Classification of microhysteroscopic images and their correlation with histologic diagnoses

Abstract

Study objective: To confirm that images observed at hysteroscopy correlate with histopathologic diagnoses.

Design: Double-blind study (Canadian Task Force classification II-2).

Setting: Gynecologic cancer center, private institute.

Subjects: One thousand four hundred thirty-six uterine cavities.

Intervention: Hysteroscopy.

Measurements and main results: Images were classified as normal hysteroscopy, benign lesion, low-risk hyperplasia, high-risk hyperplasia, and carcinoma. A hysteroscopic diagnosis was made and biopsy specimens were obtained with Kevorkian-type curettes or Sims curettes. Tissues were studied by a pathologist, after which we compared endoscopic diagnoses with anatopathologic diagnoses. Significant correlation was found between suspicion based on images and histologic confirmation (p = 0.001). The Cramer V coefficient, which measures the relationship between both methods, was high: 0.925. The Cramer V coefficient takes values ranging from zero, to indicate lack of correlation, to 1, to indicate perfect correlation. A lambda symmetry coefficient of 0.96 is interpreted as probable improvement in the prediction of histologic diagnosis based on images.

Conclusion: This classification system can be useful for a systematic approach to hysteroscopic findings and to improve communication among specialists. It is based on the degree of hysteroscopic suspicion aimed at early detection of endometrial cancer and its precursor lesions.