Premium
Optimized extraction of the medial temporal lobe for postmortem MRI based on custom 3D printed molds
Author(s) -
Lasserve Jade,
Lim Sydney A.,
Wisse Laura,
Ittyerah Ranjit,
Ravikumar Sadhana,
Lavery Madigan,
Robinson John L.,
Schuck Theresa,
Grossman Murray,
Lee Eddie B.,
Yushkevich Paul A.,
Tisdall Dylan M.,
Prabhakaran Karthik,
Mizsei Gabor,
ArtachoPerula Emilio,
Martin Maria Mercedes Iniguez de Onzono,
Jimenez Maria del Mar Arroyo,
Munoz Monica,
Romero Francisco Javier Molina,
Rabal Maria del Pilar Marcos,
Irwin David J.,
Trojanowski John Q.,
Wolk David A.,
Insausti Ricardo
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.043254
Subject(s) - magnetic resonance imaging , segmentation , computer science , 3d printed , artificial intelligence , lobe , computer vision , temporal lobe , anatomy , biomedical engineering , medicine , biology , radiology , neuroscience , epilepsy
Background Structural magnetic resonance imaging (MRI) biomarkers are important for early detection of Alzheimer’s Disease (AD). However, atrophy measures can be confounded by changes due to aging and comorbid non‐AD neurodegenerative pathologies. Linking postmortem MRI of the medial temporal lobe (MTL) to histopathology may identify focal patterns of change associated specifically with early AD. We implemented a pipeline of high‐resolution MRI of MTL specimens and serial histopathology imaging [REF]. However, the task of extracting the intact MTL specimen such that it fits into the MRI coil requires anatomical expertise and has proven error prone. Here we present an algorithm to automatically create 3D printed molds guiding MTL extraction. Method INPUTS: 7T MRI scan of a formalin‐fixed hemisphere in which the hemisphere, MTL ROI and optional second ROI to be spared during cutting (e.g. frontal lobe) have been segmented using ITK‐SNAP semi‐automatic segmentation tools. OUTPUTS: Two 3D printed molds with slits that guide cutting. Mold 1 holds the whole hemisphere, guiding four cuts orthogonal to the midsagittal plane (Figure 1). Mold 2 holds the extracted tissue block, guiding three subsequent longitudinal cuts that trim the tissue to fit into a 50mm cylindrical holder (Figure 2). The positioning of the cuts can be specified interactively by the user using ITK‐SNAP (translating and rotating 3D images representing cutting planes) or automatically by optimizing (using Powell’s method) energy functions that minimize the volume of the final piece of tissue under multiple constraints (see Figures 3 and 4). Result The algorithm with interactively positioned cut planes was used in four hemispheres; the automated version in one (Figure 5). For each MRI scan, the MTL was intact. By contrast, retrospective review revealed cutting errors in 48% of manually cut specimens. Conclusion Our image‐guided approach reduces errors and dependence on anatomical expertise; allows more tissue to be spared from each brain donation; and enables postmortem imaging at a larger scale. It is not limited to the MTL and could be of interest to brain banks and AD research centers involved in postmortem imaging.