BoneDoc

BoneDoc is a modular C++ backend service for 3D shape analysis of human bones. Developed as part of my diploma thesis, it supports the classification of bone morphology based on anatomical features of the femur, humerus, and tibia. It run as a containerized microservice and is integrated into BoneHost, a lightweight web framework for accessing backend services.

The diploma thesis included a study aimed at identifying morphological differences between ethnic groups (Asian and Caucasian) through multivariate analysis of 3D bone structures. To support this study, BoneDoc was developed as a research tool to automate anatomical parameter extraction and statistical classification using logistic regression. The results of the study showed that population-specific traits could be detected with high accuracy based on bone morphology - especially for femur and humerus bones.

The analysis was based on 296 3D-reconstructed bone surfaces (from CT scans), divided into:

• Femur: 132 samples (82 Asian, 50 Caucasian)
• Humerus: 92 samples (53 Asian, 39 Caucasian)
• Tibia: 72 samples (48 Asian, 24 Caucasian)

All meshes were preprocessed and registered using open-source tools to ensure point-wise correspondence across the dataset.

BoneDoc extracts anatomical parameters known from the literature to be relevant for population classification, including:

• Bone length
• Shaft widths (mediolateral and anteroposterior)
• Medial and lateral offsets
• Angles such as neck-shaft inclination, anteversion, retroversion, and torsion (depending on the bone)

A logistic regression model was trained on this dataset to predict the ethnic group. Cross-validation and statistical tests confirmed the robustness of the results - with some limitations for the tibia, due to fewer discriminative features.

BoneDoc enables:

• Automatic extraction of anatomical features from new 3D bone models
• Classification into Asian or Caucasian groups using trained statistical models
• Prototyping and testing of shape analysis methods in a modular environment
• Open-Source use and further development ¹⁾

Although the tool was developed in an academic context, it may serve as a foundation for further work in medical image analysis, population-based studies, or implant design.