The role of new ultrasound modalities to refine the diagnostic workup on women with cervical carcinoma

Sammanfattning: Background: Cervical cancer is the fourth most common female cancer in the world and has traditionally been clinically staged not including imaging. Despite that, clinicians depend on diagnostic imaging in the management of patients to triage them to treatment. Ultrasonography (US) is the most widely used diagnostic imaging technique by gynecologists worldwide and is equally accurate as Magnetic resonance imaging (MRI) for local assessment of cervical cancer in the pelvis in low volume disease, whereas both methods lack accuracy for detection of lymph node metastasis. New ultrasound modalities such as three-dimensional ultrasonography (3D US), contrast-enhanced ultrasonography (CEUS) and strain elastography (SE) have shown promising results in the assessment of tumours, but the clinical value in cervical cancer patients is uncertain. The aims of this thesis were to explore the clinical value of 3D US, CEUS and SE and to assess the inter-rater agreement of conventional US and MRI in the assessment of patients with cervical cancer. Methods: Study I was a multicenter study including 104 women with surgically resectable cervical cancer from 5 European institutions. Studies II-IV were based on a single center cohort of 93 women with all stages of cervical cancer, from Karolinska University Hospital. In study I subjective evaluation was compared to objective measurements of 2D and 3D US parameters for prediction of deep stromal and parametrial invasion as well as lymph node metastases. In study II semi-quantitative parameters and filling pattern of CEUS were compared in 49 patients with cervical cancer and 21 healthy controls. The features of SE in 30 patients with all stages of cervical cancer were explored in study III. In study IV the inter-rater agreement of raters with varying experience on US and MRI were compared in off-line settings for the assessment of cervical tumours in a cohort of 60 patients with all stages of cervical cancer. Results: Subjective assessment with 2D US had high accuracy to detect deep stromal and parametrial invasion (sensitivity 91% and 100%; specificity 97% and 95% respectively) but not to predict lymph node metastases. Tumour size measured by 2D and volume measured by 3D US were accurate to predict deep stromal invasion (AUC 0.83 and 0.85) but outperformed by subjective assessment. 3D vascular indices (VI, VFI, FI) had no value to predict deep stromal invasion or lymph node metastases. A focal CEUS pattern had a higher specificity 91% versus 73% (p=0.62) and a similar sensitivity 80% versus 85% (p=1.00) compared to subjective assessment using conventional US for tumour detection. The semi-quantitative CEUS parameter area under the time intensity curve had high accuracy to separate tumour lesions from healthy stroma (AUC 0.923, p<0.001). Size measures of early stage tumours were accurate with SE (mean difference -0.11 mm, p=0.66) with no bias found. An elasticity score of 4-5 was found in 45% (9/20) with early stage and 80% (8/10) with advanced disease where SE was even useful in 7/10 (70%) of cases to delineate tumour borders. The inter-rater agreement for tumour visualization, 1/3 deep stromal and parametrial invasion was moderate for US raters (0.4-0.6) and moderate-good (0.4-0.8) for MRI raters irrespective of experience. Conclusion: Subjective 2D US assessment is accurate for detecting deep stromal and parametrial invasion where it outperforms objective measurements with 2D and 3D vascular indices. A focal contrast pattern with CEUS has high specificity and high accuracy for tumour detection. SE is accurate to measure tumour size and may improve delineation of advanced tumours. In off-line settings, inter-rater agreement is moderate for US raters and moderate-good for MRI raters irrespective of experience.