Advanced imaging biomarkers in endometrial cancer
Not peer reviewed
MetadataShow full item record
Background: Endometrial cancer is the most common gynecological cancer in highdeveloped regions of the world, and the incidence has been increasing over the last half century, largely driven by a concurrent increase in population obesity. Primary treatment is surgical in most cases, but only limited preoperative risk stratification has been applied in traditional clinical practice. To enable more individualized treatment, improved methods for preoperative tumor characterization are highly warranted.
Aim: To identify and evaluate new imaging markers that may aid in the preoperative risk stratification and tailoring of treatment in endometrial cancer.
Material and methods: The studies included in this thesis are based on collected imaging-, clinical- and histological data from endometrial cancer patients treated at Haukeland University Hospital during April 2009 to November 2013. Standardized MR imaging data were acquired for 216 prospectively included patients with histologically confirmed endometrial cancer. From this cohort, four different subcohorts were included in study I-IV. In Paper I, three radiologists independently measured tumor size on conventional MR images for 212 patients. In Paper II, metabolic features were extracted from MR spectroscopy performed on 77 patients. In Paper III, texture features were extracted from MR images, using a filtrationhistogram technique in 180 patients. In Paper IV, CT imaging data were retrospectively collected and texture features extracted for 155 patients. In all studies, the respective imaging markers were evaluated as predictors of histopathological highrisk features and survival.
Results: The interobserver variability for MRI-measured tumor size is very low (ICC 0.78-0.85) (Paper I). AP diameter greater than 2 cm independently predicts deep myometrial invasion (OR 6.7, p< 0.001) and CC diameter greater than 4 cm independently predicts lymph node metastases (OR 4.9, p=0.009) when adjusting for conventional MRI reading results and risk status based on preoperative endometrial biopsy (Paper I). CC tumor diameter has an independent impact on recurrence- and progression-free survival (adjusted HR 1.04, p=0.009) (Paper I).
Tumor tissue has significantly higher MR spectroscopy-derived ratios for tCho/Creatine, tCho/Water and tCho/Noise than normal myometrial tissue (p<0.001 for all) (Paper II). High tumor tCho/Water ratio is also significantly associated with high histological tumor grade in endometrioid tumors (p=0.02) (Paper II). No significant associations are found between tumor tCho-levels and recurrence- and progression-free survival (Paper II).
When performing texture analysis of MR images, high tumor entropy in ADC-maps independently predicts deep myometrial invasion (OR 3.2, p<0.001), and high MPP in T1c images independently predicts high-risk histological subtype (OR 1.01, p=0.004) when adjusting for MRI-measured tumor volume, conventional MRI reading results and biopsy risk status (Paper III). Furthermore, high kurtosis in T1c images independently predicts reduced recurrence- and progression-free survival (adjusted HR 1.5, p<0.001) (Paper III).
When performing texture analysis of CT images, high tumor entropy independently predicts deep myometrial invasion (OR 3.7, p=0.008) and cervical stroma invasion (OR 3.9, p=0.02) when adjusting for MRI-measured tumor volume, conventional MRI reading results, age and biopsy risk status (Paper IV). High value of MPP (MPP5>24.2) independently predicts high-risk histological subtype (OR 3.7, p=0.01) (Paper IV). High tumor kurtosis tends to independently predict reduced recurrenceand progression-free survival (adjusted HR 1.1, p=0.06) (Paper IV).
Conclusions: Tumor size can be measured on preoperative conventional MRI with very low interobserver variability. Large tumor size predicts deep myometrial invasion, lymph node metastases and poor outcome in endometrial cancer, and thus, imaging markers based on tumor size may improve preoperative risk stratification (Paper I). High choline levels, measured by MR spectroscopy, differentiate tumor tissue from normal tissue in endometrial cancer patients, but do not have significant prognostic value in our study (Paper II). MRI-derived tumor texture parameters predict deep myometrial invasion, high-risk histological subtype, and reduced recurrence- and progression-free survival in endometrial cancer (Paper III). CT-derived tumor texture features predict deep myometrial invasion and cervical stroma invasion in endometrial cancer, and also tend to predict high-risk histological subtype and survival (Paper IV). The image texture features entropy, kurtosis and MPP seem to reflect tumor heterogeneity and may aid in preoperative risk assessment (Paper III and IV).