Medical Image Analysis with Deep Learning

Introduction

Processing and analysis of medical images; a rapidly growing industry expected to reach ~$4 billion by 2020. Recent advances in deep learning are helping to identify, classify, and quantify patterns in medical images. Deep Learning helps to exploit hierarchical feature representations learned exclusively from data without proper domain-specific knowledge. Deep learning is rapidly becoming the state of the art in numerous medical applications. This two days training will cover basic image processing techniques, different methods of features extractions, deep learning techniques (Autoencoders, CNN, RNN), and its application to Medical Image analysis (X-ray, OCT, Retinal Images, Brain Images, etc.). This training helps to start a career in the field of medical imaging analysis using Classical techniques and Deep Learning. You can use these skills in order to develop newer technological innovations and regularize them for high-throughput clinical translation and usage.

Course Objectives:

  • Different Medical Image Modalities (X-rays, Magnetic Resonance, Ultrasonic, etc.)
  • Basic Image operation with Python
  • Texture in Medical Images and Classical Feature extraction
  • Neural Network, Autoencoder (Sparse and de-noising)
  • Deep Learning with Convolutional Neural Network (CNN)
  • Application of Deep Learning to Medical Image Analysis

Duration

7 hours, 1 Day Course

Mode of Delivery

Classroom-based, Instructor-led Training

Course Outline

  1. Introduction to Medical Images
    1. X-ray and CT Imaging
    2. Magnetic Resonance Imaging
    3. Ultrasound Imaging
    4. Optical Microscopy and Molecular Imaging
  2. Basic Image operation with Python
    1. Read and Display Image
    2. Covert Colour Image to Grayscale Image
    3. Cropping and Resizing an Image
    4. Rotating Image
    5. Histogram Equalization
    6. Blurring an Image
  3. Texture in Medical Images
    1. Texture characterization – Statistical vs Structural
    2. Co-occurrence Matrix
    3. Orientation Histogram
    4. Local Binary Pattern (LBP)
    5. Texture from Fourier features
    6. Wavelets
    7. Feature extractions for Image (Medical/General)
  4. Neural Network for Visual Computing
    1. Simple Neuron
    2. Neural Network formulation
    3. Learning with Error Propagation
    4. Gradient Checking and Optimization
  5. Deep Learning
    1. What is Deep Learning?
    2. Families of Deep Learning
    3. Multilayer Perceptron
    4. Learning Rule
    5. Autoencoders
    6. Retinal Vessel Detection using Autoencoders
  6. Stacked, Sparse, Denoising Autoencoders
    1. Stacking Autoencoders
    2. Ladder wise pre-training and End-to-End Pre-training
    3. Denoising and Sparse Autoencoders
    4. Ladder Training
    5. End-to-End Training
    6. Medical Image classification with Stacked Autoencoders
  7. Convolutional Neural Network (ConvNet)
    1. What is ConvNet?
    2. Difference between Fully connected NN and ConvNet.
    3. Stride, Padding, and Pooling
    4. Deconvolution
    5. ReLU Transfer Function
  8. Image Classification with CNN
    1. Convolutional Autoencoder
    2. LeNet for Image Classification
    3. AlexNet for Image Classification
  9. Improving Deep Neural Network
    1. Batch Normalization, Dropout
    2. Tuning Hyper-parameters to improve performance of NN.
    3. Learning Rate Annealing
    4. Different Cost Functions
  10. Deep CNN and its application to Medical Images
    1. Vgg16, ResNet34, GooleNet, and DenseNet121
    2. Transfer Learning
    3. Pneumonia detection from Chest X-rays with Deep CNN.
    4. White blood cell classification with CNN
  11. Object Localization
    1. Activation pooling for object localization
    2. Region proposal Network
    3. Sematic segmentation
    4. UNet
    5. Retinopathy Image segmentation with UNet
  12. Spatio-Temporal Deep Learning
    1. Understanding Video analysis
    2. Recurrent Neural Network (RNN)
    3. Long Short Term Memory (LSTM)
    4. Activity recognition using 3D-CNN
    5. Analysis of Brain Images