Table 1.
Eligibility criteria for NSLBP.
Fig 1.
Experimental set up for obtaining spine flexion.
Table 2.
Patient-reported outcome measures (PROMs).
Fig 2.
Human pose estimation model utilised to estimate spine flexion performance.
Fig 3.
Sagittal 2D position of the base of the neck keypoint.
Y = base of the neck vertical coordinate (pixels), X = base of the neck antero-posterior coordinate (pixels).
Fig 4.
Visual representation of the calculation to obtain the range of movement of spine flexion.
Euclidean distance (in pixels) between hip to neck (hn), ankle to hip (ah), and ankle to neck (an).
Fig 5.
Spine flexion angle waveform of a single participant repeating the movement with the signal pre-smoothing (blue) and post-smoothing (orange).
The spine flexion repetition time (tri), and angle depth (depthi) are highlighted in the graph to give insight into the consistency of participants attainment of a specific flexion depth and the extent to which the flexion depth fluctuated across repetitions.
Table 3.
Movement and clinical features used for distinguishing between MI and MCI.
Fig 6.
Batch Norm = batch normalisation layer, ReLU = Rectified Linear Unit layer.
Fig 7.
An example of an angle waveform from a single participant performing 10 repetitions of spine flexion derived from the pose estimation model and the motion capture data (Mocap).
Blue = pose estimation model; Orange = Mocap.
Table 4.
Classification performance when all spine flexion angle and patient reported outcome measures (PROMs) features were inputted.
Table 5.
Set of optimal input features yielding the highest classification performance.
Table 6.
Classification performance when the optimal features were inputted separately and combined.