Iterative active deformational methodology for tumor delineation: Evaluation across radiation treatment stage and volume

Abstract

Purpose: To introduce, implement, and assess an iterative modification to the active deformational image segmentation method as applied to cervical cancer tumors. Materials and Methods:A comparison by Jaccard similarity (J S ) between this active deformational method and manual segmentation was performed on tumors of various sizes across preradiation, 3 weeks postradiation, and 6 weeks postradiation using a General Linear Mixed Model across 121 studies from 52 patients with Stage IIB-IV cervical cancers. Results:The deformable segmentation method produced promising levels of agreement including J S factors of 0.71 Ϯ 0.11 in the preradiation studies. The analysis illustrated a rate of improvement in J S with increasing tumor volume that differed between the preradiation and 6 weeks postradiation stage (P ϭ 0.0474). In the large preradiated tumors each additional cm 3 of volume was associated with an increase or improvement in J S of 0.0008 (95% confidence interval [CI]: 0.0003, 0.0014). In the smaller postradiation tumors, each additional cm 3 of volume was associated with a more robust improvement in J S of 0.0046 (95% CI: 0.0009, 0.0082). Conclusion:Agreement was strongly affected by tumor volume, and its performance was most impacted across volume in the later stages of radiation therapy. The deformation-based segmentation method appears to demonstrate utility for delineating cervical cancer tumors, particularly in the earliest stages of radiation treatment, where agreement is greatest.

A Feasibility Study to Determine Whether Clinical Contrast Enhanced Magnetic Resonance Imaging can Detect Increased Bladder Permeability in Patients with Interstitial Cystitis

“…All image slices were semi-automatically segmented [12], with additional corrections to remove the catheter and improve bladder wall definition. Contrast enhancement at each pixel was reported as a post-contrast (9-min)/pre-contrast ratio, and log10 transformed, for entire bladders for each image slice.…”

Deep Learning Hybrid Techniques for Brain Tumor Segmentation

“…Image segmentation becomes complicated by intensity inhomogeneity, slow segmentation speed, and narrow area of application which can be corrected by an additive bias correction (ABC) model [ 28 ]. In [ 29 ] authors created an active deformable model for cervical tumor identification in 2008. Suri and his colleagues developed a feature-based recognition and edge-based segmentation method for measuring carotid intima-media thickness (cIMT) in 2011 [ 30 ].…”