Machine Learning with Clinical Data Leila Mureebe Machine

Machine Learning with Clinical Data Leila Mureebe

Machine Learning • Use of statistical techniques to give computers the ability to progressively improve performance on a specific task with data, without being explicitly programmed

Machine Learning Techniques

Clinical Problem • Patient in CT ICU after complicated cardiac procedure • Both feet are cool • Is a consult needed?

Clinical Problem • Two samples • How would you classify them?

Survey administered to providers Various roles Various years in position Accuracy 57. 7% +/- 9. 9% Clinical Problem

Classification Problem • In machine learning and statistics, classification is the problem of identifying to which of a set of categories (sub-populations) a new observation belongs, on the basis of a training set of data containing observations (or instances) whose category membership is known

In machine learning and statistics, classification is the problem of identifying to which category an observation belongs to Classification This is based on the basis of a training set of data where the category is known This is a form of supervised learning

Goal: Create an algorithm to accurately classify continuous wave Doppler signals Initially - 2 classifiers Our Project Future - 3 classifiers

Curated data set Requirements Data curation is a broad term used to indicate processes and activities related to the organization and integration of data collected from various sources, annotation of the data, and publication and presentation of the data such that the value of the data is maintained over time, and the data remains available for reuse and preservation

100 patients undergoing ABIs 5 samples per patient Our Data Set Official report • ABI • Phasicity • Plethysmograph

Linear Classifiers: Logistic Regression, Naive Bayes Classifier Support Vector Machines Decision Trees Classification Algorithms Boosted Trees Random Forest Neural Networks Nearest Neighbor

Features Examined Temporal Frequency Rise time Bandwidth (99%) Decay time Mean frequency Mean time Peak frequency (high) Diastolic level Peak frequency (low) Systolic level Total harmonic distortion Total peak height Power

• A receiver operating characteristic curve, is a graphical plot that illustrates the diagnostic ability of a binary classifier system as its discrimination threshold is varied ROC Curve

support vector machine constructs a hyperplane or set of hyperplanes in a high- or infinite-dimensional space, which can be used for classification, regression, or other tasks like outliers detection[3]. Best Performer Intuitively, a good separation is achieved by the hyperplane that has the largest distance to the nearest training-data point of any class, since in general the larger the margin the lower the generalization error of the classifier

• H 1 does not separate the classes. H 2 does, but only with a small margin. H 3 separates them with the maximum margin SVM Theory

Survey to assess human accuracy Additional Studies Comparison of extracted audio to plethysmogram Extrapolation of CTA results to audio

Other Example • Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs • Varun Gulshan, Ph. D 1; Lily Peng, MD, Ph. D 1; Marc Coram, Ph. D 1; et al • JAMA. 2016; 316(22): 2402 -2410. doi: 10. 1001/jama. 2016. 17216