Medical-Image-Processing

This repository contains some works that I have done in Computation Radiology Lab UR Medical Center.

It mainly contains the following parts:

• Phase contrast X-ray computed tomography (PCI-CT) deep learning classifier.
• Vertebrae Segmentation in Pathological Spine CT via Fully Convolutional Neural Network.
• Preprocessing of Raw image data (.mhd/.raw/.dicom) of Annotated lymph node CT dataset and Microsoft's Spine Web dataset

Keypoints of PCT-CT deep learning project

• This project got accepted at SPIE2017 conference for oral representation. You can check out our publication here. A refined version of this paper is uploaded here.

• It makes fully use of CaffeNet and Inception-v3 Net to classify CT samples. Please check out Deep_Transfer_Learning_PCI_CT.pdf for full detail.

• Some result is shown below:

  The classification result we got from a fine-tuned CaffeNet.

  Keypoints of Vertebrea Segmentation project

• Preprocessing raw spine dataset (About 20GB), you can check out the sample spine CT scan before and after preprocessing below.

• The preprocessing techniques we have explored include: Histogram Equalization, Adaptive Histogram Equalization and Constrast Stretching

  

  

  For full detail, please refers to our report here.

• Prepared the lmdb file so that the data can be directly fed into Convolutional Neural Network.

• Used fully convolutional neural network to segment the image.