Evaluation of a computational model for mycetoma-causative agents identification.

BACKGROUND: The therapeutic strategy for mycetoma relies heavily on the identification of the causative agents, which are either fungal or bacterial. While histopathological examination of surgical biopsies is currently the most used diagnostic tool, it requires well-trained pathologists, who are lacking in most rural areas where mycetoma is endemic. In this work we propose and evaluate a machine learning approach that semi-automatically analyses histopathological microscopic images of grains and provides a classification of the disease as eumycetoma or actinomycetoma. METHODS: The computational model is based on radiomics and partial least squares. It is assessed on a dataset that includes 890 individual grains collected from 168 patients originating from the Mycetoma Research Centre in Sudan. The dataset contained 94 eumycetoma cases and 74 actinomycetoma cases, with a distribution of the species among the two causative agents that is representative of the Sudanese distribution. RESULTS: The proposed model achieved identification of causative agents with an accuracy of 91.89%, which is comparable to the accuracy of experts from the domain. The method was found to be robust to a small error in the segmentation of the grain and to changes in the acquisition protocol. Among the radiomics features, the homogeneity of mycetoma grain textures was found to be the most discriminative feature for causative agent identification. CONCLUSION: The results presented in this study support that this computational approach could greatly benefit rural areas with limited access to specialized clinical centres and also provide a second opinion for expert pathologists to implement the appropriate therapeutic strategy.

Enhancing histopathological image classification of invasive ductal carcinoma using hybrid harmonization techniques.

This study aims to develop a robust pipeline for classifying invasive ductal carcinomas and benign tumors in histopathological images, addressing variability within and between centers. We specifically tackle the challenge of detecting atypical data and variability between common clusters within the same database. Our feature engineering-based pipeline comprises a feature extraction step, followed by multiple harmonization techniques to rectify intra- and inter-center batch effects resulting from image acquisition variability and diverse patient clinical characteristics. These harmonization steps facilitate the construction of more robust and efficient models. We assess the proposed pipeline's performance on two public breast cancer databases, BreaKHIS and IDCDB, utilizing recall, precision, and accuracy metrics. Our pipeline outperforms recent models, achieving 90-95% accuracy in classifying benign and malignant tumors. We demonstrate the advantage of harmonization for classifying patches from different databases. Our top model scored 94.7% for IDCDB and 95.2% for BreaKHis, surpassing existing feature engineering-based models (92.1% for IDCDB and 87.7% for BreaKHIS) and attaining comparable performance to deep learning models. The proposed feature-engineering-based pipeline effectively classifies malignant and benign tumors while addressing variability within and between centers through the incorporation of various harmonization techniques. Our findings reveal that harmonizing variabilities between patches from different batches directly impacts the learning and testing performance of classification models. This pipeline has the potential to enhance breast cancer diagnosis and treatment and may be applicable to other diseases.

Radiomics prognostic analysis of PET/CT images in a multicenter head and neck cancer cohort: investigating ComBat strategies, sub-volume characterization, and automatic segmentation.

PURPOSE: This study aimed to investigate the impact of several ComBat harmonization strategies, intra-tumoral sub-volume characterization, and automatic segmentations for progression-free survival (PFS) prediction through radiomics modeling for patients with head and neck cancer (HNC) in PET/CT images. METHODS: The HECKTOR MICCAI 2021 challenge set containing PET/CT images and clinical data of 325 oropharynx HNC patients was exploited. A total of 346 IBSI-compliant radiomic features were extracted for each patient's primary tumor volume defined by the reference manual contours. Modeling relied on least absolute shrinkage Cox regression (Lasso-Cox) for feature selection (FS) and Cox proportional-hazards (CoxPH) models were built to predict PFS. Within this methodological framework, 8 different strategies for ComBat harmonization were compared, including before or after FS, in feature groups separately or all features directly, and with center or clustering-determined labels. Features extracted from tumor sub-volume clustering were also investigated for their prognostic additional value. Finally, 3 automatic segmentations (2 threshold-based and a 3D U-Net) were also compared. All results were evaluated with the concordance index (C-index). RESULTS: Radiomics features without harmonization, combined with clinical factors, led to models with C-index values of 0.69 in the testing set. The best version of ComBat harmonization, i.e., after FS, for feature groups separately and relying on clustering-determined labels, achieved a C-index of 0.71. The use of features extracted from tumor sub-volumes further improved the C-index to 0.72. Models that relied on the automatic segmentations yielded close but slightly lower prognostic performance (0.67-0.70) compared to reference contours. CONCLUSION: A standard radiomics pipeline allowed for prediction of PFS in a multicenter HNC cohort. Applying a specific strategy of ComBat harmonization improved the performance. The extraction of intra-tumoral sub-volume features and automatic segmentation could contribute to the improvement and automation of prognosis modeling, respectively.

QU-BraTS: MICCAI BraTS 2020 Challenge on Quantifying Uncertainty in Brain Tumor Segmentation - Analysis of Ranking Scores and Benchmarking Results.

Deep learning (DL) models have provided state-of-the-art performance in various medical imaging benchmarking challenges, including the Brain Tumor Segmentation (BraTS) challenges. However, the task of focal pathology multi-compartment segmentation (e.g., tumor and lesion sub-regions) is particularly challenging, and potential errors hinder translating DL models into clinical workflows. Quantifying the reliability of DL model predictions in the form of uncertainties could enable clinical review of the most uncertain regions, thereby building trust and paving the way toward clinical translation. Several uncertainty estimation methods have recently been introduced for DL medical image segmentation tasks. Developing scores to evaluate and compare the performance of uncertainty measures will assist the end-user in making more informed decisions. In this study, we explore and evaluate a score developed during the BraTS 2019 and BraTS 2020 task on uncertainty quantification (QU-BraTS) and designed to assess and rank uncertainty estimates for brain tumor multi-compartment segmentation. This score (1) rewards uncertainty estimates that produce high confidence in correct assertions and those that assign low confidence levels at incorrect assertions, and (2) penalizes uncertainty measures that lead to a higher percentage of under-confident correct assertions. We further benchmark the segmentation uncertainties generated by 14 independent participating teams of QU-BraTS 2020, all of which also participated in the main BraTS segmentation task. Overall, our findings confirm the importance and complementary value that uncertainty estimates provide to segmentation algorithms, highlighting the need for uncertainty quantification in medical image analyses. Finally, in favor of transparency and reproducibility, our evaluation code is made publicly available at https://github.com/RagMeh11/QU-BraTS.

Quantitative analysis of the formation of nucleoprotein complexes between HIV-1 Gag protein and genomic RNA using transmission electron microscopy.

In HIV, the polyprotein precursor Gag orchestrates the formation of the viral capsid. In the current view of this viral assembly, Gag forms low-order oligomers that bind to the viral genomic RNA triggering the formation of high-ordered ribonucleoprotein complexes. However, this assembly model was established using biochemical or imaging methods that do not describe the cellular location hosting Gag-gRNA complex nor distinguish gRNA packaging in single particles. Here, we studied the intracellular localization of these complexes by electron microscopy and monitored the distances between the two partners by morphometric analysis of gold beads specifically labeling Gag and gRNA. We found that formation of these viral clusters occurred shortly after the nuclear export of the gRNA. During their transport to the plasma membrane, the distance between Gag and gRNA decreases together with an increase of gRNA packaging. Point mutations in the zinc finger patterns of the nucleocapsid domain of Gag caused an increase in the distance between Gag and gRNA as well as a sharp decrease of gRNA packaged into virions. Finally, we show that removal of stem loop 1 of the 5'-untranslated region does not interfere with gRNA packaging, whereas combined with the removal of stem loop 3 is sufficient to decrease but not abolish Gag-gRNA cluster formation and gRNA packaging. In conclusion, this morphometric analysis of Gag-gRNA cluster formation sheds new light on HIV-1 assembly that can be used to describe at nanoscale resolution other viral assembly steps involving RNA or protein-protein interactions.

Asymmetric Ensemble of Asymmetric U-Net Models for Brain Tumor Segmentation With Uncertainty Estimation.

Accurate brain tumor segmentation is crucial for clinical assessment, follow-up, and subsequent treatment of gliomas. While convolutional neural networks (CNN) have become state of the art in this task, most proposed models either use 2D architectures ignoring 3D contextual information or 3D models requiring large memory capacity and extensive learning databases. In this study, an ensemble of two kinds of U-Net-like models based on both 3D and 2.5D convolutions is proposed to segment multimodal magnetic resonance images (MRI). The 3D model uses concatenated data in a modified U-Net architecture. In contrast, the 2.5D model is based on a multi-input strategy to extract low-level features from each modality independently and on a new 2.5D Multi-View Inception block that aims to merge features from different views of a 3D image aggregating multi-scale features. The Asymmetric Ensemble of Asymmetric U-Net (AE AU-Net) based on both is designed to find a balance between increasing multi-scale and 3D contextual information extraction and keeping memory consumption low. Experiments on 2019 dataset show that our model improves enhancing tumor sub-region segmentation. Overall, performance is comparable with state-of-the-art results, although with less learning data or memory requirements. In addition, we provide voxel-wise and structure-wise uncertainties of the segmentation results, and we have established qualitative and quantitative relationships between uncertainty and prediction errors. Dice similarity coefficient for the whole tumor, tumor core, and tumor enhancing regions on BraTS 2019 validation dataset were 0.902, 0.815, and 0.773. We also applied our method in BraTS 2018 with corresponding Dice score values of 0.908, 0.838, and 0.800.

In situ Analytical Quality Control of chemotherapeutic solutions in infusion bags by Raman spectroscopy.

Analytical Quality Control (AQC) in centralised preparation units of oncology centers is a common procedure relying on the identification and quantification of the prepared chemotherapeutic solutions for safe intravenous administration to patients. Although the use of Raman spectroscopy for AQC has gained much interest, in most applications it remains coupled to a flow injection analyser (FIA) requiring withdrawal of the solution for analysis. In addition to current needs for more rapid and cost-effective analysis, the risk of exposure of clinical staff to the toxic molecules during daily handling is a serious concern to address. Raman spectroscopic analysis, for instance by Confocal Raman Microscopy (CRM), could enable direct analysis (non-invasive) for AQC directly in infusion bags. In this study, 3 anticancer drugs, methotrexate (MTX), 5-fluorouracil (5-FU) and gemcitabine (GEM) have been selected to highlight the potential of CRM for withdrawal free analysis. Solutions corresponding to the clinical range of each drug were prepared in 5% glucose and data was collected from infusion bags placed under the Raman microscope. Firstly, 100% discrimination has been obtained by Partial Least Squares Discriminant Analysis (PLS-DA) confirming that the identification of drugs can be performed. Secondly, using Partial Least Squares Regression (PLSR), quantitative analysis was performed with mean % error of predicted concentrations of respectively 3.31%, 5.54% and 8.60% for MTX, 5-FU and GEM. These results are in accordance with the 15% acceptance criteria used for the current clinical standard technique, FIA, and the Limits of Detection for all drugs were determined to be substantially lower than the administered range, thus highlighting the potential of confocal Raman spectroscopy for direct analysis of chemotherapeutic solutions.

Differentiated Cells in Prolonged Hypoxia Produce Highly Infectious Native-Like Hepatitis C Virus Particles.

BACKGROUND AND AIMS: Standard hepatitis C virus (HCV) cell-culture models present an altered lipid metabolism and thus produce lipid-poor lipoviral particles (LVPs). These models are thereby weakly adapted to explore the complete natural viral life cycle. APPROACH AND RESULTS: To overcome these limitations, we used an HCV cell-culture model based on both cellular differentiation and sustained hypoxia to better mimic the host-cell environment. The long-term exposure of Huh7.5 cells to DMSO and hypoxia (1% O2 ) significantly enhanced the expression of major differentiation markers and the cellular hypoxia adaptive response by contrast with undifferentiated and normoxic (21% O2 ) standard conditions. Because hepatocyte-like differentiation and hypoxia are key regulators of intracellular lipid metabolism, we characterized the distribution of lipid droplets (LDs) and demonstrated that experimental cells significantly accumulate larger and more numerous LDs relative to standard cell-culture conditions. An immunocapture (IC) and transmission electron microscopy (TEM) method showed that differentiated and hypoxic Huh7.5 cells produced lipoproteins significantly larger than those produced by standard Huh7.5 cell cultures. The experimental cell culture model is permissive to HCV-Japanese fulminant hepatitis (JFH1) infection and produces very-low-buoyant-density LVPs that are 6-fold more infectious than LVPs formed by standard JFH1-infected Huh7.5 cells. Finally, the IC-TEM approach and antibody-neutralization experiments revealed that LVPs were highly lipidated, had a global ultrastructure and a conformation of the envelope glycoprotein complex E1E2 close to that of the ones circulating in infected individuals. CONCLUSIONS: This relevant HCV cell culture model thus mimics the complete native intracellular HCV life cycle and, by extension, can be proposed as a model of choice for studies of other hepatotropic viruses.

Study of influence of the glutamatergic concentration of [18F]FPEB binding to metabotropic glutamate receptor subtype 5 with N-acetylcysteine challenge in rats and SRM/PET study in human healthy volunteers.

Altered glutamate signaling is thought to be involved in a myriad of psychiatric disorders. Positron emission tomography (PET) imaging with [18F]FPEB allows assessing dynamic changes in metabotropic glutamate receptor 5 (mGluR5) availability underlying neuropathological conditions. The influence of endogenous glutamatergic levels into receptor binding has not been well established yet. The purpose of this study was to explore the [18F]FPEB binding regarding to physiological fluctuations or acute changes of glutamate synaptic concentrations by a translational approach; a PET/MRS imaging study in 12 healthy human volunteers combined to a PET imaging after an N-acetylcysteine (NAc) pharmacological challenge in rodents. No significant differences were observed with small-animal PET in the test and retest conditions on the one hand and the NAc condition on the other hand for any regions. To test for an interaction of mGuR5 density and glutamatergic concentrations in healthy subjects, we correlated the [18F]FPEB BPND with Glu/Cr, Gln/Cr, Glx/Cr ratios in the anterior cingulate cortex VOI; respectively, no significance correlation has been revealed (Glu/Cr: r = 0.51, p = 0.09; Gln/Cr: r = -0.46, p = 0.13; Glx/Cr: r = -0.035, p = 0.92).These data suggest that the in vivo binding of [18F]FPEB to an allosteric site of the mGluR5 is not modulated by endogenous glutamate in vivo. Thus, [18F]FPEB appears unable to measure acute fluctuations in endogenous levels of glutamate.

Normal volumetric and T1 relaxation time values at 1.5 T in segmented pediatric brain MRI using a MP2RAGE acquisition.

OBJECTIVES: This study introduced a tailored MP2RAGE-based brain acquisition for a comprehensive assessment of the normal maturing brain. METHODS: Seventy normal patients (35 girls and 35 boys) from 1 to 16 years of age were recruited within a prospective monocentric study conducted from a single University Hospital. Brain MRI examinations were performed at 1.5 T using a 20-channel head coil and an optimized 3D MP2RAGE sequence with a total acquisition time of 6:36 min. Automated 38 region segmentation was performed using the MorphoBox (template registration, bias field correction, brain extraction, and tissue classification) which underwent a major adaptation of three age-group T1-weighted templates. Volumetry and T1 relaxometry reference ranges were established using a logarithmic model and a modified Gompertz growth respectively. RESULTS: Detailed automated brain segmentation and T1 mapping were successful in all patients. Using these data, an age-dependent model of normal brain maturation with respect to changes in volume and T1 relaxometry was established. After an initial rapid increase until 24 months of life, the total intracranial volume was found to converge towards 1400 mL during adolescence. The expected volumes of white matter (WM) and cortical gray matter (GM) showed a similar trend with age. After an initial major decrease, T1 relaxation times were observed to decrease progressively in all brain structures. The T1 drop in the first year of life was more pronounced in WM (from 1000-1100 to 650-700 ms) than in GM structures. CONCLUSION: The 3D MP2RAGE sequence allowed to establish brain volume and T1 relaxation time normative ranges in pediatrics. KEY POINTS: • The 3D MP2RAGE sequence provided a reliable quantitative assessment of brain volumes and T1 relaxation times during childhood. • An age-dependent model of normal brain maturation was established. • The normative ranges enable an objective comparison to a normal cohort, which can be useful to further understand, describe, and identify neurodevelopmental disorders in children.

Qualitative and quantitative analysis of therapeutic solutions using Raman and infrared spectroscopy.

Anticancer drugs are prescribed and administrated to an increasing number of patients on a daily basis. As a consequence, a number of concerns have been raised about the patient health and safety in the case that the drugs administered are not at the required concentration or even worse not the correct ones. Quality control of therapeutic solutions has therefore been extensively implemented in hospital environments, in order to avoid any failure in the intense workflow faced by administering pharmacists. In the present study, infrared (IR) and Raman spectroscopy have been employed for the analysis of 3 commercially available therapeutic solutions TEVA®, MYLAN®, CERUBIDINE®, respectively containing doxorubicin, epirubicin and daunorubicin. They perfectly illustrate the analytical difficulties encountered, as these 3 chemotherapeutic drugs are isomers, hardly distinguishable with conventional approaches such as UV/VIS spectrometry. Any analytical failure to identify these molecules can lead to delays in patient treatment. While Partial Least Squares Regression analysis demonstrates that both Raman and IR can deliver satisfactory quantitative analysis in the clinical range, with respective Root Mean Square Error of Cross Validation (RMSECV) between 0.0127 - 0.0220 g·L-1 and 0.0573 - 0.0759 g·L-1, the identification rate between the 2 techniques differs substantially. Indeed, Principal Component Analysis - Factorial Discriminant Analysis (PCA-FDA) highlights that, depending on the data preprocessing applied to Raman spectra, the discrimination between the 3 drugs is decreased, with in some cases specificity and sensitivity below 50%. However, IR analysis displays encouraging results with an overall specificity and sensitivity between 99 and 100%, suggesting that reliable validation of the therapeutic solution for administration to patients can be achieved. IR and Raman spectroscopy could assist and support quality control of chemotherapeutic solutions prepared in personalised concentrations for each patient. The effective and reliable characterisation of therapeutic solutions could have a lot to offer to improve current practices in a near future.

The first MICCAI challenge on PET tumor segmentation.

INTRODUCTION: Automatic functional volume segmentation in PET images is a challenge that has been addressed using a large array of methods. A major limitation for the field has been the lack of a benchmark dataset that would allow direct comparison of the results in the various publications. In the present work, we describe a comparison of recent methods on a large dataset following recommendations by the American Association of Physicists in Medicine (AAPM) task group (TG) 211, which was carried out within a MICCAI (Medical Image Computing and Computer Assisted Intervention) challenge. MATERIALS AND METHODS: Organization and funding was provided by France Life Imaging (FLI). A dataset of 176 images combining simulated, phantom and clinical images was assembled. A website allowed the participants to register and download training data (n = 19). Challengers then submitted encapsulated pipelines on an online platform that autonomously ran the algorithms on the testing data (n = 157) and evaluated the results. The methods were ranked according to the arithmetic mean of sensitivity and positive predictive value. RESULTS: Sixteen teams registered but only four provided manuscripts and pipeline(s) for a total of 10 methods. In addition, results using two thresholds and the Fuzzy Locally Adaptive Bayesian (FLAB) were generated. All competing methods except one performed with median accuracy above 0.8. The method with the highest score was the convolutional neural network-based segmentation, which significantly outperformed 9 out of 12 of the other methods, but not the improved K-Means, Gaussian Model Mixture and Fuzzy C-Means methods. CONCLUSION: The most rigorous comparative study of PET segmentation algorithms to date was carried out using a dataset that is the largest used in such studies so far. The hierarchy amongst the methods in terms of accuracy did not depend strongly on the subset of datasets or the metrics (or combination of metrics). All the methods submitted by the challengers except one demonstrated good performance with median accuracy scores above 0.8.