How does the technical status of medical ultrasound equipment affect image quality? Studies based on human observer experiments and a novel method for automatic detection of defective transducers

Sammanfattning: Ultrasound as an imaging modality has a fundamental role in modern healthcare. The equipment, especially transducers, is fragile and handled daily, and transducers are known to sometimes become defective. The overall aim of this thesis was to evaluate how the technical status of the equipment affects clinical image quality. In study 1 a novel automatic method to detect defects in linear transducers by analysing the images that are saved for documentation was developed which makes it possible to monitor many transducers without interference with clinical activity. The method was evaluated in study 2 by comparing the result obtained with the automatic method to the well-established in-air method, which is used for quality control, for 81 linear and curvilinear transducers. The methods showed good agreement, the area under curve for the receiver operating characteristics evaluation being 0.88 (SD 0.06). There is a variety of different scanners, from regular to high-end. In study 3 a comparison of a regular and a high-end scanner was made, the images from the clinically relevant objects in a greyscale phantom being assessed by six observers in a detection study. Visibility was substantially better for the advanced scanner in four out of 16 combinations of depth, contrast and size. The method from study 1 was also used in study 4 to find images from defective transducers that had been used clinically. 160 images from four defective transducers, with defects that had different grades of severity, were collected, and were compared to 160 images from non-defective transducers in an observer study with four experienced radiologists. There were four mandatory questions concerning whether the defects were detectable, whether the possible defects might affect the diagnosis, the visibility of structural details and total image quality. For three of the transducers the defects were detectable, the visibility of structural details being assessed as worse for all the four defective transducers. The total image quality was assessed as worse for three of the defective transducers. Out of the assessments of the images produced by the defective transducers, 19 % were “confident that the artifacts could affect the diagnosis”. This thesis shows that differences in image quality between different kind of ultrasound scanners can be detected using human observers. The thesis also shows that clinical image quality can be affected by defective transducers and that there is a risk of misdiagnosis if transducers with severe defects are used. The developed method for automatic detection of non-uniformities in clinical images could be used as a complement to normal quality control for earlier detection of defective transducers.