Method to classify elderly subjects as fallers and non‐fallers based on gait energy image. Issue 3 (25th September 2014)

• Record Type: Journal Article
• Authors: Gandomkar, Ziba; Bahrami, Fariba
• Is Part Of: Healthcare technology letters. Volume 1:Issue 3(2014)
• Page Start: 110
• Page End: 114
• Publication Date: 2014-09-25
• Subjects: image classification -- image colour analysis -- image sequences -- geriatrics -- medical image processing -- gait analysis -- data visualisation; elderly subject classification -- gait energy image -- GEI -- coloured gait energy image -- CGEI -- nonfallers -- gait pattern visualization -- timed up and go test -- TUG test -- fall risk assessment -- clinical tools -- gait sequences -- cognitive load -- gait cycles -- walking sequence -- colour components -- histogram‐based features -- correct classification rate; Biomedical engineering -- Periodicals; Medical technology -- Periodicals; 610.28
• DOI: 10.1049/htl.2014.0065
• Languages: English
• ISSNs: 2053-3713

Abstract

Abstract : Falls are one of the leading causes of injuries among the elderly. Therefore, distinguishing fallers and performing preventive actions is vitally important. A new variation of the gait energy image (GEI) called coloured gait energy image (CGEI) is proposed for classifying subjects as fallers and non‐fallers and for visualising their gait patterns. Eight elderly fallers, eight elderly non‐fallers and eight young subjects performed timed up and go (TUG) test, which is one of the well‐known clinical tools for fall risk assessment and contains two gait sequences. Subjects were also asked to perform two other variations of the TUG test, namely TUG with manual load and TUG with cognitive load. Gait sequences were extracted from the TUG test based on the opinion of three human observers. Then the gait cycles were automatically extracted from the walking sequence and divided into three phases, corresponding to double support and first and second half of single support. Next, the GEI of each phase was generated and formed one of the colour components of CGEI. Histogram‐based features obtained from CGEI were then used to classify the video collected from walking sequences of elderly fallers and non‐fallers. Correct classification rate was improved by approximately 27% compared with the standard TUG test.