Comparative analysis of machine learning algorithms for computer-assisted reporting based on fully automated cross-lingual RadLex mappings

Abstract

Computer-assisted reporting (CAR) tools were suggested to improve radiology report quality by context-sensitively recommending key imaging biomarkers. However, studies evaluating machine learning (ML) algorithms on cross-lingual ontological (RadLex) mappings for developing embedded CAR algorithms are lacking. Therefore, we compared ML algorithms developed on human expert-annotated features against those developed on fully automated cross-lingual (German to English) RadLex mappings using 206 CT reports of suspected stroke. Target label was whether the Alberta Stroke Programme Early CT Score (ASPECTS) should have been provided (yes/no:154/52). We focused on probabilistic outputs of ML-algorithms including tree-based methods, elastic net, support vector machines (SVMs) and fastText (linear classifier), which were evaluated in the same 5 × fivefold nested cross-validation framework. This allowed for model stacking and classifier rankings. Performance was evaluated using calibration metrics (AUC, brier score, log loss) and -plots. Contextual ML-based assistance recommending ASPECTS was feasible. SVMs showed the highest accuracies both on human-extracted- (87%) and RadLex features (findings:82.5%; impressions:85.4%). FastText achieved the highest accuracy (89.3%) and AUC (92%) on impressions. Boosted trees fitted on findings had the best calibration profile. Our approach provides guidance for choosing ML classifiers for CAR tools in fully automated and language-agnostic fashion using bag-of-RadLex terms on limited expert-labelled training data.

[Retracted] Traffic State Detection Based on Multidimensional Data Fusion System of Internet of Things

“…Positioning system (GPS) is an interdependent assembly, receiving equipment or positioning components to determine the position of the system, and has three parts: space components a GPS constellation, Ground Part-Ground Monitoring System, and user equipment part a GPS signal receiver. Various positioning systems greatly improve the working efficiency of the Internet of Things and provide interoperable communication for frequent activities of various industries through precise positioning information[29,30] Cloud Computing Technology. Cloud computing mainly takes the Internet as the center; establishes large capacity storage servers; provides safe, fast, and convenient network services on the basis of open standards and services; and makes the "cloud" of the Internet become the data center and computing center of all walks of life (IPv6.…”

Large-Scale Standardized Image Integration for Secondary Use Research Projects

“…The use of NLP methods (R3) had to be stalled, as anonymization is a particular challenge, especially in the free-text findings. The reports for the image series were accordingly imported and registered in the index, but the extraction into the archetypes "Imaging examination result" [10] for the staging in the openEHR repository with the aid of NLP methods [11] is the next integration goal. With respect to data privacy (R8), the UKSH has already a broad consent in use since 2017 allowing re-use of clinical data for research [12].…”

Lightweight Visual Transformers Outperform Convolutional Neural Networks for Gram-Stained Image Classification: An Empirical Study

“…Tanwani et al from Google [58] proposed RepsNet which generates automated medical reports in natural language from medical images. Regardless of the advanced architectures of novel deep learning models such as VTs, simple statistical methods or shallow ML algorithms often outperform these models or offer at least a sufficient enough performance, especially on limited, medical domain-specific tasks as demonstrated for anomaly detection in neuroimaging [59] and cross-lingual radiological report classification [60]. Deploying lightweight DL models to an augmented reality (AR) device [61] also has promising applications.…”