Using virtual microscopy for the development of sampling strategies in quantitative histology and design-based stereology.

Only a fraction of specimens under study are usually selected for quantification in histology. Multilevel sampling or tissue probes, slides and fields of view (FOVs) in the regions of interest (ROIs) are required. In general, all parts of the organs under study should be given the same probability to be taken into account; that is, the sampling should be unbiased on all levels. The objective of our study was to provide an overview of the use of virtual microscopy in the context of developing sampling strategies of FOVs for stereological quantification. We elaborated this idea on 18 examples from multiple fields of histology, including quantification of extracellular matrix and muscle tissue, quantification of organ and tumour microvessels and tumour-infiltrating lymphocytes, assessing osseointegration of bone implants, healing of intestine anastomoses and osteochondral defects, counting brain neurons, counting nuclei in vitro cell cultures and others. We provided practical implications for the most common situations, such as exhaustive sampling of ROIs, sampling ROIs of different sizes, sampling the same ROIs for multiple histological methods, sampling more ROIs with variable intensities or using various objectives, multistage sampling and virtual sampling. Recommendations were provided for pilot studies on systematic uniform random sampling of FOVs as a part of optimizing the efficiency of histological quantification to prevent over- or undersampling. We critically discussed the pros and cons of using virtual sections for sampling FOVs from whole scanned sections. Our review demonstrated that whole slide scans of histological sections facilitate the design of sampling strategies for quantitative histology.

Why Use Position Features in Liver Segmentation Performed by Convolutional Neural Network.

Liver volumetry is an important tool in clinical practice. The calculation of liver volume is primarily based on Computed Tomography. Unfortunately, automatic segmentation algorithms based on handcrafted features tend to leak segmented objects into surrounding tissues like the heart or the spleen. Currently, convolutional neural networks are widely used in various applications of computer vision including image segmentation, while providing very promising results. In our work, we utilize robustly segmentable structures like the spine, body surface, and sagittal plane. They are used as key points for position estimation inside the body. The signed distance fields derived from these structures are calculated and used as an additional channel on the input of our convolutional neural network, to be more specific U-Net, which is widely used in medical image segmentation tasks. Our work shows that this additional position information improves the results of the segmentation. We test our approach in two experiments on two public datasets of Computed Tomography images. To evaluate the results, we use the Accuracy, the Hausdorff distance, and the Dice coefficient. Code is publicly available at: https://gitlab.com/hachaf/liver-segmentation.git.

Semantic Segmentation of Intralobular and Extralobular Tissue from Liver Scaffold H&E Images.

Decellularized tissue is an important source for biological tissue engineering. Evaluation of the quality of decellularized tissue is performed using scanned images of hematoxylin-eosin stained (H&E) tissue sections and is usually dependent on the observer. The first step in creating a tool for the assessment of the quality of the liver scaffold without observer bias is the automatic segmentation of the whole slide image into three classes: the background, intralobular area, and extralobular area. Such segmentation enables to perform the texture analysis in the intralobular area of the liver scaffold, which is crucial part in the recellularization procedure. Existing semi-automatic methods for general segmentation (i.e., thresholding, watershed, etc.) do not meet the quality requirements. Moreover, there are no methods available to solve this task automatically. Given the low amount of training data, we proposed a two-stage method. The first stage is based on classification of simple hand-crafted descriptors of the pixels and their neighborhoods. This method is trained on partially annotated data. Its outputs are used for training of the second-stage approach, which is based on a convolutional neural network (CNN). Our architecture inspired by U-Net reaches very promising results, despite a very low amount of the training data. We provide qualitative and quantitative data for both stages. With the best training setup, we reach 90.70% recognition accuracy.

Novel morphological multi-scale evaluation system for quality assessment of decellularized liver scaffolds.

Decellularized scaffolds can serve as an excellent three-dimensional environment for cell repopulation. They maintain tissue-specific microarchitecture of extracellular matrix proteins with important spatial cues for cell adhesion, migration, growth, and differentiation. However, criteria for quality assessment of the three-dimensional structure of decellularized scaffolds are rather fragmented, usually study-specific, and mostly semi-quantitative. Thus, we aimed to develop a robust structural assessment system for decellularized porcine liver scaffolds. Five scaffolds of different quality were used to establish the new evaluation system. We combined conventional semi-quantitative scoring criteria with a quantitative scaffold evaluation based on automated image analysis. For the quantitation, we developed a specific open source software tool (ScaffAn) applying algorithms designed for texture analysis, segmentation, and skeletonization. ScaffAn calculates selected parameters characterizing structural features of porcine liver scaffolds such as the sinusoidal network. After evaluating individual scaffolds, the total scores predicted scaffold interaction with cells in terms of cell adhesion. Higher scores corresponded to higher numbers of cells attached to the scaffolds. Moreover, our analysis revealed that the conventional system could not identify fine differences between good quality scaffolds while the additional use of ScaffAn allowed discrimination. This led us to the conclusion that only using the combined score resulted in the best discrimination between different quality scaffolds. Overall, our newly defined evaluation system has the potential to select the liver scaffolds most suitable for recellularization, and can represent a step toward better success in liver tissue engineering.

Generating standardized image data for testing and calibrating quantification of volumes, surfaces, lengths, and object counts in fibrous and porous materials using X-ray microtomography.

Quantification of the structure and composition of biomaterials using micro-CT requires image segmentation due to the low contrast and overlapping radioopacity of biological materials. The amount of bias introduced by segmentation procedures is generally unknown. We aim to develop software that generates three-dimensional models of fibrous and porous structures with known volumes, surfaces, lengths, and object counts in fibrous materials and to provide a software tool that calibrates quantitative micro-CT assessments. Virtual image stacks were generated using the newly developed software TeIGen, enabling the simulation of micro-CT scans of unconnected tubes, connected tubes, and porosities. A realistic noise generator was incorporated. Forty image stacks were evaluated using micro-CT, and the error between the true known and estimated data was quantified. Starting with geometric primitives, the error of the numerical estimation of surfaces and volumes was eliminated, thereby enabling the quantification of volumes and surfaces of colliding objects. Analysis of the sensitivity of the thresholding upon parameters of generated testing image sets revealed the effects of decreasing resolution and increasing noise on the accuracy of the micro-CT quantification. The size of the error increased with decreasing resolution when the voxel size exceeded 1/10 of the typical object size, which simulated the effect of the smallest details that could still be reliably quantified. Open-source software for calibrating quantitative micro-CT assessments by producing and saving virtually generated image data sets with known morphometric data was made freely available to researchers involved in morphometry of three-dimensional fibrillar and porous structures in micro-CT scans.

Stereological quantification of microvessels using semiautomated evaluation of X-ray microtomography of hepatic vascular corrosion casts.

PURPOSE: Quantitative description of hepatic microvascular bed could contribute to understanding perfusion CT imaging. Micro-CT is a useful method for the visualization and quantification of capillary-passable vascular corrosion casts. Our aim was to develop and validate open-source software for the statistical description of the vascular networks in micro-CT scans. METHODS: Porcine hepatic microvessels were injected with Biodur E20 resin, and the resulting corrosion casts were scanned with 1.9-4.7 [Formula: see text] resolution. The microvascular network was quantified using newly developed QuantAn software both in randomly selected volume probes (n = 10) and in arbitrarily outlined hepatic lobules (n = 4). The volumes, surfaces, lengths, and numbers of microvessel segments were estimated and validated in the same data sets with manual stereological counting. Calculations of tortuosity, radius histograms, length histograms, exports of the skeletonized vascular network into open formats, and an assessment of the degree of their anisotropy were performed. RESULTS: Within hepatic lobules, the microvessels had a volume fraction of 0.13 [Formula: see text] 0.05, surface density of 21.0 [Formula: see text] 2.0 [Formula: see text], length density of 169.0 [Formula: see text] 40.2 [Formula: see text], and numerical density of 588.5 [Formula: see text] 283.1 [Formula: see text]. Sensitivity analysis of the automatic analysis to binary opening, closing, threshold offset, and aggregation radius of branching nodes was performed. CONCLUSION: The software QuantAn and its source code are openly available to researchers working in the field of stochastic geometry of microvessels in micro-CT scans or other three-dimensional imaging methods. The implemented methods comply with reproducible stereological techniques, and they were highly consistent with manual counting. Preliminary morphometrics of the classical hepatic lobules in pig were provided.