University of Twente Student Theses


Automated diagnosis of shock in Malawian children using point-of-care cardiac ultrasound

Bock, E.J.E. de (2022) Automated diagnosis of shock in Malawian children using point-of-care cardiac ultrasound.

Full text not available from this repository.

Full Text Status:Access to this publication is restricted
Abstract:Rationale: Shock, or circulatory insufficiency, is one of the most life-threatening conditions, especially in low-income countries such as Malawi. Methods for early detection of shock are lacking. Point-of-care ultrasound (US) might help to diagnose shock, but image interpretation is difficult, which delays shock recognition. Objectives: The study aimed to accelerate and improve the diagnostics of shock in Malawian children using US. The goal of this study was to investigate the possibilities of the automated classification of cardiac images of the paediatric Rapid US in Shock (p-RUSH) exam. Methods: Cardiac p-RUSH images of Malawian children with (n=30) and without shock (n=102) were used. The machine learning (ML) methods logistic regression, decision tree (DT), random forest and support vector machine (SVM) were explored, and their use in shock classification based on echocardiographic features was evaluated. Interobserver variability of manual cardiac function measurements was determined. Automatic measurements’ feasibility and similarity to manual measurements were evaluated. Results: Ejection fraction (EF), fractional shortening (FS) and stroke volume index (SVI) were a relevant set of shock classifiers. ML models were able to detect shock in the training set, but not in the test set. Interobserver variability for manual EF and FS was high. Automatic EF measurements could be performed in half of the patients, but showed poor to moderate correlation with manual EF and high bias. Conclusions: In the studied setting, ML-based classification algorithms theoretically have the potential to accelerate and improve shock diagnostics, but the heterogeneity of the study population calls for a larger sample size to further investigate this. The use of automated cardiac function measurements is currently limited by poor image quality. In future work, DT and SVM classification algorithms based on (automatically computed) EF, FS and SVI are promising to explore in order to improve early detection of shock, which might guide treatment and increase survival in Malawian children in shock.
Item Type:Essay (Master)
Faculty:TNW: Science and Technology
Subject:44 medicine
Programme:Technical Medicine MSc (60033)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page