Predictive Behavior of a Computational Foot/Ankle Model through Artificial Neural Networks

Table 3

Percent differences, ANN-predictions versus target stiffness.

Ligament	Target stiffness (N/mm)	Percent difference
		FFN	RBFN
Tibiocalcaneal	75.0	0.89%	6.22%
Tibionavicular	5.0	−5.79%	19.44%
Tibiospring 1	7.6	−5.79%	19.44%
Tibiospring 2	25.0	−5.79%	19.44%
Anterior tibiotalar	90.0	2.98%	2.00%
Posterior tibiotalar	117.0	2.98%	2.00%
Talocalcaneal interosseous	33.8	6.32%	6.58%
Plantar fascia 1	30.0	6.14%	3.61%
Plantar fascia 2	45.0	6.14%	3.61%
Plantar fascia 3	37.5	6.14%	3.61%
Plantar fascia 4	15.0	6.14%	3.61%
Plantar fascia 5	150.0	6.10%	2.72%
Spring 1	16.9	3.22%	8.73%
Spring 2	2.3	3.39%	29.90%

Ligament stiffness values for the patient-specific foot/ankle model alongside percent differences between target values and the ANN-predicted stiffness. (Negative percentages indicate a decrease relative to the target stiffness value.) Within the foot/ankle model, multiple linear elements comprise a single ligament; therefore, the components listed in the table represent each uniquely assigned stiffness value.