Abstract
BACKGROUND: Different types of testicular tumours require different treatments. Malignant germ cell tumours usually require radical orchiectomy, while benign tumours may only require excision of the mass. Currently, conventional imaging examinations are difficult to diagnose testicular pathology types. Since testicular tumours are mainly malignant germ cell tumours, puncture of such tumours may lead to tumour metastasis. Therefore, testicular tumours are generally not recommended for puncture biopsy, and radical orchiectomy is performed directly, which inevitably leads to the mistaken removal of testicles in some patients with benign tumours.Since ultrasound texture analysis can evaluate tumour echoes and uniformity in greater detail and with greater accuracy than grey-scale ultrasound, it is helpful in distinguishing tumour pathology types. METHODS: This study is a retrospective analysis that included 86 patients (with a total of 89 testicular lesions) with pathologically confirmed primary testicular tumours at our hospital between February 2017 and June 2021. The patients were divided into four groups based on pathological results: epidermoid cyst group, sertoli-leydig cell tumour group, lymphoma group, and malignant germ cell tumour group. A retrospective analysis was conducted on conventional ultrasound and its textural analysis performance, which was then compared with pathological findings. Count data were assessed using a paired chi-square test to evaluate the diagnostic performance of both examination methods.Subsequently, a binary logistic regression analysis was conducted on the nine texture feature parameters alongside age and size to derive predictive probability values, thereby achieving comprehensive combined diagnosis. These parameters and the combined results were then plotted on a ROC curve, with the corresponding area under the curve (AUC) calculated. RESULTS: This study included a total of 89 testicular lesions, including 10 cases of sertoli-leydig cell tumours; 18 cases of epidermoid cysts; 19 cases of lymphoma; and 42 cases of malignant germ cell tumours. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of conventional ultrasound diagnosis for sertoli-leydig cell tumours were 40.0%, 96.2%, 57.1%, 92.7%, and 89.9%, respectively. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of conventional ultrasound diagnosis of testicular epidermoid cysts were 22.2%, 100.0%, 100.0%, 83.5%, and 84.3%, respectively. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of conventional ultrasound diagnosis for lymphoma were 15.8%, 90.0%, 30.0%, 79.7%, and 74.2%, respectively. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of conventional ultrasound diagnosis for malignant germ cell tumours were 19.0%, 89.4%, 61.5%, 55.3%, and 56.2%, respectively. Using the one-sample K-S test to examine the normality of texture parameters (min gray, max gray, standard, skewness, contrast, sum average, difference variance, difference entropy, and dissimilarity) in testicular tumours alongside patient age and tumour size, it was found that standard, sum average, and tumour size conformed to normal distribution, whilst the remaining parameters did not. Subsequently, these nine texture feature parameters were combined with age and size in a binary logistic regression analysis to derive predictive probability values for comprehensive joint diagnosis. ROC curves were plotted for the nine texture feature parameters, age, size, and the full joint model. Plot ROC curves for the sertoli-leydig cell tumour group and the unsertoli-leydig cell tumour group. resulting in a maximum AUC of 0.992 for the combined groups. Based on the AUC curve, the combined groups with an AUC > 0.148 demonstrated sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 100.0%, 96.2%, 76.9%, 100.0%, and 96.6%, respectively, for diagnosing sertoli-leydig cell tumours. ROC curves were plotted for the epidermoid cyst group and the non-epidermoid cyst group. yielding a maximum AUC of 0.970 for the combined group. Based on the AUC curve, the combined group with an AUC > 0.357 demonstrated sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 94.4%, 95.8%, 85.0%, 98.6%, and 95.5%, respectively, for diagnosing epidermoid cysts. ROC curves were plotted for the lymphoma group and the non-lymphoma group, yielding a maximum AUC of 0.971 for all combined variables. Based on the AUC curve, the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of all combined variables > 0.136 for diagnosing lymphoma were 100.0%, 87.1%, 67.9%, 100.0%, and 89.9%, respectively. ROC curves were plotted for the malignant germ cell tumour group and the non-malignant germ cell tumour group. yielding a maximum AUC of 0.809 for the combined groups. Based on the AUC curve, the combined groups with an AUC > 0.560 demonstrated sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 66.7%, 93.6%, 90.3%, 75.9%, and 80.9%, respectively, for diagnosing malignant germ cell tumours. The combined use of all methods significantly improved the accuracy of conventional ultrasound diagnosis, with statistically significant differences. CONCLUSION: Ultrasound texture analysis combined with age and size can provide a relatively accurate pathological classification diagnosis of epidermoid cysts, lymphoma, sertoli-leydig cell tumours, and malignant germ cell tumours, thereby improving the accuracy of routine ultrasound diagnosis. Different pathological types of testicular tumours require different treatment plans, and preoperative ultrasound diagnosis is of great significance as it can guide clinical treatment and preserve the testicles of some patients.