%0 Journal Article %A Andersen, Rasmus Elbaek %A Arendt-Nielsen, Lars %A Madeleine, Pascal %D 2016 %I Begell House %K vibration arthrometry, knee disorders, accelerometer, vibroarthrographic processing, classifiers, feature selection, osteoarthritis %N 1-2 %P 13-32 %R 10.1615/CritRevPhysRehabilMed.2016017185 %T A Review of Engineering Aspects of Vibroarthography of the Knee Joint %U https://www.dl.begellhouse.com/journals/757fcb0219d89390,7d05545f5ad8aa9c,47e4a7bb4f6fe824.html %V 28 %X Vibroarthrography (VAG) of the knee has been a noninvasive diagnostic tool for more than 100 years. During the last two decades, electronic sensors and digital signal processing have become central key elements of VAG research. Several sensor types and processing approaches have been used, but to date no consensus on the most adequate way to record and process VAG signals has been reached. In this review, we investigate two central aspects: the recording techniques and the processing chains used in the field of knee VAG. In the majority of VAG studies, a single accelerometer is used as the sensor and signals are recorded in a frequency range below 1000 Hz. Time−frequency features, statistical features, spatiotemporal features and combinations of these feature types have been extracted from VAG recordings. The most frequently used classifiers have been neural network−based classifiers and vector support machines. Classifier accuracies of >90% and area under the receiver operating characteristic curve >0.9 have been achieved, showing that VAG has great potential as a diagnostic tool. In the future, multi-channel VAG recordings may lead to better discrimination of the severity of and the location of the knee injury or disorder. %8 2016-12-13