BACKGROUND AND PURPOSE Non-echo-planar DWI MR imaging (including the HASTE sequence) has been shown to be highly sensitive and specific for large cholesteatomas. The purpose of this study was to determine the diagnostic accuracy of HASTE DWI for the detection of incipient cholesteatoma in high-risk retraction pockets. MATERIALS AND METHODS This was a prospective study of 16 patients who underwent MR imaging with HASTE DWI before surgery. Surgeons were not informed of the results, and intraoperative findings were compared against the radiologic diagnosis. Sensitivity, specificity, and positive and negative predictive values were calculated. RESULTS Among the 16 retraction pockets, 10 cholesteatomas were diagnosed intraoperatively (62.5%). HASTE showed 90% sensitivity, 100% specificity, 100% positive predictive value, and 85.7% negative predictive value in this group of patients. We found only 1 false-negative finding in an infected cholesteatoma. CONCLUSIONS We demonstrate a high correlation between HASTE and surgical findings, suggesting that this technique could be useful for the early detection of primary acquired cholesteatomas arising from retraction pockets and could help to avoid unnecessary surgery.