Detection of parametrial invasion in women with uterine cervical cancer using diffusion tensor imaging at 1.5T MRI

Abstract

Purpose: The purpose of this study was to prospectively evaluate the capability of diffusion tensor imaging (DTI) of the lumbosacral plexus to identify parametrial invasion by uterine cervical cancer.

Materials and methods: Twenty-seven women with biopsy-proven cervical cancer were prospectively enrolled and underwent DTI at 1.5 TMRI. Fractional anisotropy (FA) values were calculated at the level of right and left L5 and S1 roots. The two sides of each patient were considered independently in two groups, according to the presence or absence of parametrial invasion. Differences between FA values of invaded parametria and those of non-invaded parametria were searched using Student t-test. Receiver operating characteristic (ROC) analysis was performed to identify the cut-off value of FA that yielded best sensitivity, specificity and accuracy for the diagnosis of parametrial invasion.

Results: A total of 54 parametria in 27 participants (mean age, 52.9 ± 12 years; age range, 30-81 years) were analyzed. Invasion was present in 37/54 (68%) parametria and absent in 17/54 (31%) parametria. FA was greater in parametrial invasion (mean, 0.321 ± 0.036; range: 0.285-0.357) than in the absence of parametrial invasion (0.292 ± 0.02; range: 0.272-0.312) (P = 0.01). At ROC analysis, best cut-off value of FA for the diagnosis of parametrial invasion was >0.3099 (AUC, 0.681; 95% CI: 0.583- 0.768), yielding 62% sensitivity (95% CI: 50.3-73.64), 73% specificity (95% CI: 50.6-85.27) and 66% accuracy (95% CI: 54.62-73.91).

Conclusion: Using >0.3099 as cut off-value for FA of L5-S1 roots, DTI has an accuracy of 73% in the diagnosis of parametrial invasion by uterine cervical cancer.

Keywords: Anisotropy; Diffusion tensor imaging; Parametrial invasion; Prospective studies; Uterine cervical neoplasms.