Abstract. The majority of current methods in object classification use the one-against-rest training scheme. We argue that when applied to a large number of classes, this strategy is problematic: as the number of classes increases, the negative class becomes a very large and complicated collection of images. The resulting classification problem then becomes extremely unbalanced, and kernel SVM classifiers trained on such sets require long training time and are slow in prediction. To address these problems, we propose to consider the negative class as a background and characterize it by a prior distribution. Further, we propose to construct "hybrid" classifiers, which are trained to separate this distribution from the samples of the positive class. A typical classifier first projects (by a function which may be non-linear) the inputs to a one-dimensional space, and then thresholds this projection. Theoretical results and empirical evaluation suggest that, after projection, the background has a relatively simple distribution, which is much easier to parameterize and work with. Our results show that hybrid classifiers offer an advantage over SVM classifiers, both in performance and complexity, especially when the negative (background) class is large.
Advanced medical imaging algorithms (such as bone removal, vessel segmentation, or a lung nodule detection) can provide extremely valuable information to the radiologists, but they might sometimes be very time consuming. Being able to run the algorithms in advance can be a possible solution. However, we do not know which algorithm to run on a given dataset before it is actually used. It is possible to manually insert matching rules for preprocessing algorithms, but it requires high maintenance and does not work well in practice. This paper presents a dynamic machine learning solution for predicting which advanced visualization (AV) algorithm needs to be applied on a given series. The system gets a handful of free text DICOM tags as an input and builds a model in the clinical setting. It incorporates a Bag of Words (BOW) feature extractor and a Random Forest classifier. The approach was tested on two datasets from clinical sites which use different languages and varying scanner models. We show that even without feature extraction, sensitivity of above 90% can be reached on both of them. By using BOW feature extractor, precision and sensitivity can usually be further improved. Even on a noisy and highly unbalanced dataset, only around 100 samples were needed to reach sensitivity of above 80% and specificity of above 97%. We show how the solution can be part of a Smart Preprocessing mechanism in a viewing software. Using such a system will ultimately minimize the time to launch studies and improve radiologists reading time efficiency.
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.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.