Vertebral corners detection on sagittal X-rays based on shape modelling, random forest classifiers and dedicated visual features (vol 7, pg 134, 2019)

Quantitative measurements of spine shape parameters on planar X-rays is critical for clinical applications, but remains tedious and with no fully automated solution demonstrated on the whole spine. This study aims to limit manual input, while demonstrating precise vertebrae corners positioning and shape parameter measurements from sagittal radiographs of the cervical and lumbar regions, exploiting novel dedicated visual features and specialised classifiers. A database of manually annotated X-rays is used to train specialised Random Forest classifiers for each spine region and corner type. An original combination of local gradient characteristics, Haar-like features and contextual features based on patch intensity and contrast is used as visual features. The proposed method is trained and evaluated on 109 sagittal X-rays of asymptomatic and pathological subjects, from multiple imaging sites, and with a large age range (6–69 years old). Performance is first evaluated for positioning a 2D spine shape model, where precisely detected corners enable to adjust the model via an original multilinear statistical regression. Root-mean-square errors of corners localisation and vertebra orientations are reported, demonstrating state-of-the-art precision compared to existing methods, but with minimal manual input. The method is then evaluated for the extraction of additional vertebrae shape characteristics, such as vertebral centre positioning, endplate centres positioning and endplate length measures, rarely studied in previous literature. The proposed method enables, with minimal initialisation, fast and precise individual vertebrae delineations on sagittal radiographs of normal and pathological cases, with a precision and robustness level required for objective support for diagnosis and therapy decision-making. © 2018, © 2018 Informa UK Limited, trading as Taylor & Francis Group.