Objectives: The aims of the study were to evaluate ultrasound (US) in the diagnosis of parametrial endometriosis compared with surgical-pathological findings, and to define an optimal cut-off of lesion size for correlation between torus/uterosacral ligaments (USL) or rectosigmoid and parametrial involvement by deep endometriosis. Methods: Longitudinal study of 60 patients referred for surgical management of pelvic endometriosis, who underwent sonography performed by an experienced sonographer. The presence of parametrial endometriosis shown by US was compared with surgical and histologic findings. The presence of endometrial cysts, deep endometriotic locations, and their potential association with parametrial endometriosis was assessed by US. The sensitivity, specificity, predictive values (PV), accuracy, and positive and negative likelihood ratios (LR) of US for predicting parametrial endometriosis and other pelvic endometriotic locations were assessed. Descriptive analysis, optimal cut-off analysis, categorical (Fisher’s exact test) and non-categorical variables (Mann—Whitney) were calculated. Results: Parametrial, ovarian, and deep endometriosis were found by surgery in respectively 30 (50%), 21 (35%), and 59 (98.63%) of the 60 patients. The sensitivity, specificity, PPV, NPV, accuracy, PLR, and NLR of US for the sonographic diagnosis of parametrial endometriosis were 40%, 96.7%, 92.3%, 61.7%, 68.3%, 12, and 0.62 respectively. In patients with parametrial involvement, a relation with rectosigmoid ( p = 0.005) and USL ( p = 0.0074) endometriosis was noted. For isolated torus/USL and rectosigmoid endometriosis, optimal cut-offs suggesting parametrial involvement by DE were 11 mm and 30 mm, respectively. Conclusions: US has low sensitivity but high specificity to diagnose parametrial endometriosis and could be used to rule in diagnosis before surgery.