Histological Image Analysis of Scaffold‐Based Cell Products to Estimate Percent Viability

Abstract

IntroductionRegulatory agencies have an expectation that cell‐based products in clinical development be characterized for their percent (%) viability. However, many new cell‐based products do not readily lend themselves to determining this metric through classical methods such as enzymatic dissociation and cell enumeration. Vital Therapies, Inc. (VTL) uses four metabolically active cartridges containing VTL C3A cells in its investigational (phase 3) ELAD® System, an extracorporeal human allogeneic cellular treatment being evaluated for severe alcoholic hepatitis. This study objective was to develop and verify new image processing methods for automatically characterizing the % viability of ELAD C3A cell cartridge histological samples.MethodsELAD C3A cell cartridges were clinically manufactured by introducing VTL C3A cells into the extracapillary space and perfusing growth medium through the lumens of the porous polysulfone hollow fibers. At maturity, cartridges were fixed for histology by recirculating 10% neutral buffered formalin through the hollow fiber lumens for approximately 24 hr. Cross‐sectional discs (~2‐cm thick) were cut using a saw and further cut into 16‐grid samples prior to paraffin embedding. Thin (4‐mm) sections were sliced from the grid blocks, stained with hematoxylin and eosin, and imaged using a high‐resolution scanner (Olympus VS120 or Hamamatsu NanoZoomer 2.0). MATLAB software was utilized to develop a custom algorithm to automatically isolate, count, and mask the hollow fiber cross‐sections through image processing techniques (e.g., color thresholding and object feature segmentation). Grayscale histogram analysis of the differently stained cells in the remaining space was utilized to separate viable from non‐viable areas. The % viability for an entire histology grid image could then be reported as the integrated ratio of viable tissue area to total tissue area. This method was compared to blinded, independent mass measurements of printed and hand‐cut images to separate hollow fiber, viable tissue, and non‐viable tissue areas. Additional comparisons were made against blinded visual assessment.ResultsThis methodology demonstrated that reliable % viability estimates could be estimated from histological images despite fixation and/or sectioning artifacts (e.g., tissue separating from the hollow fibers). The automated image‐processing method was able to complete the analysis very rapidly (e.g., an 8000×8000 pixel image containing over 600 fibers is analyzed in a couple minutes), such that a large library of histological samples could be completed quickly and the results tallied for inter‐grid, intra‐ and inter‐cartridge, and intra‐ and inter‐lot comparisons.ConclusionsNew cell‐based products require creative approaches to address regulatory expectations. The image processing algorithm developed herein was shown to be in agreement with, and more reproducible and rapid than, visual or manual assessments. It was further recognized that this method could be extended to other types of histological stains (e.g., trichrome) or to other types of biomaterial/scaffold‐based tissue‐engineered products.This abstract is from the Experimental Biology 2018 Meeting. There is no full text article associated with this abstract published in The FASEB Journal.