Deep learning-based Scapular morphology assessment pipeline for glenoid segmentation and landmarks localization

by Kuan Liu et al.

Introduction

This repository is the code implementation for our paper "Deep Learning-Based Scapular Morphology Assessment Pipeline for Glenoid Segmentation and Landmark Localization".

We developed two deep learning models to segment the glenoid shape and localize five key scapular landmarks (trigonum spinae (TS), angulus inferior (AI), processus coracoideus (PC), acromion (AC), and angulus acromialis (AA)), respectively.

Get Started

This repository is based on Python 3.11.5 + PyTorch 2.1.0+cu118 + MONAI 1.3.0.

Training WorkFlow

1. Dataset preperation.

We provide dataset generation scripts in ./dataset folder.

2. Train the 2D Glenoid Segmentation Model.

python main_2d.py

3. Train the 3D Scapular Landmark Localization Model.

python main_3d.py

Inference WorkFlow

We provide inference scripts for both the 2D segmentation and 3D landmark localization models:

python test_2d.py

python test_3d.py

We also develop 3D Slicer plugin for scapular morphology assessment integrating both inference models Slicer-Scapular-Morphology.

Pretrained model weights and sample data can be downloaded from https://github.com/liukuan5625/SlicerScapularMorphology/releases.