Morphologic examination of tissue biopsies is essential for histopathological diagnosis. However, accurate and scalable cellular quantification in human samples remains challenging. Here, we present a deep learning–based approach for antigen-specific cellular morphometrics in human kidney biopsies, which combines indirect immunofluorescence imaging with U-Net–based architectures for image-to-image translation and dual segmentation tasks, achieving human-level accuracy. In the kidney, podocyte loss represents a hallmark of glomerular injury and can be estimated in diagnostic biopsies. Thus, we profiled over 27,000 podocytes from 110 human samples, including patients with antineutrophil cytoplasmic antibody–associated glomerulonephritis (ANCA-GN), an immune-mediated disease with aggressive glomerular damage and irreversible loss of kidney function. We identified previously unknown morphometric signatures of podocyte depletion in patients with ANCA-GN, which allowed patient classification and, in combination with routine clinical tools, showed potential for risk stratification. Our approach enables robust and scalable molecular morphometric analysis of human tissues, yielding deeper biological insights into the human kidney pathophysiology.

Abstract. Automatic visual speech recognition is an interesting problem in pattern recognition especially when audio data is noisy or not readily available. It is also a very challenging task mainly because of the lower amount of information in the visual articulations compared to the audible utterance. In this work, principle component analysis is applied to the image patches -extracted from the video data -to learn the weights of a two-stage convolutional network. Block histograms are then extracted as the unsupervised learning features. These features are employed to learn a recurrent neural network with a set of long short-term memory cells to obtain spatiotemporal features. Finally, the obtained features are used in a tandem GMM-HMM system for speech recognition. Our results show that the proposed method has outperformed the baseline techniques applied to the OuluVS2 audiovisual database for phrase recognition with the frontal view cross-validation and testing sentence correctness reaching 79% and 73%, respectively, as compared to the baseline of 74% on cross-validation.The final publication is available at Springer via http://dx

Expansion microscopy physically enlarges biological specimens to achieve nanoscale resolution using diffraction-limited microscopy systems1. However, optimal performance is usually reached using laser-based systems (for example, confocal microscopy), restricting its broad applicability in clinical pathology, as most centres have access only to light-emitting diode (LED)-based widefield systems. As a possible alternative, a computational method for image resolution enhancement, namely, super-resolution radial fluctuations (SRRF)2,3, has recently been developed. However, this method has not been explored in pathology specimens to date, because on its own, it does not achieve sufficient resolution for routine clinical use. Here, we report expansion-enhanced super-resolution radial fluctuations (ExSRRF), a simple, robust, scalable and accessible workflow that provides a resolution of up to 25 nm using LED-based widefield microscopy. ExSRRF enables molecular profiling of subcellular structures from archival formalin-fixed paraffin-embedded tissues in complex clinical and experimental specimens, including ischaemic, degenerative, neoplastic, genetic and immune-mediated disorders. Furthermore, as examples of its potential application to experimental and clinical pathology, we show that ExSRRF can be used to identify and quantify classical features of endoplasmic reticulum stress in the murine ischaemic kidney and diagnostic ultrastructural features in human kidney biopsies.

Morphologic examination of tissue biopsies is essential for histopathological diagnosis. However, accurate and scalable cellular quantification in human samples remains challenging. Here, we present a deep learning-based approach for antigen-specific cellular morphometrics in human kidney biopsies, which combines indirect immunofluorescence imaging with U-Net-based architectures for image-to-image translation and dual segmentation tasks, achieving human-level accuracy. In the kidney, podocyte loss represents a hallmark of glomerular injury and can be estimated in diagnostic biopsies. Thus, we profiled over 27,000 podocytes from 110 human samples, including patients with anti-neutrophil cytoplasmic antibody-associated glomerulonephritis (ANCA-GN), an immune-mediated disease with aggressive glomerular damage and irreversible loss of kidney function. Previously unknown morphometric signatures of podocyte depletion were identified in patients with ANCA-GN, which allowed patient classification and showed potential for risk stratification in combination with routine clinical tools. Together, our approach enables robust and scalable molecular morphometric analysis of human tissues, yielding deeper biological insights into the human kidney pathophysiology.SummaryDeep learning enables robust and scalable molecular morphometric analysis of human tissues, yielding deeper biological insights into the human kidney pathophysiology.

Visual speech recognition is a challenging research problem with a particular practical application of aiding audio speech recognition in noisy scenarios. Multiple camera setups can be beneficial for the visual speech recognition systems in terms of improved performance and robustness. In this paper, we explore this aspect and provide a comprehensive study on combining multiple views for visual speech recognition. The thorough analysis covers fusion of all possible view angle combinations both at feature level and decision level. The employed visual speech recognition system in this study extracts features through a PCA-based convolutional neural network, followed by an LSTM network. Finally, these features are processed in a tandem system, being fed into a GMM-HMM scheme. The decision fusion acts after this point by combining the Viterbi path log-likelihoods. The results show that the complementary information contained in recordings from different view angles improves the results significantly. For example, the sentence correctness on the test set is increased from 76% for the highest performing single view (30 • ) to up to 83% when combining this view with the frontal and 60 • view angles.

No abstract

No abstract

No abstract