Radiomics Based on Multimodal Transvaginal Ultrasound Imaging in Predicting Endometrial Cancer and Cervical Stromal Invasion

Endometrial Cancer is the second most common gynecological cancer in China and the first in Western countries. The common clinical symptom of endometrial cancer is vaginal bleeding, which occurs in about 10% of postmenopausal women. Most patients with postmenopausal vaginal bleeding are diagnosed with benign diseases, and less than 10% of patients are diagnosed with endometrial cancer. Early diagnosis is crucial for the prognosis of patients with endometrial cancer. The 5-year survival rate of patients with endometrial cancer which lesions localized to the uterus is about 95%, while the survival rate of patients with regional and distant metastasis is reduced to less than 70% and 20%.

Surgery is the main treatment of endometrial cancer. CSI is one of the main criteria for determining the follow-up treatment. According to NCCN guidelines, Total Hysterectomy and Bilateral Salpingo-Oophorectomy (THBSO) are standard treatments for patients with endometrioid carcinoma without CSI. While extensive hysterectomy or surgery after radiotherapy is appropriate for patients with CSI. Therefore, accurate assessment of CSI status in patients with EC before operation is important for the formulation of accurate surgical strategies.

Endometrial biopsy has been considered the gold standard for assessing endometrial cancer. However, it is limited because of increased cost, sample errors, related complications such as pain, bleeding, inability to evaluate the extent of tumor invasion and easy to cause tumor spread. CT/MR are alternative ways with high cost and complications. Transvaginal ultrasound examination is considered as the first imaging investigation for endometrial cancer. ESGO/ESTRO/ESP guidelines for the management of patients with endometrial carcinoma indicate that transvaginal ultrasound can be used instead of magnetic resonance imaging to detect cervical stromal infiltration under the operation of experienced doctors. Improving the performance of ultrasonic diagnosis is significant for how to choose the follow-up treatment and reduce the cost and risk of overtreatment.

Radiomics refers to high-throughput mining of quantitative image features from medical imaging. Radiomics derived data, when combined with other pertinent clinicopathological features, can produce accurate and robust evidence-based decision-making systems. Multimodal radiomics can provide more imaging feature information than single modal radiomics, which showed better diagnostic performance in previous study of kinds of cancer diseases.