Standardized Reporting of Core Needle Biopsy TortuosityBiopTort—A Computer-Aided Protocol for Categorizing Tortuosity
Context.—
Core needle biopsies (CNBs) are among the most common biopsy procedures, yielding long, thin deformation-prone cores. Core deformation results in both distorted histologic features and tortuosity (ie, deviation from the ideal straight path) along the length of the core. Tortuosity creates challenges in slide preparation and downstream pathologist assessment (eg, absent diagnostically relevant tissue after sectioning), indicating a clinical need for its minimization. Despite this need, there is no standard protocol for reporting CNB tortuosity.
Objectives.—
To (1) establish an actionable protocol for scoring CNB tortuosity, (2) build BiopTort, a software tool assigning CNB images interpretable, tiered scores consistent with the protocol, and (3) investigate our protocol and BiopTort’s impact on interpathologist concordance and scoring time.
Design.—
Using a held-out CNB data set (N = 167), 3 pathologists assessed tortuosity in 3 sequential stages: (1) Baseline: a 3-tier scale based on prior experience, (2) Protocol: rescored with our 4-tier Tortuosity Scoring Protocol, (3) BiopTort-Aided: rescored with the protocol alongside BiopTort. Interrater concordance for each stage, using the Fleiss κ, was recorded.
Results.—
Fleiss κ was 0.19, 0.40, and 0.63, and average time per slide was 8.0, 24.2, and 13.3 seconds for stages 1 to 3, respectively. BiopTort was statistically noninferior to human interrater pairwise concordance.
Conclusions.—
These results suggest our protocol lessens interpathologist variability, with BiopTort further reducing variability and scoring protocol employment time cost. When deployed in CNB quality control workflows, our computer-aided protocol appears well situated to facilitate standardized, efficient, and actionable reporting of CNB tortuosity. BiopTort is open-source (http://bioptort.com).
Contributor Notes
This work was supported by the National Institutes of Health under award numbers 1R01LM013864, 1U01DK133090, and R01DK118431.
The authors have no relevant financial interest in the products or companies described in this article.
Supplemental digital content is available for this article. See text for hyperlink.