Predictive modeling of compressive strength for additively manufactured PEEK spinal fusion cages using machine learning techniques

Sivakumar, NK, Palaniyappan, S, Bodaghi, M ORCID: https://orcid.org/0000-0002-0707-944X, Azeem, PM, Nandhakumar, GS, Basavarajappa, S, Pandiaraj, S and Hashem, MI, 2024. Predictive modeling of compressive strength for additively manufactured PEEK spinal fusion cages using machine learning techniques. Materials Today Communications. ISSN 2352-4928

Text 1862189_Bodaghi.pdf - Post-print Full-text access embargoed until 6 February 2025. Download (680kB)

Abstract

The current study delves into the utilization of Machine Learning (ML) algorithms to evaluate the mechanical properties of additively manufactured PEEK spinal fusion cages. In this research, a range of ML models, including Linear Regression (LiR), Lasso Regression (LaR), Decision tree (DT), and K-Nearest Neighbor (KNN) are harnessed to enhance compressive strength prediction. Ensemble learning techniques such as bagging, boosting, and stacking are applied to identify the most accurate ML model in terms of achieving heightened accuracy and minimized errors. To facilitate this, spinal fusion cages are 3D printed using the Fused Filament Fabrication (FFF) technique and subsequently tested using a Universal Testing Machine (UTM). The development of ML models involves the exploration of independent material-extrusion factors, encompassing layer height (0.1 mm, 0.2 mm, 0.3 mm), printing temperature (370℃, 390℃, 410℃), printing speed (10 mm/sec, 30 mm/sec, 50 mm/sec), infill density (40%, 70%, 100%), build orientation (0º, 45º, 90º), and line width (0.1 mm, 0.2 mm, 0.3 mm). The robustness and effectiveness of the developed ML models in predicting compressive strength properties are optimized through comprehensive error metric analysis. The results indicate that the LiR model, particularly when implemented under the boosting ensemble technique, demonstrates the highest accuracy with a Mean Absolute Error (MAE) of 0.657, Root Mean Square Error (RMSE) of 0.758, and Median Absolute Error (MedAE) of 0.634. This underscores the potential of LiR for precise compressive strength prediction in 3D-printed PEEK spinal fusion cages for spinal and maxillomandibular reconstruction.

Item Type:	Journal article
Publication Title:	Materials Today Communications
Creators:	Sivakumar, N.K., Palaniyappan, S., Bodaghi, M., Azeem, P.M., Nandhakumar, G.S., Basavarajappa, S., Pandiaraj, S. and Hashem, M.I.
Publisher:	Elsevier BV
Date:	6 February 2024
ISSN:	2352-4928
Identifiers:	Number Type 10.1016/j.mtcomm.2024.108307 DOI 1862189 Other
Divisions:	Schools > School of Science and Technology
Record created by:	Laura Ward
Date Added:	13 Feb 2024 11:55
Last Modified:	13 Feb 2024 11:55
URI:	https://irep.ntu.ac.uk/id/eprint/50862

Actions (login required)

Edit View

Statistics

Views

Downloads