Bookstore

This textbook presents deep learning models and their healthcare applications. It focuses on rich health data and deep learning models that can effectively model health data. Healthcare data: Among all healthcare technologies, electronic health records (EHRs) had vast adoption and a significant impact on healthcare delivery in recent years. One crucial benefit of EHRs is to capture all the patient encounters with rich multi-modality data. Healthcare data include both structured and unstructured information. Structured data include various medical codes for diagnoses and procedures, lab results, and medication information. Unstructured data contain 1) clinical notes as text, 2) medical imaging data such as X-rays, echocardiogram, and magnetic resonance imaging (MRI), and 3) time-series data such as the electrocardiogram (ECG) and electroencephalogram (EEG). Beyond the data collected during clinical visits, patient self-generated/reported data start to grow thanks to wearable sensors' increasing use.  The authors  present deep learning case studies on all data described.

Deep learning models: Neural network models are a class of machine learning methods with a long history. Deep learning models are neural networks of many layers, which can extract multiple levels of features from raw data. Deep learning applied to healthcare is a natural and promising direction with many initial successes. The authors cover deep neural networks, convolutional neural networks, recurrent neural networks, embedding methods, autoencoders, attention models, graph neural networks, memory networks, and generative models. It's presented with concrete healthcare case studies such as clinical predictive modeling, readmission prediction, phenotyping, x-ray classification, ECG diagnosis, sleep monitoring, automatic diagnosis coding from clinical notes, automatic deidentification, medication recommendation, drug discovery (drug property prediction and molecule generation), and clinical trial matching.

This textbook targets graduate-level students focused on deep learning methods and their healthcare applications. It can be used for the concepts of deep learning and its applications as well. Researchers working in this field will also find this book to be extremely useful and valuable for their research.

Introduction to Deep Learning for Healthcare

ContentsI IntroductionI.1 Who should read this book?I.2 Book organizationII Health DataII.1 The growth of EHR AdoptionII.2 Health DataII.2.1 Life cycle of health dataII.2.2 Structured Health DataII.2.3 Unstructured clinical notesII.2.4 Continuous signalsII.2.5 Medical Imaging DataII.2.6 Biomedical data for in silico drug Discovery II.3 Health Data StandardsIII Machine Learning BasicsIII.1 Supervised LearningIII.1.1 Logistic RegressionIII.1.2 Softmax RegressionIII.1.3 Gradient DescentIII.1.4 Stochastic and Minibatch Gradient DescentIII.2 Unsupervised LearningIII.2.1 Principal component analysisIII.2.2 t-distributed stochastic neighbor embedding (t-SNE)III.2.3 ClusteringIII.3 Assessing Model PerformanceIII.3.1 Evaluation Metrics for Regression TasksIII.3.2 Evaluation Metrics for Classification TasksIII.3.3 Evaluation Metrics for Clustering TasksIII.3.4 Evaluation StrategyIII.4 Modeling ExerciseIII.5 Hands-On Practice34 CONTENTSIVDeep Neural Networks (DNN)IV.1 A Single neuronIV.1.1 Activation functionIV.1.2 Loss FunctionIV.1.3 Train a single neuronIV.2 Multilayer Neural NetworkIV.2.1 Network RepresentationIV.2.2 Train a Multilayer Neural NetworkIV.2.3 Summary of the Backpropagation AlgorithmIV.2.4 Parameters and Hyper-parametersIV.3 Readmission Prediction from EHR Data with DNNIV.4 DNN for Drug Property PredictionV EmbeddingV.1 OverviewV.2 Word2VecV.2.1 Idea and Formulation of Word2VecV.2.2 Healthcare application of Word2VecV.3 Med2Vec: two-level embedding for EHRV.3.1 Med2Vec MethodV.4 MiME: Embed Internal StructureV.4.1 Notations of MIMEV.4.2 Description of MIMEV.4.3 Experiment results of MIMEVI Convolutional Neural Networks (CNN)VI.1 CNN intuitionVI.2 Architecture of CNNVI.2.1 Convolution layer - 1DVI.2.2 Convolution layer - 2DVI.2.3 Pooling LayerVI.2.4 Fully Connected LayerVI.3 Backpropagation Algorithm in CNN*VI.3.1 Forward and Backward Computation for 1-D DataVI.3.2 Forward Computation and Backpropagation for 2-D ConvolutionLayer . VI.3.3 Special CNN ArchitectureVI.4 Healthcare Applications VI.5 Automated surveillance of cranial images for acute neurologic eventsVI.6 Detection of Lymph Node Metastases from Pathology ImagesVI.7 Cardiologist-level arrhythmia detection and classification in ambulatoryECGCONTENTS 5VIIRecurrent Neural Networks (RNN)VII.1Basic Concepts and NotationsVII.2Backpropagation Through Time (BPTT) algorithmVII.2.1Forward PassVII.2.2 Backward PassVII.3RNN VariantsVII.3.1 Long Short-Term Memory (LSTM)VII.3.2 Gated Recurrent Unit (GRU)VII.3.3 Bidirectional RNNVII.3.4 Encoder-Decoder Sequence-to-Sequence ModelsVII.4Case Study: Early detection of heart failureVII.5Case Study: Sequential clinical event predictionVII.6Case Study: De-identification of Clinical NotesVII.7Case Study: Automatic Detection of Heart Disease from electrocardiography(ECG) DataVIIAIutoencoders (AE)VIII.1OverviewVIII.2AutoencodersVIII.3Sparse AutoencodersVIII.4Stacked AutoencodersVIII.5Denoising AutoencodersVIII.6Case Study: "Deep Patient" via stacked denoising autoencodersVIII.7Case Study: Learning from Noisy, Sparse, and Irregular ClinicaldataIX Attention ModelsIX.1 OverviewIX.2 Attention MechanismIX.2.1 Attention based on Encoder-Decoder RNN ModelsIX.2.2 Case Study: Attention Model over Longitudinal EHRIX.2.3 Case Study: Attention model over a Medical OntologyIX.2.4 Case Study: ICD Classification from Clinical NotesX Memory NetworksX.1 Original Memory NetworksX.2 End-to-end Memory NetworksX.3 Case Study: Medication RecommendationX.4 EEG-RelNet: Memory Derived from DataX.5 Incorporate Memory from Unstructured Knowledge BaseXIGraph Neural NetworksXI.1 OverviewXI.2 Graph Convolutional NetworksXI.2.1 Basic Setting of GCNXI.2.2 Spatial Convolution on Graphs6 CONTENTSXI.2.3 Spectral Convolution on GraphsXI.2.4 Approximate Graph ConvolutionXI.2.5 Neighborhood AggregationXI.3 Neural Fingerprinting: Drug Molecule Embedding with GCNXI.4 Decagon: Modeling Polypharmacy Side Effects with GCNXI.5 Case Study: Multiview Drug-drug Interaction PredictionXIIGenerative ModelsXII.1Generative adversarial networks (GAN)XII.1.1 The GAN FrameworkXII.1.2 The Cost Function of DiscriminatorXII.1.3 The Cost Function of GeneratorXII.2Variational Autoencoders (VAE)XII.2.1 Latent Variable ModelsXII.2.2Objective FormulationXII.2.3Objective ApproximationXII.2.4 Reparameterization TrickXII.3Case Study: Generating Patient RecordsXII.4Case Study: Small Molecule Generation for Drug DiscoveryXII CIonclusionXIII.1Model SetupXIII.2Model TrainingXIII.3Testing and Performance EvaluationXIII.4Result VisualizationXIII.5Case StudiesXIVAppendixXIV.1Regularization*XIV.1.1Vanishing or Exploding Gradient ProblemXIV.1.2DropoutXIV.1.3Batch normalizationXIV.2Stochastic Gradient Descent and Minibatch gradient descent*XIV.3Advanced optimization*XIV.3.1MomentumXIV.3.2Adagrad, Adadelta, and RMSpropXIV.3.3Adam