Application of Neural Network and Finite Element Method for Multiscale Prediction of Bone Fatigue Crack Growth in Cancellous Bone - Multiscale Computer Modeling in Biomechanics and Biomedical Engineering - page 23

Microcrack in a single trabecula

(a) Damage. (b) Crack propagation. (c) Microcracks at trabeculae

under fatigue test.

Fig. 12 Predicted fatigue damage (a) and cracking (b) of a single trabecula of a transversally
loaded trabecula; c SEM image of a fractured trabecula [ 10 ]

(a) (b)

Fig. 13 Comparison between NN Prediction and FE verification of Cr.Dn and Cr.Le
accumulation versus cycles. Example of sample (BV = TV ¼ 25 : 035 %) (Table 1 ). First, NN
prediction was performed. Second, FE verification was applied to check the validity of the NN.
a Crack density, b crack length

prediction and the FE one. The NN computation time was about 1 s. The FE
computation time was about 1 h (64 GB dual-core computer). The results indicate
that the NN model was able to predict rapidly and accurately the average Cr.Dn and
Cr.Le accumulation within trabecular bone under varying applied compressive
stress.

6.2 Fatigue Cracks Within a Proximal Femur
Organ: Multiscale Results

The previously trained NN was exploited as a tool to rapidly estimate the Cr.Dn and
Cr.Le accumulation at a given bone site for a human proximal femur. Since the
proposed NN model incorporates the modeling of the real scanned 3D trabecular
networks, stress concentration due to the trabecular shapes and connections can be