Introduction

---

• Algorithms can be used to detect disease in chest X-rays.

Visualisation

---

• In NumPy, RGB images are usually stored as 3-dimensional arrays.

Data preparation

---

• Data augmentation can help to avoid overfitting.

Neural networks

---

• Dense layers, also known as fully connected layers, are an important building block in most neural network architectures. In a dense layer, each neuron is connected to every neuron in the preceeding layer.
• Dropout is a method that helps to prevent overfitting by temporarily removing neurons from the network.
• The Rectified Linear Unit (ReLU) is an activation function that outputs an input if it is positive, and outputs zero if it is not.
• Convolutional neural networks are typically used for imaging tasks.

Training and evaluation

---

• During the training process we iteratively update the model to minimise error.

Explainability

---

• Saliency maps are a popular form of explainability for imaging models.
• Saliency maps should be used cautiously.