Background
Surgical treatment of deep infiltrating endometriosis (DIE) is complex, and preoperative diagnosis benefits both surgeon and patient. Studies in expert centres have reported high accuracy for transvaginal ultrasound (TVUS) diagnosis of DIE. External validation of these findings has been limited, and no information is available on how quickly these skills can be acquired. The aim of this study was to measure the learning curve of DIE‐TVUS and to identify the causes for inaccuracies in the diagnosis of bowel lesions and Pouch of Douglas (POD) obliteration.

Methods
Following one week of training at the University of São Paulo (Brazil), 205 consecutive women with a history of endometriosis symptoms were prospectively assessed by TVUS after minimal bowel preparation. TVUS findings were correlated with laparoscopic findings in eighty‐five cases to assess the accuracy. The LC‐CUSUM and CUSUM were used to assess the learning curve and maintenance of competency, respectively.

Results
The sensitivity and specificity for DIE of the bladder, vagina and bowel were 33% and 100%, 80% and 100%, and 88% and 93%, respectively. The sensitivity and specificity for the presence of POD obliteration were 88% and 90%, respectively. LC‐CUSUM analysis confirmed that competency for DIE‐TVUS was achieved within 38 scans for the detection of POD obliteration and within 36 scans for the detection of bowel nodules. Competency was maintained for the remainder of the scans as assessed by the CUSUM.

Conclusions
After one week of DIE‐TVUS training, competency can be achieved within forty procedures, allowing diagnosis of DIE with similar diagnostic accuracy as reported by centres of excellence.