Dissertation Ivar Mjåland Salte
On May 28th 2024, Cand.med. Ivar Mjåland Salte defended his thesis "Artificial intelligence to improve measurement reproducibility of left ventricular function in echocardiography" for the degree of PhD at the Faculty of Medicine, Institute of Clinical Medicine, University of Oslo. The trial lecture was titled "Overcoming Bias in Artificial Intelligence for Equitable Healthcare".
Photo: Tore Salte
Public Defence: Ivar Mjåland Salte - Institute of Clinical Medicine
Adjudication committee
- First opponent: Professor Partho P. Sengupta, Rutgers Robert Wood Johnson Medical School, United States
- Second opponent: Professor Maja-Lisa Løchen, UiT - The Artic University of Norway, Norway
- Third member and chair of the evaluation committee: Associate professor John-Peder Escobar Kvitting, University of Oslo
Chair of the Defence
- Professor Emeritus Odd Geiran, University of Oslo
Principal Supervisor
- Professor II Thor Edvardsen, University of Oslo
Summary
The left ventricle plays a pivotal role as a dynamic pump supplying blood to the systemic circulation under pressure. Accurate assessment of left ventricular function is crucial in diagnosing and managing patients with various cardiovascular conditions. Ultrasound imaging of the heart (echocardiography) is cost-effective, easily accessible and harmless, and therefore a preferred diagnostic tool for assessing left ventricular function.
Left ventricular Global longitudinal strain (GLS) is a key measurement for quantification of left ventricular function, reflecting the longitudinal shortening of the left ventricular wall during the cardiac cycle. GLS is traditionally performed using a semi-automatic computer program, in which the operator annotates the relevant images and defines the region of interest. Such semi-automatic measurements can be time-consuming, as there is often a need for several manual adjustments. Moreover, these manual adjustments may cause clinically significant measurement variability between different operators.
In this thesis, Salte and colleagues developed and explored the use of novel artificial intelligence (AI) methods, including deep learning technologies, for fast and fully automated measurements of left ventricular GLS in echocardiography.
The results demonstrated the technical feasibility of employing AI for automated GLS measurements in echocardiography. The AI method surpassed the accuracy of a traditional computer algorithm, when estimating motion in ultrasound images. AI-based GLS measurements displayed strong agreement with one of the most widely used and clinically available semi-automatic programs. Moreover, AI-based measurements demonstrated superior reproducibility compared to two different operators using a semi-automatic method.
In summary, AI could successfully measure left ventricular GLS in ultrasound images, resulting in efficient measurements and improved reproducibility compared to conventional semi-automatic methods relying on human input.