Semi-supervised classification of anatomy on orthopedic X-ray data

1 Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany.

Motivation

• Manual surgery planning/verification is complex, prone to errors, and non-sterile.
• Clinical need for (semi)-automatic guidance.

Research objective

Development of a classification algorithm that 1) infers the type of anatomy (and 2) the viewing direction) from a 2D X-ray image.

Result

In this project, I implement different architectures of machine learning models such as Densenet, Pre-Trained Densnet Models on imageNet and pre-trained vgg16 models on imageNet successfully. The result of classification is really good since I get the accuracy- over 90 percentages.