OBJECTIVES: The most commonly used summary metric in neuroimaging is the mean value, but this pays little attention to the shape of the data distribution and can therefore be insensitive to subtle changes that alter the data distribution.
METHODS: We propose a distributional-based metric called the normalized histogram similarity measure (HSM) for characterization of quantitative images. We applied HSM to quantitative magnetic resonance imaging T1 relaxation data of 44 patients with mild traumatic brain injury and compared with data of 43 age-matched controls.
RESULTS: Significant differences were found between the patients and the controls in 8 gray matter regions using the HSM whereas in only 1 gray matter region based on the mean values.
CONCLUSIONS: Our results show that HSM is more sensitive than the standard mean values in detecting brain tissue changes. Future studies on brain tissue properties using quantitative magnetic resonance imaging should consider the use of HSM to properly capture any tissue changes.