rtificial intelligence (AI) algorithms have existed for decades and have recently been propelled to the forefront of medical imaging research. To a large extent, this is related to improvements in computing power, availability of a large amount of training data, and innovative and improved neural network architectures, with the recognition that certain types of algorithms are well suited to image analysis. The latter discovery was accelerated by the ImageNet competition and represents a fundamental transformation in research mechanics and methods in computer vision.Currently, in most studies, researchers collect data, perform analysis, and publish results. The same researchers may continue to augment and expand the data set and perform subsequent analysis with resulting publications. The data for each study are held quite closely and are rarely shared among institutions outside of multicenter trials. Competitions represent a different model of research: Research data are made available to the public, usually with a baseline performance metric. Groups around the world are invited to analyze the data and create algorithms to beat the performance of the prior generation. For example, the baseline performance metric for this challenge was set by the previous skeletal age model developed by Larson et al (1).The Radiological Society of North America (RSNA) Pediatric Bone Age Machine Learning Challenge was created to evaluate the performance of computer algorithms in executing a common image analysis activity that is familiar to many pediatric radiologists: estimating the bone age of pediatric patients based on radiographs of their hand (1-5). This challenge used a data set of pediatric
This work extends PneumoCAD, a Computer-Aided Diagnosis system for detecting pneumonia in infants using radiographic images [1], with the aim of improving the system's accuracy and robustness. We implement and compare five contemporary machine learning classifiers, namely: Naïve Bayes, K-Nearest Neighbor (KNN), Support Vector Machines (SVM), Multi-Layer Perceptron (MLP) and Decision Tree, combined with three dimensionality reduction algorithms: Sequential Forward Selection (SFS), Principal Component Analysis (PCA) and Kernel Principal Component Analysis (KPCA). Current results demonstrate that Naïve Bayes classifier combined with KPCA produces the best overall results.
Supervised learning with irregularly sampled time series have been a challenge to Machine Learning methods due to the obstacle of dealing with irregular time intervals. Some papers introduced recently recurrent neural network models that deals with irregularity, but most of them rely on complex mechanisms to achieve a better performance. This work propose a novel method to represent timestamps (hours or dates) as dense vectors using sinusoidal functions, called Time Embeddings. As a data input method it and can be applied to most machine learning models. The method was evaluated with two predictive tasks from MIMIC III, a dataset of irregularly sampled time series of electronic health records. Our tests showed an improvement to LSTM-based and classical machine learning models, specially with very irregular data.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.