Machine Learning

Comparative Study of Machine Learning Techniques on Musculoskeletal Abnormality Detection

A growing demand and public expectation for radiology services have been revealed by recent studies. As a result, problems such as mismatch between human labor and demand as well as potential hazards in management have been observed in hospitals. This challenge can be potentially tackled by introducing machine learning techniques to augment human diagnosis. This project is motivated to produce a holistic comparison among the existing machine learning solution paradigms in detecting abnormal radiographics. The result of this study can serve as a reference guide for researchers to evaluate existing approaches and design new solutions according to constraints in resources, time and expectation on performance.

This project applied various machine learning techniques and compared their performance in the context of classifying and detecting abnormal musculoskeletal radiographs. 3 paradigms of machine learning methods have been experimented:

  1. Classic machine learning models such as Support Vector Machine (SVM) fed by features extracted by convolutional autoencoder. This approach achieves an accuracy of 0.598.
  2. A semi-supervised anomaly detection method where an convolutional autoencoder-decoder is trained to reconstruct only the normal images and detect abnormal images with high reconstruction error. This approach turns out to be ineffective, possibly due to subtle differences between normal and abnormal radiographs with low resolution.
  3. Deep learning models including VGGnet, ResNet and DenseNet. These models demonstrate high capability in tackling the problem and manage to achieve an highest accuracy of 0.8365.