Deep-learning dark-blood LGE provided comparable lesion-to-myocardium contrast and overall image quality to bright-blood LGE in pediatric patients, with faster acquisition times.
Observational (n=50)
No
Does a deep-learning assisted dark-blood LGE sequence provide comparable image quality and contrast to a bright-blood LGE sequence in pediatric patients?
Deep-learning assisted dark-blood LGE provides comparable image quality and contrast to bright-blood LGE but with faster acquisition, making it a viable alternative for pediatric patients.
Abstract Background Late gadolinium enhancement (LGE) imaging detects focal myocardial fibrosis and inflammation, for example in myocarditis. Short-axis stacks typically require multiple breath-holds over several minutes, which children may find difficult to perform. A deep-learning assisted dark-blood LGE sequence may reduce acquisition time compared to a deep-learning bright-blood phase sensitive inversion recovery sequence. Purpose We hypothesized that dark-blood imaging shows focal myocardial hyperenhancement, with better scar-to-myocardium contrast and with faster acquisition than bright-blood imaging. Methods Images from consecutive patients at a tertiary congenital heart hospital who had undergone LGE imaging with both bright-blood and dark-blood sequences were analyzed. Two experienced readers independently recorded location and extent of LGE positive lesions and measured signal intensity both in lesions and in normal myocardium. Signal intensity was averaged over three regions of interest in every sample, and contrast was calculated as the difference in signal intensities between lesion and myocardium. Overall image quality and visibility of anatomical details was graded on a scale from 1=poor to 4=excellent. Results 25 patients with LGE were identified (age 11.1 +/- 5.2 years; 8 female), as well as 25 age-matched controls (age 11.2 +/- 5.3 years; p = 0.965; 8 female). Overall image quality did not differ between sequences, while one of two observers found anatomical details to be better distinguishable by bright-blood imaging. The signal intensity contrast between lesions and myocardium was comparable between sequences. Conclusion(s) The deep-learning dark-blood LGE sequence shows focal myocardial hyperenhancement with comparable contrast to bright-blood LGE. Anatomical detail is possibly slightly less visible with dark-blood compared to bright-blood LGE, but overall image quality is similar. Dark-blood LGE is acquired much faster than bright-blood LGE, thus making it a viable alternative in pediatric patients.
Burkhardt et al. (2026) conducted an observational in Myocardial fibrosis and inflammation (n=50). Deep-learning assisted dark-blood LGE sequence vs. Deep-learning bright-blood phase sensitive inversion recovery sequence was evaluated on Signal intensity contrast between lesions and myocardium, and overall image quality. Deep-learning dark-blood LGE provided comparable lesion-to-myocardium contrast and overall image quality to bright-blood LGE in pediatric patients, with faster acquisition times.