InSpect: INtegrated SPECTral Component Estimation and Mapping for Multi-contrast Microstructural MRI

• Published in: Information Processing in Medical Imaging pp. 755 - 766
• Main Authors: Slator, Paddy J., Hutter, Jana, Marinescu, Razvan V., Palombo, Marco, Young, Alexandra L., Jackson, Laurence H., Ho, Alison, Chappell, Lucy C., Rutherford, Mary, Hajnal, Joseph V., Alexander, Daniel C.
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing
• Series: Lecture Notes in Computer Science
• ISBN: 3030203506; 9783030203504
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-20351-1_59

We introduce a novel algorithm for deriving meaningful maps from multi-contrast MRI experiments. Such experiments enable the estimation of multidimensional correlation spectra, in domains such as T1-diffusivity, T2-diffusivity, or T1-T2. These spectra combine information from complementary MR properties, and therefore have the potential for improved quantification of distinct tissue types compared to single-contrast analyses. However, spectral estimation is an ill-conditioned problem which is highly sensitive to noise and requires significant regularisation. We propose an Expectation-Maximisation based method - which we term InSpect - for unified analysis of multi-contrast MR images. The algorithm simultaneously estimates canonical spectra associated with distinct tissue types within an image, and produces maps quantifying the spatial distribution of these spectra. We test the algorithm’s capabilities on simulated data, then apply to placental diffusion-relaxometry data. On placental data we identified significant within-organ and across-subject variation in T2*-ADC spectra - showing the potential of InSpect for detailed separation and quantification of distinct microstructural environments.