell components, and lymphoma stroma with a higher internal microvascular density [
24,
26‐
28]. In addition, the study also found that only kurtosis showed a difference during VP. We tried to explain this conclusion using its histopathological features and vascular structure [
32]. Histopathology, ALs have a high microvascular count and high cellularity, in contrast, PAs are with a large amount of mucinous matrix and rare epithelial components. Characteristic dynamic CT of ALs presents with rapid contrast enhancement at AP, decreased enhancement at VP, and shows slow flushing and shows steady horizontal enhancement, while PA gradually increases, which we speculate may be due to slow leakage of contrast medium from a small amount of microvascular into the vascular space and mucus-like matrix.
ROC analyses were performed for the ability of these parameters with statistically significant differences, individually and combined, to discriminate between PA and AL. The AUC of the mean CT value of AP was 0.79. The AUC of the AP net reinforcement value was 0.78. The maximum AUC of 90th in the FSP parameter was 0.87. The maximum AUC of 90th in the AP parameters was 0.83, and only the kurtosis values in the VP was statistically significant between the two groups, with an AUC of 0.71. The multi-parameter combined diagnostic ROC curve showed the best diagnostic efficacy of Mean of FSP + The 90th percentile of AP + Kurtosis of VP with an AUC of 0.93 and a sensitivity and specificity of 0.90 and 0.85, respectively.
It is of certain value to make the differential diagnosis of PA and AL by enhancing the maximum-level texture histogram analysis of CT images without extending the patient examination time and increasing the economic burden, and it can provide new ideas for clinical and imaging diagnosis. Limitations of this study: First, this study is retrospective, the sample size is small, there is a certain selectivity bias, and the sample size of the included studies between groups varies greatly, which may lead to error in the results. Although the sample size was small, we still found a strong association between texture features and the diagnosis of PA and AL, and we should proceed to investigate the association of broader radiological features with parotid gland tumors in a multicenter large sample. Second, although the ROI of tumor lesions is outlined with the participation of radiologists and oral and maxillofacial surgeons with extensive diagnostic experience, the selection of ROI is done manually, which leads to sampling bias. Third, this study only analyzes the first-order parameters of the most commonly used texture analysis method, histogram analysis, and the second-order parameters and high-order parameters have not yet been included in the study, and the later research will be further explored by algorithms such as auto-segmentation of radiomics and volume integration neural network.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.