Pixel precise unsupervised detection of viral particle proliferation in cellular imaging data

Abstract

Cellular and molecular imaging techniques and models have been developed to characterize single stages of viral proliferation after focal infection of cells in vitro . The fast and automatic classification of cell imaging data may prove helpful prior to any further comparison of representative experimental data to mathematical models of viral propagation in host cells. Here, we use computer generated images drawn from a reproduction of an imaging model from a previously published study of experimentally obtained cell imaging data representing progressive viral particle proliferation in host cell monolayers. Inspired by experimental time-based imaging data, here in this study viral particle increase in time is simulated by a one-by-one increase, across images, in black or gray single pixels representing dead or partially infected cells, and hypothetical remission by a one-by-one increase in white pixels coding for living cells in the original image model. The image simulations are submitted to unsupervised learning by a Self-Organizing Map (SOM) and the Quantization Error in the SOM output (SOM-QE) is used for automatic classification of the image simulations as a function of the represented extent of viral particle proliferation or cell recovery. Unsupervised classification by SOM-QE of 160 model images, each with more than three million pixels, is shown to provide a statistically reliable, pixel precise, and fast classification model that outperforms human computer-assisted image classification by RGB image mean computation. The automatic classification procedure proposed here provides a powerful approach to understand finely tuned mechanisms in the infection and proliferation of virus in cell lines in vitro or other cells.

Human Symmetry Uncertainty Detected by a Self-Organizing Neural Network Map

“…This provides further data showing that artificial neural networks are capable of detecting human uncertainty in perceptual judgment tasks [34]. The capability of the SOM-QE to capture such uncertainty in human choice responses to the symmetry of shapes with local variations in color parameters is tightly linked to the proven selectivity of this neural network metric to local contrast and color variations in large variety of complex image data [36][37][38][39][40][41][42][43][44]. Here, the metric is revealed as a measure of both variance in the image input data, and uncertainty in specific human decisions in response to such data.…”

“…The conceptual background and method of neural network analysis follows the same principle and protocol already described in our latest previous work on biological cell imaging data analysis by SOM [37,44]. It is described here again in full detail, for the benefit of the reader.…”

“…In our previous studies, we demonstrated functional properties of the SOM-QE such as a sensitivity to the spatial extent, intensity, and sign or color of local image contrasts in unsupervised image classification by the neural network. The metric reliably detects the finest, clinically or functionally relevant variations in local image contrast contents [36][37][38][39][40][41][42][43], often invisible to the human eye [38,39,42,43]. Here it will be shown that the SOM-QE as a neural network state metric reliably captures, or correlates with, varying levels of human uncertainty in the detection of symmetry of shape pairs with varying local color information contents.…”

“…To this end, we exploit a neural network metric in the output of a biologically inspired Self-Organizing Map (SOM). The Quantization Error (SOM-QE) [35][36][37][38][39][40][41][42][43][44] is a measure of output variance and quantifies the difference between neural network states at different stages of unsupervised image learning. In our previous studies, we demonstrated functional properties of the SOM-QE such as a sensitivity to the spatial extent, intensity, and sign or color of local image contrasts in unsupervised image classification by the neural network.…”

Deep Reinforcement Learning for the Control of Robotic Manipulation: A Focussed Mini-Review

“…As we know, in supervised deep learning, a training set of input-output pairs are fed to neutral networks to construct an approximation that maps an input to an output [54,55]. This learned target function can then be used for labeling new examples when a test set is given.…”