Performance discrepancy across tumours |
ImFusion |
‘Very good for clear-cut, simple tumours... [those which were] highly contrast enhancing, homogeneous, with well-defined boundaries and minimal adjacent blood vessels.’ |
All five participants (100%) |
‘Complex tumours threw up errors which required a complete restart.’ |
Compromise between thoroughness and timing |
ITK-SNAP |
‘I am a perfectionist... if we weren’t timed it would take me much longer.’ |
Four out of five (80%) |
‘I made lots of small mistakes... but it would have taken too long to correct.’ |
‘It was very fiddly.’ |
Unpredictable outcome after drawing labels |
ImFusion |
‘a one-trick pony...if you make alterations to the initial segmentation you may worsen it.’ |
Three out of five (60%) |
‘if we wanted perfection...we would have to go back again and again.’ |
‘I do not know if the changes I make will improve of worsen the segmentation.’ |
Speed of segmentation |
ImFusion |
‘Much faster so it would be great for my work.’ |
Four out of five (80%) |
‘The algorithm works very quickly.’ |
UI and tools |
Both |
‘[Using ImFusion] was a much nicer experience...and a sleek UI.’ |
- |
‘[ImFusion] is better for visualization.’ |
‘[In ImFusion] I would like to have a paintbrush tool which draws and erases exactly what I want it to...there is too much prediction required...scribbles I make should not affect the whole segmentation.’ |
Study design |
– |
‘It was helpful to have someone here to explain and provide feedback.’ |
- |
‘Would have been good to define the goal more clearly...do we want a very accurate segmentation or a rough volume estimate.’ |
‘You could have gone through all the tools I might need during the training phase.’ |