The Automatic Quantification of Morphological Features of Pectus Excavatum Based on Three-Dimensional Images

Abstract

Visual examination and quantification of severity are essential for clinical decision making in patients with pectus excavatum. Yet, visual assessment is prone to inter- and intra-observer variability and current quantitative methods are inadequate. This study aims to develop and evaluate a novel, automatic and non-invasive method to objectively quantify pectus excavatum morphology based on three-dimensional images. Key steps of the automatic analysis are normalization of image orientation, slicing, and computation of the morphological features encompassing pectus depth, width, length, volume, position, steepness, flaring, asymmetry and mean cross-sectional area. A digital phantom mimicking a patient with pectus excavatum was used to verify the analysis method. Prospective three-dimensional imaging and subsequent surface analysis in patients with pectus excavatum was performed to assess clinical feasibility. Verification of the developed analysis tool demonstrated 100% reproducibility of all morphological feature values. Calculated parameters compared to the predetermined phantom dimensions were accurate for all but four features. The pectus width, length, volume and steepness showed an error of 4 mm (4%), 2 mm (2%), 12 mL (5%) and 1 degree (3%), respectively. Prospective imaging of 52 patients (88% males) demonstrated the feasibility of the developed tool to quantify morphological features of pectus excavatum in the clinical setting. Mean duration to calculate all features in one patient was 7.6 seconds. We have developed and presented a non-invasive pectus excavatum surface analysis tool, that is feasible to automatically quantify morphological features based on three-dimensional images with promising accuracy and reproducibility.