Registration of multiphoton optical images of cervical tissue to quantitative ultrasound data

Collagen remodeling, leading to cervical softening and ripening, is essential for normal vaginal delivery. If this process happens too early there is increased risk of preterm birth, which puts the child at risk for life long health complications and early death. Collagen content and organization in the cervix plays a large role in cervical function. The cervix is highly aligned and layered, and this organization breaks down during remodeling. Cervical collagen remodeling begins early in pregnancy, but there is currently no reliable, non-invasive, quantitative metric for determining whether a cervix is changing too rapidly, thus no effective measure of the risk of preterm birth. We have been developing quantitative ultrasound (QUS) techniques to measure the alignment, and corroborating those QUS measurements with optical microscopy and quantitative image analysis in the form of curvelet transforms (a multi-scale transform similar to wavelets). We show how different choices of scale for the curvelet transform characterize features at different spatial scales. In particular we are investigating which spatial scale, chosen for curvelet analysis, is most predictive of the ultrasound alignment measurements. The spatially-registered US echo power loss measurements are consistent with gross inspection of the optical image itself and with a curvelets based analysis done at the largest spatial scale.