RevGel-seq: instrument-free single-cell RNA sequencing using a reversible hydrogel for cell-specific barcoding.

Progress in sample preparation for scRNA-seq is reported based on RevGel-seq, a reversible-hydrogel technology optimized for samples of fresh cells. Complexes of one cell paired with one barcoded bead are stabilized by a chemical linker and dispersed in a hydrogel in the liquid state. Upon gelation on ice the complexes are immobilized and physically separated without requiring nanowells or droplets. Cell lysis is triggered by detergent diffusion, and RNA molecules are captured on the adjacent barcoded beads for further processing with reverse transcription and preparation for cDNA sequencing. As a proof of concept, analysis of PBMC using RevGel-seq achieves results similar to microfluidic-based technologies when using the same original sample and the same data analysis software. In addition, a clinically relevant application of RevGel-seq is presented for pancreatic islet cells. Furthermore, characterizations carried out on cardiomyocytes demonstrate that the hydrogel technology readily accommodates very large cells. Standard analyses are in the 10,000-input cell range with the current gelation device, in order to satisfy common requirements for single-cell research. A convenient stopping point after two hours has been established by freezing at the cell lysis step, with full preservation of gene expression profiles. Overall, our results show that RevGel-seq represents an accessible and efficient instrument-free alternative, enabling flexibility in terms of experimental design and timing of sample processing, while providing broad coverage of cell types.

High-sensitivity SARS-CoV-2 group testing by digital PCR among symptomatic patients in hospital settings.

BACKGROUND: Worldwide demand for SARS-CoV-2 RT-PCR testing is still high as testing remains central to follow the disease spread and vaccine efficacy. Group testing has been proposed as a solution to expand testing capabilities but sensitivity concerns may limit its impact on the management of the pandemic. Digital PCR (RT-dPCR) has been shown to be highly sensitive and could help by providing larger testing capabilities without compromising sensitivity. METHODS: We implemented RT-dPCR based COVID-19 group testing on a commercially available system and assay (naica® system from Stilla Technologies) and investigated the sensitivity of the method in real life conditions of a university hospital in Paris, France, in May 2020. We tested the protocol in a direct comparison with reference RT-PCR testing on 448 samples split into groups of 8, 16 and 32 samples for RT-dPCR analysis. RESULTS: Individual RT-PCR testing identified 25/448 positive samples. Using 56 groups of 8, RT-dPCR identified 23 groups as positive, corresponding to 26 positive samples by individual PCR (positive percentage agreement 95.2% [95% confidence interval: 76.2-99.9%]) and including 2 samples not detected by individual RT-PCR but confirmed positive by further investigation. 15 of 28 groups of 16 tested positive, corresponding to 25 positive samples by individual PCR (positive percentage agreement 87.5% [95% confidence interval: 61.7-98.4%]). 14 groups of 32 were fully concordant with individual PCR testing but will need to be confirmed on larger datasets. CONCLUSIONS: Our proposed approach of group testing by digital PCR has similar diagnostic sensitivity compared to individual RT-PCR testing for group up to 16 samples. This approach reduces the quantity of reagent needed by up to 80% while reducing costs and increasing capabilities of testing up to 10-fold.

Correction: EGFR C797S, EGFR T790M and EGFR sensitizing mutations in non-small cell lung cancer revealed by six-color crystal digital PCR.

[This corrects the article DOI: 10.18632/oncotarget.26446.].

The effects of iatrogenic blood contamination on total nucleated cell counts and protein concentrations in canine cerebrospinal fluid.

BACKGROUND: Cerebrospinal fluid (CSF) might be altered by iatrogenic blood contamination, precluding accurate diagnostic interpretation. OBJECTIVES: Available formulas to correct for iatrogenic blood contamination are likely unreliable. Study objectives were to determine the effects of blood contamination on total nucleated cell counts (NCCs) and protein concentrations in canine CSF. METHODS: Two methods were followed to evaluate the effect of blood contamination on total NCC and protein concentrations in CSF. First, records from the Colorado State University Veterinary Teaching Hospital were retrospectively searched for dogs where CSF analysis was performed. Total NCCs, RBC counts, protein concentrations, and cytologic interpretations were recorded. Second, CSF from 4 canine patients and 3 research hounds was prospectively analyzed before and after known dilutions of whole blood were added. RESULTS: Of the 787 clinical samples analyzed, 108 samples had a cytologic diagnosis of blood contamination. RBC counts for all clinical samples ranged from 0 to 210,000 cells/µL. No correlation between total NCCs or protein concentrations with RBC counts were found when all samples were evaluated. Total NCCs and RBCs were weakly correlated in samples with a cytologic diagnosis of blood contamination and when ≥500 RBC/µL was present. When serial dilutions of whole blood were added to normal CSF, no significant changes were observed in the total NCCs of uncontaminated aliquots and contaminated aliquots containing up to 8480 RBC/µL. CONCLUSIONS: Erythrocyte counts in blood-contaminated canine CSF poorly correlate with total NCCs and protein concentrations. Using formulas to correct total NCCs and protein concentrations for the number of RBCs in CSF is inappropriate.

Detection of lipoarabinomannan in urine and serum of HIV-positive and HIV-negative TB suspects using an improved capture-enzyme linked immuno absorbent assay and gas chromatography/mass spectrometry.

TB diagnosis and treatment monitoring in resource limited regions rely heavily on serial sputum smear microscopy and bacterial culture. These microbiological methods are time-consuming, expensive and lack adequate sensitivity. The WHO states that improved TB diagnosis and treatment is imperative to achieve an end to the TB epidemic by 2030. Commercially available lipoarabinomannan (LAM) detection tools perform at low sensitivity that are highly dependent on the underlying immunological status of the patient; those with advanced HIV infection perform well. In this study, we have applied two novel strategies towards the sensitive diagnosis of TB infection based on LAM: Capture ELISA to detect LAM in paired urine and serum samples using murine and human monoclonal antibodies, essentially relying on LAM as an 'immuno-marker'; and, secondly, detection of α-d-arabinofuranose and tuberculostearic acid (TBSA)- 'chemical-markers' unique to mycobacterial cell wall polysaccharides/lipoglycans by our recently developed gas chromatography/mass spectrometry (GC/MS) method. Blinded urine specimens, with microbiologically confirmed active pulmonary TB or non TB (HIV+/HIV-) were tested by the aforementioned assays. LAM in patient urine was detected in a concentration range of 3-28 ng/mL based on GC/MS detection of the two LAM-surrogates, d-arabinose and tuberculostearic acid (TBSA) correctly classifying TB status with sensitivity > 99% and specificity = 84%. The ELISA assay had high sensitivity (98%) and specificity (92%) and the results were in agreement with GC/MS analysis. Both tests performed well in their present form particularly for HIV-negative/TB-positive urine samples. Among the HIV+/TB+ samples, 52% were found to have >10 ng/mL urinary LAM. The detected amounts of LAM present in the urine samples also appears to be associated with the gradation of the sputum smear, linking elevated LAM levels with higher mycobacterial burden (odds ratio = 1.08-1.43; p = 0.002). In this small set, ELISA was also applied to parallel serum samples confirming that serum could be an additional reservoir for developing a LAM-based immunoassay for diagnosis of TB.

Nod2 is required for antigen-specific humoral responses against antigens orally delivered using a recombinant Lactobacillus vaccine platform.

Safe and efficacious orally-delivered mucosal vaccine platforms are desperately needed to combat the plethora of mucosally transmitted pathogens. Lactobacillus spp. have emerged as attractive candidates to meet this need and are known to activate the host innate immune response in a species- and strain-specific manner. For selected bacterial isolates and mutants, we investigated the role of key innate immune pathways required for induction of innate and subsequent adaptive immune responses. Co-culture of murine macrophages with L. gasseri (strain NCK1785), L. acidophilus (strain NCFM), or NCFM-derived mutants-NCK2025 and NCK2031-elicited an M2b-like phenotype associated with TH2 skewing and immune regulatory function. For NCFM, this M2b phenotype was dependent on expression of lipoteichoic acid and S layer proteins. Through the use of macrophage genetic knockouts, we identified Toll-like receptor 2 (TLR2), the cytosolic nucleotide-binding oligomerization domain containing 2 (NOD2) receptor, and the inflammasome-associated caspase-1 as contributors to macrophage activation, with NOD2 cooperating with caspase-1 to induce inflammasome derived interleukin (IL)-1ß in a pyroptosis-independent fashion. Finally, utilizing an NCFM-based mucosal vaccine platform with surface expression of human immunodeficiency virus type 1 (HIV-1) Gag or membrane proximal external region (MPER), we demonstrated that NOD2 signaling is required for antigen-specific mucosal and systemic humoral responses. We show that lactobacilli differentially utilize innate immune pathways and highlight NOD2 as a key mediator of macrophage function and antigen-specific humoral responses to a Lactobacillus acidophilus mucosal vaccine platform.

Dynamic remodeling of lipids coincides with dengue virus replication in the midgut of Aedes aegypti mosquitoes.

We describe the first comprehensive analysis of the midgut metabolome of Aedes aegypti, the primary mosquito vector for arboviruses such as dengue, Zika, chikungunya and yellow fever viruses. Transmission of these viruses depends on their ability to infect, replicate and disseminate from several tissues in the mosquito vector. The metabolic environments within these tissues play crucial roles in these processes. Since these viruses are enveloped, viral replication, assembly and release occur on cellular membranes primed through the manipulation of host metabolism. Interference with this virus infection-induced metabolic environment is detrimental to viral replication in human and mosquito cell culture models. Here we present the first insight into the metabolic environment induced during arbovirus replication in Aedes aegypti. Using high-resolution mass spectrometry, we have analyzed the temporal metabolic perturbations that occur following dengue virus infection of the midgut tissue. This is the primary site of infection and replication, preceding systemic viral dissemination and transmission. We identified metabolites that exhibited a dynamic-profile across early-, mid- and late-infection time points. We observed a marked increase in the lipid content. An increase in glycerophospholipids, sphingolipids and fatty acyls was coincident with the kinetics of viral replication. Elevation of glycerolipid levels suggested a diversion of resources during infection from energy storage to synthetic pathways. Elevated levels of acyl-carnitines were observed, signaling disruptions in mitochondrial function and possible diversion of energy production. A central hub in the sphingolipid pathway that influenced dihydroceramide to ceramide ratios was identified as critical for the virus life cycle. This study also resulted in the first reconstruction of the sphingolipid pathway in Aedes aegypti. Given conservation in the replication mechanisms of several flaviviruses transmitted by this vector, our results highlight biochemical choke points that could be targeted to disrupt transmission of multiple pathogens by these mosquitoes.

EGFR C797S, EGFR T790M and EGFR sensitizing mutations in non-small cell lung cancer revealed by six-color crystal digital PCR.

Background: Detection of EGFR sensitizing and p.T790M and p.C797S resistance mutations is particularly important for non-small cell lung cancer (NSCLC) patient therapy management. Non-invasive blood-based monitoring of these mutations may pave the way to a fine-tuned personalized treatment. Digital PCR has emerged as an extremely sensitive method to detect rare mutations, however its major limitation is the number of hotspots that can be simultaneously differentiated. Methods: We developed a 6-color digital PCR assay for the detection and quantification of 19 most prevalent EGFR sensitizing and resistance mutations and evaluated this assay on 82 tumor and plasma samples from NSLC patients. Results: Limits of detection (LOD) for the 6-color digital PCR assay were assessed on serial dilutions of DNA standards. We found that the 6-color assay enabled detection of mutant fractions as low as 1 mutant in 1025 wild-type molecules, depending on the mutation targeted, when assayed in a background of 10 000 wild-type DNA copies. EGFR mutant allelic fraction was also measured on tumor and plasma samples by 6-color digital PCR, and displayed a highly significant correlation with next generation sequencing and 3-color digital PCR. Lastly, the 6-color digital PCR assay was performed on several longitudinal plasma samples from four patients and revealed levels of sensitizing and resistance EGFR mutations that reflected well the course of the disease. Conclusion: This 6-color Crystal digital PCR assay could represent a robust solution using digital PCR for the monitoring of EGFR mutations.

Crystal digital droplet PCR for detection and quantification of circulating EGFR sensitizing and resistance mutations in advanced non-small cell lung cancer.

Over the past years, targeted therapies using tyrosine kinase inhibitors (TKI) have led to an increase in progression-free survival and response rate for a subgroup of non-small cell lung cancer (NSCLC) patients harbouring specific gene abnormalities compared with chemotherapy. However long-lasting tumor regression is rarely achieved, due to the development of resistant tumoral subclones, which requires alternative therapeutic approaches. Molecular profile at progressive disease is a challenge for making adaptive treatment decisions. The aim of this study was to monitor EGFR-mutant tumors over time based on the quantity of mutant DNA circulating in plasma (ctDNA), comparing two different methods, Crystal™ Digital™ PCR and Massive Parallel Sequencing (MPS). In plasma circulating cell free DNA (cfDNA) of 61 advanced NSCLC patients we found an overall correlation of 78% between mutated allelic fraction measured by Crystal Digital PCR and MPS. 7 additional samples with sensitizing mutations and 4 additional samples with the resistance mutation were detected with Crystal Digital PCR, but not with MPS. Monitoring levels of both mutation types over time showed a correlation between levels and trends of mutated ctDNA detected and clinical assessment of disease for the 6 patients tested. In conclusion, Crystal Digital PCR exhibited good performance for monitoring mutational status in plasma cfDNA, and also appeared as better suited to the detection of known mutations than MPS in terms of features such as time to results.

Metabolic differentiation of early Lyme disease from southern tick-associated rash illness (STARI).

Lyme disease, the most commonly reported vector-borne disease in the United States, results from infection with Borrelia burgdorferi. Early clinical diagnosis of this disease is largely based on the presence of an erythematous skin lesion for individuals in high-risk regions. This, however, can be confused with other illnesses including southern tick-associated rash illness (STARI), an illness that lacks a defined etiological agent or laboratory diagnostic test, and is coprevalent with Lyme disease in portions of the eastern United States. By applying an unbiased metabolomics approach with sera retrospectively obtained from well-characterized patients, we defined biochemical and diagnostic differences between early Lyme disease and STARI. Specifically, a metabolic biosignature consisting of 261 molecular features (MFs) revealed that altered N-acyl ethanolamine and primary fatty acid amide metabolism discriminated early Lyme disease from STARI. Development of classification models with the 261-MF biosignature and testing against validation samples differentiated early Lyme disease from STARI with an accuracy of 85 to 98%. These findings revealed metabolic dissimilarity between early Lyme disease and STARI, and provide a powerful and new approach to inform patient management by objectively distinguishing early Lyme disease from an illness with nearly identical symptoms.

Type 1 Reaction in Patients With Leprosy Corresponds to a Decrease in Proresolving Lipid Mediators and an Increase in Proinflammatory Lipid Mediators.

Background: Type 1 reaction (T1R) is an acute T-helper type 1 (Th1) inflammatory episode in patients with leprosy. While immunological responses associated with T1R have been investigated, the corresponding metabolic responses that could contribute to T1R pathology have received little attention. Methods: Metabolomics-based analyses of sera from 7 patients with and 9 without T1R were conducted via liquid chromatography-mass spectrometry. Serum metabolites present at levels that significantly differed (P < .05) with a log2 fold change of ≥ 1.0 between patient groups were interrogated against known metabolic pathways. The structural identification of targeted metabolites was confirmed and abundance changes validated by mass spectrometry and enzyme-linked immunoassay. Results: Forty metabolic pathways were perturbed in patients with T1R, with 71 dysregulated metabolites mapping to pathways for lipid mediators of inflammation. Of note was an increase in the abundance of the proinflammatory leukotriene B4 (LTB4) and a corresponding decrease in the level of proresolving resolvin D1 (RvD1). Also, levels of prostaglandin D2 (PGD2) and lipoxin A4 (LXA4) in patients with T1R were significantly increased, while the level of prostaglandin E2 (PGE2) was decreased. Conclusions: The dysregulation of metabolic pathways leading to abundance shifts between proinflammatory and proresolving lipid mediators provides a link between metabolic and cellular immune responses that result in the Th1-mediated pathology of T1R.

Temporal and Spatial Variability of Entomological Risk Indices for West Nile Virus Infection in Northern Colorado: 2006-2013.

West Nile virus (WNV) is enzootic in northern Colorado. Annual surveillance activities in Fort Collins, CO, include collecting female Culex mosquitoes and testing them for the presence of WNV RNA in order to calculate 1) Culex female abundance, 2) WNV infection rate, and 3) the vector index (VI). These entomological risk indices inform public policy regarding the need for emergency adulticiding. Currently, these are calculated on a city-wide basis. In this study, we present descriptive data from historical surveillance records spanning 2006-2013 to discern seasonal and yearly patterns of entomological risk for WNV infection. Also, we retrospectively test the hypothesis that entomological risk is correlated with human transmission risk and is heterogeneous within the City of Fort Collins. Four logistically relevant zones within the city were established and used to test this hypothesis. Zones in the eastern portion of the city consistently had significantly higher Culex abundance and VI compared with zones in the west, leading to higher entomological risk indicators for human WNV infection in the east. Moreover, the relative risk of a reported human case of WNV infection was significantly higher in the eastern zones of the city. Our results suggest that a more spatially targeted WNV management program may better mitigate human risk for WNV infection in Fort Collins, and possibly other cities where transmission is enzootic, while at the same time reducing pesticide use.

Semi-automated query construction for content-based endomicroscopy video retrieval.

Content-based video retrieval has shown promising results to help physicians in their interpretation of medical videos in general and endomicroscopic ones in particular. Defining a relevant query for CBVR can however be a complex and time-consuming task for non-expert and even expert users. Indeed, uncut endomicroscopy videos may very well contain images corresponding to a variety of different tissue types. Using such uncut videos as queries may lead to drastic performance degradations for the system. In this study, we propose a semi-automated methodology that allows the physician to create meaningful and relevant queries in a simple and efficient manner. We believe that this will lead to more reproducible and more consistent results. The validation of our method is divided into two approaches. The first one is an indirect validation based on per video classification results with histopathological ground-truth. The second one is more direct and relies on perceived inter-video visual similarity ground-truth. We demonstrate that our proposed method significantly outperforms the approach with uncut videos and approaches the performance of a tedious manual query construction by an expert. Finally, we show that the similarity perceived between videos by experts is significantly correlated with the inter-video similarity distance computed by our retrieval system.

Software for automated classification of probe-based confocal laser endomicroscopy videos of colorectal polyps.

AIM: To support probe-based confocal laser endomicroscopy (pCLE) diagnosis by designing software for the automated classification of colonic polyps. METHODS: Intravenous fluorescein pCLE imaging of colorectal lesions was performed on patients undergoing screening and surveillance colonoscopies, followed by polypectomies. All resected specimens were reviewed by a reference gastrointestinal pathologist blinded to pCLE information. Histopathology was used as the criterion standard for the differentiation between neoplastic and non-neoplastic lesions. The pCLE video sequences, recorded for each polyp, were analyzed off-line by 2 expert endoscopists who were blinded to the endoscopic characteristics and histopathology. These pCLE videos, along with their histopathology diagnosis, were used to train the automated classification software which is a content-based image retrieval technique followed by k-nearest neighbor classification. The performance of the off-line diagnosis of pCLE videos established by the 2 expert endoscopists was compared with that of automated pCLE software classification. All evaluations were performed using leave-one-patient-out cross-validation to avoid bias. RESULTS: Colorectal lesions (135) were imaged in 71 patients. Based on histopathology, 93 of these 135 lesions were neoplastic and 42 were non-neoplastic. The study found no statistical significance for the difference between the performance of automated pCLE software classification (accuracy 89.6%, sensitivity 92.5%, specificity 83.3%, using leave-one-patient-out cross-validation) and the performance of the off-line diagnosis of pCLE videos established by the 2 expert endoscopists (accuracy 89.6%, sensitivity 91.4%, specificity 85.7%). There was very low power (< 6%) to detect the observed differences. The 95% confidence intervals for equivalence testing were: -0.073 to 0.073 for accuracy, -0.068 to 0.089 for sensitivity and -0.18 to 0.13 for specificity. The classification software proposed in this study is not a "black box" but an informative tool based on the query by example model that produces, as intermediate results, visually similar annotated videos that are directly interpretable by the endoscopist. CONCLUSION: The proposed software for automated classification of pCLE videos of colonic polyps achieves high performance, comparable to that of off-line diagnosis of pCLE videos established by expert endoscopists.

Learning semantic and visual similarity for endomicroscopy video retrieval.

Content-based image retrieval (CBIR) is a valuable computer vision technique which is increasingly being applied in the medical community for diagnosis support. However, traditional CBIR systems only deliver visual outputs, i.e., images having a similar appearance to the query, which is not directly interpretable by the physicians. Our objective is to provide a system for endomicroscopy video retrieval which delivers both visual and semantic outputs that are consistent with each other. In a previous study, we developed an adapted bag-of-visual-words method for endomicroscopy retrieval, called "Dense-Sift," that computes a visual signature for each video. In this paper, we present a novel approach to complement visual similarity learning with semantic knowledge extraction, in the field of in vivo endomicroscopy. We first leverage a semantic ground truth based on eight binary concepts, in order to transform these visual signatures into semantic signatures that reflect how much the presence of each semantic concept is expressed by the visual words describing the videos. Using cross-validation, we demonstrate that, in terms of semantic detection, our intuitive Fisher-based method transforming visual-word histograms into semantic estimations outperforms support vector machine (SVM) methods with statistical significance. In a second step, we propose to improve retrieval relevance by learning an adjusted similarity distance from a perceived similarity ground truth. As a result, our distance learning method allows to statistically improve the correlation with the perceived similarity. We also demonstrate that, in terms of perceived similarity, the recall performance of the semantic signatures is close to that of visual signatures and significantly better than those of several state-of-the-art CBIR methods. The semantic signatures are thus able to communicate high-level medical knowledge while being consistent with the low-level visual signatures and much shorter than them. In our resulting retrieval system, we decide to use visual signatures for perceived similarity learning and retrieval, and semantic signatures for the output of an additional information, expressed in the endoscopist own language, which provides a relevant semantic translation of the visual retrieval outputs.

Online blind calibration of non-uniform photodetectors: application to endomicroscopy.

We present an original method for the online blind calibration of non-uniform photodetectors. The disparity of the detectors may arise from both irregular spatial arrangement and distinct slowly time-varying photometric transfer functions. As natural images are mostly continuous, the signal collected by neighboring detectors is strongly correlated over time. The core idea of our method is to translate the calibration problem into relative pairwise calibrations between neighboring detectors followed by the regularized inversion of a system akin to gradient-based surface recovery. From our blind calibration procedure, we design an online blind calibration pipeline compatible with clinical practice. Online blind calibration is proved to be statistically better than standard offline calibration for reconstructing endomicroscopy sequences.

Retrieval evaluation and distance learning from perceived similarity between endomicroscopy videos.

Evaluating content-based retrieval (CBR) is challenging because it requires an adequate ground-truth. When the available groundtruth is limited to textual metadata such as pathological classes, retrieval results can only be evaluated indirectly, for example in terms of classification performance. In this study we first present a tool to generate perceived similarity ground-truth that enables direct evaluation of endomicroscopic video retrieval. This tool uses a four-points Likert scale and collects subjective pairwise similarities perceived by multiple expert observers. We then evaluate against the generated ground-truth a previously developed dense bag-of-visual-words method for endomicroscopic video retrieval. Confirming the results of previous indirect evaluation based on classification, our direct evaluation shows that this method significantly outperforms several other state-of-the-art CBR methods. In a second step, we propose to improve the CBR method by learning an adjusted similarity metric from the perceived similarity ground-truth. By minimizing a margin-based cost function that differentiates similar and dissimilar video pairs, we learn a weight vector applied to the visual word signatures of videos. Using cross-validation, we demonstrate that the learned similarity distance is significantly better correlated with the perceived similarity than the original visual-word-based distance.

A smart atlas for endomicroscopy using automated video retrieval.

To support the challenging task of early epithelial cancer diagnosis from in vivo endomicroscopy, we propose a content-based video retrieval method that uses an expert-annotated database. Motivated by the recent successes of non-medical content-based image retrieval, we first adjust the standard Bag-of-Visual-Words method to handle single endomicroscopic images. A local dense multi-scale description is proposed to keep the proper level of invariance, in our case to translations, in-plane rotations and affine transformations of the intensities. Since single images may have an insufficient field-of-view to make a robust diagnosis, we introduce a video-mosaicing technique that provides large field-of-view mosaic images. To remove outliers, retrieval is followed by a geometrical approach that captures a statistical description of the spatial relationships between the local features. Building on image retrieval, we then focus on efficient video retrieval. Our approach avoids the time-consuming parts of the video-mosaicing by relying on coarse registration results only to account for spatial overlap between images taken at different times. To evaluate the retrieval, we perform a simple nearest neighbors classification with leave-one-patient-out cross-validation. From the results of binary and multi-class classification, we show that our approach outperforms, with statistical significance, several state-of-the art methods. We obtain a binary classification accuracy of 94.2%, which is quite close to clinical expectations.

An image retrieval approach to setup difficulty levels in training systems for endomicroscopy diagnosis.

Learning medical image interpretation is an evolutive process that requires modular training systems, from non-expert to expert users. Our study aims at developing such a system for endomicroscopy diagnosis. It uses a difficulty predictor to try and shorten the physician learning curve. As the understanding of video diagnosis is driven by visual similarities, we propose a content-based video retrieval approach to estimate the level of interpretation difficulty. The performance of our retrieval method is compared with several state of the art methods, and its genericity is demonstrated with two different clinical databases, on the Barrett's Esophagus and on colonic polyps. From our retrieval results, we learn a difficulty predictor against a ground truth given by the percentage of false diagnoses among several physicians. Our experiments show that, although our datasets are not large enough to test for statistical significance, there is a noticeable relationship between our retrieval-based difficulty estimation and the difficulty experienced by the physicians.