Frailty Assessment Using Temporal Gait Characteristics and a Long Short-Term Memory Network

Detecting Fall Risk and Frailty in Elders with Inertial Motion Sensors: A Survey of Significant Gait Parameters

“…The studies included patients with frailty, fall risk, both, or other diseases: 11 studies were focused on patients with frailty [ 15 , 24 , 26 , 27 , 28 , 29 , 31 , 35 , 36 , 41 , 42 ], 6 studies patients with fall risk [ 6 , 18 , 30 , 33 , 37 , 39 ], and 2 studies combining frailty and fall risk: [ 25 , 38 ]. Two studies included patients with frailty plus mind and peripheral artery disease [ 32 , 40 ], respectively; and one study [ 34 ] included patients with fall risk and Parkinson’s.…”

“…A large part of the studies used a single sensor, usually located on the back (lower back, L3 or L5 vertebrae) [ 24 , 27 , 30 , 37 , 39 , 42 ], or in some cases on the sternum [ 29 , 41 ], chest [ 32 ], or foot [ 6 ]. Three studies employed two IMUs, which were located on the same part of the body: feet [ 26 ], shins [ 35 ], and heels [ 36 ]. Studies that employed 3 IMUs [ 18 , 33 ], 5 IMUS [ 15 , 25 , 31 , 38 , 40 ], or 6 IMUs [ 28 , 34 ] combined different body positions as the thighs, shins, feet, shanks, back, ankles, pelvis, or sternum.…”

“…The most frequently repeated tests consisted of simply walking a fixed distance or time, normally between 3 and m 7 m or 0.5 min and 30 min [ 6 , 26 , 30 , 31 , 33 , 34 , 35 , 36 , 37 , 38 , 42 ]. Some studies realized more complex walking tests, in which the subject was performing another task such as counting (dual-task tests) [ 15 , 18 , 40 ].…”

“…Other methods such as partial least-squares discriminant analysis were applied by [ 37 , 39 ], and decision tree models were used by [ 24 , 32 , 42 ]. Moreover, there were studies that supported the use of new machine learning techniques such as neural network models [ 26 , 35 ] or support vector machine methods [ 25 ].…”

Video_5.AVI

“…Artificial neural networks might also help to further investigate the frailty of gait. Dawoon recently analysed the gait statistics gathered from gyroscopes placed on the feet using a long short-term memory networkbased classifier (Jung et al, 2021). Akbari performed a Kinectsensor machine learning methodology as a frailty classifier via functional assessment exercises including a walking test (Akbari et al, 2021).…”