Semantic Segmentation of Micro-CT Images to Analyze Bone Ingrowth into Biodegradable Scaffolds

Area of Science:

• Biomaterials Science
• Medical Imaging
• Orthopedic Surgery

Background:

• Large bone fractures often result in non-union, necessitating surgical intervention.
• Current treatments like autografting have limitations, driving the need for alternative solutions.
• Biodegradable scaffolds offer a promising alternative for promoting bone regeneration.

Purpose of the Study:

• To develop and validate a novel 3-stage pipeline for accurate quantification of bone ingrowth into biodegradable scaffolds using micro-computed tomography (μ-CT).
• To improve the accuracy of differentiating bone, scaffold, and pores in μ-CT images for better analysis of bone regeneration.
• To quantitatively assess bone ingrowth and scaffold degradation for evaluating scaffold efficacy.

Main Methods:

• A 3-stage computational pipeline was developed, starting with a convolutional neural network for image segmentation.
• The pipeline incorporates 3D rigid image registration to define regions of interest (VOI).
• Custom algorithms were implemented for quantitative analysis of bone ingrowth and scaffold degradation.

Main Results:

• The convolutional neural network achieved high segmentation accuracy, with a Dice similarity coefficient of 90.1%, Intersection over Union of 83.9%, and pixel accuracy of 93.1% on unseen data.
• The developed pipeline enables precise quantification of bone ingrowth and scaffold degradation.
• The methodology provides a significant improvement over existing segmentation techniques for μ-CT analysis.

Conclusions:

• The proposed 3-stage pipeline offers a robust and accurate method for analyzing bone ingrowth into biodegradable scaffolds using μ-CT.
• This approach enhances the evaluation of bone regeneration, crucial for developing effective treatments for large bone defects.
• The improved quantification accuracy supports the advancement of biomaterial-based therapies for non-union fractures.