The use of semi-automated and automated strategies for co registration and correlation between radiographic and pathologic image analysis in the central nervous system - a feasibility study

• Grant number: 15/02770-3
• Support Opportunities: Scholarships abroad - Research
• Start date: May 01, 2015
• End date: April 30, 2016
• Field of knowledge: Health Sciences - Medicine - Pathological Anatomy and Clinical Pathology
• Principal Investigator: Luiz Fernando Ferraz da Silva
• Grantee: Luiz Fernando Ferraz da Silva
• Host Investigator: Lilla Zollei
• Host Institution: Faculdade de Medicina (FM). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Institution abroad: Massachusetts General Hospital, United States
• Associated research grant: 13/21728-2 - The use of modern autopsy techniques to investigate human diseases (MODAU), AP.TEM

Abstract

INTRODUCTION AND OBJECTIVE The autopsy is considered the most reliable tool to validate the results obtained through imaging techniques, an area with recent advances in terms of contrast and spatial resolution for medical diagnosis. A key point in the integration of radiological and pathological quantitative results is the co-registration, whose applicability must consider a three-dimensional level (between the image and organ) and a two-dimensional (between the slice of the organ and the sampled tissue fragment). There are available semi-automated tools and some automated algorithms that take into account these levels and allow more reliable interaction and potentially better correlation between the findings.Our center have been developing highlighted research using autopsy and had recently implemented an imaging platform with a 7 Tesla Magnetic Resonance Imaging (7TMRI). In this context, the improvement of integrative strategies and comparative analysis for radiological and pathological parameters is critical to achieve new standards in these projects.The objective of the present study is to evaluate the applicability of different co-registration and co-localization strategies allowing radiological and pathological quantitative image analysis correlation in the brain that can be replicated for other organic systems.STRATEGY AND METHODS: The project includes two phases integrating radiological and pathological information, using the central nervous system as a model once it`s the area in which co-register strategies are most advanced, mainly for studies using 7TMRI.Phase 1: We will compare two image co-registration tools in the assessment and identification of focal lesions in the central nervous system, the White Matter Hyper Intensities (WMH), correlating radiological and pathological findings. We will include 4 cases from which we already have in-situ post-mortem MRI images, ex-situ MRI performed 120 days after fixation, brain slices pictures and histological slides of the WMH areas. The spatial semiautomated co-registration and reconstruction was performed using Amira software. From these data we will be applied automated co-registration algorithms using FreeSurfer tool for comparison. Using the existing stained slides with HE, Nissl, Kluever-Barrera, and immunohistochemistry anti-NF, anti-GFAP and anti-MBP we will be able to quantify pathological parameters such as the positive area, total area, optical density, number of vessels, neurons and glial cells for correlation with the radiological parameters such as proton diffusion measure.Phase 2: We will use automated co-registration techniques to study the structure of the thalamus and thalamic-cortical connections using 7T MRI and anatomic-pathologic correlation. We will prospectively include 2 children who died between 0 and 6 months of age. An in-situ post-mortem 7TMRI will be held. The brain, after 60 days formalin fixation will be re-scanned with the same parameters and sliced (vertex-frontal orientation) and pictures will be taken. Semi-automated and automated co-registration techniques will be applied. For automated co-registration we will use the FreeSurfer tool and its algorithms. Thalamic regions will be mapped from the RM7T and, using the co-registration data, fragments of about 2,5x2,5 cm will be removed from slices for microscopy and pathological image analysis. The slides will be subject to the same colors mentioned for the first subproject and the same parameters quantified for comparison.RELEVANCE: The Laboratory of Computational Neuroimaging (Martinos' Center / MGH / Harvard) is a leader in co-registration techniques and will train the proponent for the application these tools. The proponent will additionally perform the pathological evaluation and measurements. The ultimate goal will be the replication of co-registration, co-localization and histological-pathological correlation strategies to other organs and systems in Brazil.