Endometrial cancer (EC) is one of the most common gynecologic cancers in developed countries. Presently, it is imperative to develop a reliable, non-invasive, or minimally invasive detection method for EC. We explored the possibility of using DNA methylation marker from endometrial brush samples (with a “Tao brush”) and cervical scrapes (with a “Pap brush”) for early detection of EC. We analyzed the methylation data of EC and normal endometrial tissues from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) datasets. Optimized methylation-sensitive restriction enzyme combined real-time fluorescent quantitative PCR (MSRE-qPCR) was used for methylation detection. Included in the training set were 143 endometrial tissues, 103 Tao and 109 Pap brush samples. The validation set included 110 Tao and 112 Pap brush samples. PCDHGB7 was significantly hypermethylated in EC compared with normal endometrial tissues in TCGA and GEO datasets (AUC > 0.95), which was verified in clinical samples. In Pap brush samples, the AUC was 0.86 with 80.65% sensitivity and 82.81% specificity, while Tao brush samples exhibited higher specificity (95.31%). The combination of Tao and Pap brush samples significantly increased the sensitivity to 90.32%. In the validation set, the final model yielded sensitivity of 98.61%, specificity of 60.53%, positive predictive value (PPV) of 82.56%, and negative predictive value (NPV) of 95.83%. These results demonstrate the potential application of the novel methylation marker, hypermethylated PCDHGB7, in cervical scrapings and endometrial brush, which provides a viable, non-invasive, or minimally invasive method for early endometrial cancer detection across different clinical features and histologies to supplement current hysteroscopy diagnosis.