During metastasis, tumor cells dynamically change their cytoskeleton to traverse through a variety of non-adherent microenvironments, including the vasculature or lymphatics. Due to the challenges of imaging drift in non-adhered tumor cells, the dynamic cytoskeletal phenotypes are poorly understood. We present a new approach to analyze the dynamic cytoskeletal phenotypes of non-adhered cells that support microtentacles (McTNs), which are cell surface projections implicated in metastatic reattachment. Combining a recently-developed cell tethering method with a novel image analysis framework allowed McTN attribute extraction. Full cell outlines, number of McTNs, and distance of McTN tips from the cell body boundary were calculated by integrating a rotating anisotropic filtering method for identifying thin features with retinal segmentation and active contour algorithms. Tethered cells behave like free-floating cells; however tethering reduces cell drift and improves the accuracy of McTN measurements. Tethering cells does not significantly alter McTN number, but rather allows better visualization of existing McTNs. In drug treatment experiments, stabilizing tubulin with paclitaxel significantly increases McTN length, while destabilizing tubulin with colchicine significantly decreases McTN length. Finally, we quantify McTN dynamics by computing the time delay autocorrelations of 2 composite phenotype metrics (cumulative McTN tip distance, cell perimeter:cell body ratio). Our automated analysis demonstrates that treatment with paclitaxel increases total McTN amount and colchicine reduces total McTN amount, while paclitaxel also reduces McTN dynamics. This analysis method enables rapid quantitative measurement of tumor cell drug responses within non-adherent microenvironments, using the small numbers of tumor cells that would be available from patient samples.