High-throughput Image Analysis Research Articles

Évaluation immunohistochimique du risque métastasique dans les cancers du sein débutants sur microbiopsies

We examined a series of 667 patients with node-negative breast carcinomas in order to identify prognostic immunohistochemical molecular signatures for the prediction of early metastasis, and potential new therapeutic targets. We used a standardized quantitative immunocytochemical approach with 37 antibodies, based on high-throughput tissue microarrays and image analysis, and analyzed the results with respect to metastatic status after a mean follow-up of 86 months. Complete data were obtained for 586 patients. The predictive value of the markers was first analyzed individually by univariate analysis (log rank test) in 586 node-negative tumors, according to metastatic status during follow-up. Twenty-seven markers had significant prognostic value. ROC curve analysis (logistic regression) was then used to determine the marker combination that best classified the patients with and without metastases. A 15-marker signature (Bcl2, P16, P21, P27, P53, CD34, CA IX, c-kit, FGF-R1, P38, JAK, pSTAT3, CK9, STAT1 and ER) correctly classified 84.8% of the patients (sensitivity 85.5%, specificity 84.6%). When ER, PR and c-erb B2 were excluded from the analysis, a very similar signature, in which CK8-18 replaced ER, correctly classified 83.6% of the patients (sensitivity 84.5%, specificity 83.4%). These results show that quantitative immunoprofiling, independently of ER, PR and c-erb B2 status, can provide a basal-like signature, can properly predict the metastatic risk of node-negative breast carcinomas at diagnostic time. This may be helpful for selecting patients who do not need aggressive adjuvant chemotherapy, and for developing tailored therapies.

A scalable and embedded FPGA architecture for efficient computation of grey level co-occurrence matrices and Haralick textures features

This paper presents a novel and optimized embedded architecture based FPGA for an efficient and fast computation of grey level co-occurrence matrices (GLCM) and Haralick textures features for use in high throughput image analysis applications where time performance is critical. The originality of this architecture allows for a scalable and a totally embedded on Chip FPGA for the processing of large images. The architecture was implemented on Xilinx Virtex-FPGAs without the use of external memory and/or host machine. The implementations demonstrate that our proposed architecture can deliver a high reduction of the memory and FPGA logic requirements when compared with the state of the art counterparts and it also achieves much improved processing times when compared against optimized software implementation running on a conventional general purpose processor.

Image Analysis in High Content Screening

The field of High Content Screening (HCS) has evolved from a technology used exclusively by the pharmaceutical industry for secondary drug screening, to a technology used for primary drug screening and basic research in academia. The size and the complexity of the screens have been steadily increasing. This is reflected in the fact that the major challenges facing the field at the present are data mining and data storage due to the large amount of data generated during HCS. On the one hand, technological progress of fully automated image acquisition platforms, and on the other hand advances in the field of automated image analysis have made this technology more powerful and more accessible to less specialized users. Image analysis solutions for many biological problems exist and more are being developed to increase both the quality and the quantity of data extracted from the images acquired during the screens. We highlight in this review some of the major challenges facing automatic high throughput image analysis and present some of the software solutions available on the market or from academic open source solutions.

Quantifying cellular interaction dynamics in 3D fluorescence microscopy data

The wealth of information available from advanced fluorescence imaging techniques used to analyze biological processes with high spatial and temporal resolution calls for high-throughput image analysis methods. Here, we describe a fully automated approach to analyzing cellular interaction behavior in 3D fluorescence microscopy images. As example application, we present the analysis of drug-induced and S1P(1)-knockout-related changes in bone-osteoclast interactions. Moreover, we apply our approach to images showing the spatial association of dendritic cells with the fibroblastic reticular cell network within lymph nodes and to microscopy data regarding T-B lymphocyte synapse formation. Such analyses that yield important information about the molecular mechanisms determining cellular interaction behavior would be very difficult to perform with approaches that rely on manual/semi-automated analyses. This protocol integrates adaptive threshold segmentation, object detection, adaptive color channel merging, and neighborhood analysis and permits rapid, standardized, quantitative analysis and comparison of the relevant features in large data sets.

Regional rates of neocortical atrophy from normal aging to early Alzheimer disease

To evaluate the spatial pattern and regional rates of neocortical atrophy from normal aging to early Alzheimer disease (AD). Longitudinal MRI data were analyzed using high-throughput image analysis procedures for 472 individuals diagnosed as normal, mild cognitive impairment (MCI), or AD. Participants were divided into 4 groups based on Clinical Dementia Rating Sum of Boxes score (CDR-SB). Annual atrophy rates were derived by calculating percent cortical volume loss between baseline and 12-month scans. Repeated-measures analyses of covariance were used to evaluate group differences in atrophy rates across regions as a function of impairment. Planned comparisons were used to evaluate the change in atrophy rates across levels of disease severity. In patients with MCI-CDR-SB 0.5-1, annual atrophy rates were greatest in medial temporal, middle and inferior lateral temporal, inferior parietal, and posterior cingulate. With increased impairment (MCI-CDR-SB 1.5-2.5), atrophy spread to parietal, frontal, and lateral occipital cortex, followed by anterior cingulate cortex. Analysis of regional trajectories revealed increasing rates of atrophy across all neocortical regions with clinical impairment. However, increases in atrophy rates were greater in early disease within medial temporal cortex, whereas increases in atrophy rates were greater at later stages in prefrontal, parietal, posterior temporal, parietal, and cingulate cortex. Atrophy is not uniform across regions, nor does it follow a linear trajectory. Knowledge of the spatial pattern and rate of decline across the spectrum from normal aging to Alzheimer disease can provide valuable information for detecting early disease and monitoring treatment effects at different stages of disease progression.

Intact Flagellar Motor of Borrelia burgdorferi Revealed by Cryo-Electron Tomography: Evidence for Stator Ring Curvature and Rotor/C-Ring Assembly Flexion

The bacterial flagellar motor is a remarkable nanomachine that provides motility through flagellar rotation. Prior structural studies have revealed the stunning complexity of the purified rotor and C-ring assemblies from flagellar motors. In this study, we used high-throughput cryo-electron tomography and image analysis of intact Borrelia burgdorferi to produce a three-dimensional (3-D) model of the in situ flagellar motor without imposing rotational symmetry. Structural details of B. burgdorferi, including a layer of outer surface proteins, were clearly visible in the resulting 3-D reconstructions. By averaging the 3-D images of approximately 1,280 flagellar motors, a approximately 3.5-nm-resolution model of the stator and rotor structures was obtained. flgI transposon mutants lacked a torus-shaped structure attached to the flagellar rod, establishing the structural location of the spirochetal P ring. Treatment of intact organisms with the nonionic detergent NP-40 resulted in dissolution of the outermost portion of the motor structure and the C ring, providing insight into the in situ arrangement of the stator and rotor structures. Structural elements associated with the stator followed the curvature of the cytoplasmic membrane. The rotor and the C ring also exhibited angular flexion, resulting in a slight narrowing of both structures in the direction perpendicular to the cell axis. These results indicate an inherent flexibility in the rotor-stator interaction. The FliG switching and energizing component likely provides much of the flexibility needed to maintain the interaction between the curved stator and the relatively symmetrical rotor/C-ring assembly during flagellar rotation.

Quantitative intensity/area, semi‐automated controls for multiplex immunofluorescence tissue‐based assays

After developing two 5‐antigen multiplex (Mx) immunofluorescence assays using qualitative tissue controls in formalin‐fixed, paraffin‐embedded (FFPE) prostate cancer (CaP) tissue sections, we proposed to optimize control evaluation by quantification of signal‐derived features. Three CaP cell lines (LNCaP, PC‐3, DU‐145) were used as substrates for a nuclear antigen in each Mx [androgen receptor (AR) and Ki‐67]. Cells suspended in agar were FFPE as arrays (CLA) containing all three cell lines. Mx results are recorded by a computer‐driven microscope system with a Nuance(tm) camera (CRI, Woburn MA) and filters to capture serial images 10 nm apart (wave length range: 520 ‐ 720 nm) including emission spectra for all fluors used. Grey scale images identifying antigen distribution by its spectral profile are created by unmixing fluor signals in each image series. High throughput image analysis scripts and algorithms previously developed to interrogate tissue sections were adapted to measure intensity/area features of AR or Ki‐67 in CLA nuclei (classified as "DAPI objects"), expressed as (1) mean intensity of all cells, and (2) mean and (3) percentage of positive cells above a threshold derived from negative control sections. Results from a test run are thus validated by control metrics within established antigen/cell line ranges, an objective approach surpassing in accuracy and efficiency subjective visual criteria.

High throughput quantification of mutant huntingtin aggregates

Mutant protein aggregates are an important biomarker in Huntington's and other neurodegenerative diseases however their quantification has typically relied on manual imaging and counting, or cell-free assays, which do not allow for concurrent analysis of cell viability. Here we describe four automated high throughput image analysis methods, developed using Metamorph™ software, to quantify mutant huntingtin aggregates in a cellular context. Imaging of aggregate-forming cells was also automated, using a Discovery-1™ automated fluorescence microscope. All four analysis methods measured aggregate formation accurately in relation to manual counting, but with differing throughput. Our in-house PolyQ assay gave the highest throughput, processing images at 0.31 s per image. The Cell Scoring assay gave lower throughput, at 19.5 s per image, but offered accurate quantification of the proportion of cells which formed aggregates, without bias from cell death. These image analysis tools provide rapid and objective alternatives to manual counting in studies of aggregate formation, to facilitate the discovery of drugs to treat Huntington's and related neurodegenerative diseases.

Image-Based High-Throughput Quantification of Cellular Fat Accumulation

A number of biochemical methods are available for measuring fat accumulation in cell culture. The authors report a simple image-based method for measuring fat accumulation in adipocytes using a combination of high-throughput brightfield microscopy and image analysis, which was validated biochemically using Oil-Red-O. The quickest and most accurate method of analysis was one based on thresholding brightfield images and determining the area of fat droplets per image. Thus, the authors have developed a simple high-throughput, label-free method for measuring fat accumulation that is applicable to any cell or tissue type where fat droplets are visible under light microscopy.

High content screen microscopy analysis of Aβ 1–42-induced neurite outgrowth reduction in rat primary cortical neurons: Neuroprotective effects of α7 neuronal nicotinic acetylcholine receptor ligands

β-Amyloid peptide 1–42 (Aβ 1–42) is generated from amyloid precursor protein (APP) and associated with neurodegeneration in Alzheimer's disease (AD). Aβ 1–42 has been shown to be cytotoxic when incubated with cultured neurons. However, APP transgenic mice over-expressing Aβ 1–42 do not show substantial loss of neurons, despite deficits in learning and memory. It is thus emerging that Aβ 1–42-induced memory deficits may involve subtler neuronal alternations leading to synaptic deficits, prior to frank neurodegeneration in AD brains. In this study, high content screen (HCS) microscopy, an advanced high-throughput cellular image processing and analysis technique, was utilized in establishing an in vitro model of Aβ 1–42-induced neurotoxicity utilizing rat neonatal primary cortical cells. Neurite outgrowth was found to be significantly reduced by Aβ 1–42 (300 nM to 30 μM), but not by the scrambled control peptide control, in a time- and concentration-dependent manner. In contrast, no reduction in the total number of neurons was observed. The Aβ 1–42-induced reduction of neurite outgrowth was attenuated by the NMDA receptor antagonist memantine and the α7 nicotinic acetylcholine receptor (nAChR) selective agonist PNU-282987. Interestingly, the α7 nAChR antagonist methyllycaconitine also significantly prevented reduction in Aβ 1–42-induced neurite outgrowth. The observed neuroprotective effects could arise either from interference of Aβ 1–42 interactions with α7 nAChRs or by modification of receptor-mediated signaling pathways. Our studies demonstrate that reduction of neurite outgrowth may serve as a model representing Aβ 1–42-mediated neuritic and synaptic toxicity, which, in combination of HCS, provides a high-throughput cell-based assay that can be used to evaluate compounds with neuroprotective properties in neurons.

FPGA architecture for fast parallel computation of co-occurrence matrices

This paper presents a novel architecture for fast parallel computation of co-occurrence matrices in high throughput image analysis applications for which time performance is critical. The architecture was implemented on a Xilinx Virtex-XCV2000E-6 FPGA using VHDL. The symmetry and sparseness of the co-occurrence matrices are exploited to achieve improved processing times, and smaller, flexible area utilization as compared with the state of the art. The performance of the proposed architecture is evaluated using input images of various dimensions, in comparison with an optimized software implementation running on a conventional general purpose processor. Simulations of the architecture on contemporary FPGA devices show that it can deliver a speedup of two orders of magnitude over software.

CellProfiler: image analysis software for identifying and quantifying cell phenotypes

CellProfiler, the first free, open-source system for flexible and high-throughput cell image analysis is described.

Proteome analysis of adipogenesis

Adipose tissue plays a crucial endocrine role in controlling whole body glucose homeostasis and insulin sensitivity. Given the substantial rise in obesity and obesity-related diseases such as diabetes, it is important to understand the molecular basis of adipocyte differentiation and its control. Many studies have successfully exploited gene array technology to monitor changes in the profile of expressed genes during adipocyte differentiation, although this method only measures changes at the level of individual mRNA species. Using two-dimensional polyacrylamide gel electrophoresis, high-throughput image analysis, and candidate picking coupled with sequencing mass spectrometry, we have followed the changes in protein expression profile that occur during the differentiation of 3T3-L1 fibroblasts into adipocytes in response to dexamethasone, isobutyl methyl xanthine and insulin, or to the PPARgamma agonist, ciglitazone. Using this technique we have found alterations in the profile of over 2000 protein species during adipogenesis. Our studies reveal previously unknown alterations during adipogenesis in the expression or mobility (on sodium dodecyl sulfate-polyacrylamide gel electrophoresis) of coactosin, which promotes actin filament destabilization, several signalling molecules, including RhoGDI-1, RhoGDI-2 and EHD1, and NEDD5 a protein involved in cytokinesis.

Amyloid deposition in the hippocampus and entorhinal cortex: quantitative analysis of a transgenic mouse model.

Various transgenic mouse models of Alzheimer's disease (AD) have been developed that overexpress mutant forms of amyloid precursor protein in an effort to elucidate more fully the potential role of beta-amyloid (A beta) in the etiopathogenesis of the disease. The present study represents the first complete 3D reconstruction of A beta in the hippocampus and entorhinal cortex of PDAPP transgenic mice. A beta deposits were detected by immunostaining and thioflavin fluorescence, and quantified by using high-throughput digital image acquisition and analysis. Quantitative analysis of amyloid load in hippocampal subfields showed a dramatic increase between 12 and 15 months of age, with little or no earlier detectable deposition. Three-dimensional reconstruction in the oldest brains visualized previously unrecognized sheets of A beta coursing through the hippocampus and cerebral cortex. In contrast with previous hypotheses, compact plaques form before significant deposition of diffuse A beta, suggesting that different mechanisms are involved in the deposition of diffuse amyloid and the aggregation into plaques. The dentate gyrus was the hippocampal subfield with the greatest amyloid burden. Sublaminar distribution of A beta in the dentate gyrus correlated most closely with the termination of afferent projections from the lateral entorhinal cortex, mirroring the selective vulnerability of this circuit in human AD. This detailed temporal and spatial analysis of A beta and compact amyloid deposition suggests that specific corticocortical circuits express selective, but late, vulnerability to the pathognomonic markers of amyloid deposition, and can provide a basis for detecting prior vulnerability factors.