Showing papers by "Peter Bajcsy published in 2020"

Deep learning predicts function of live retinal pigment epithelium from quantitative microscopy.

Abstract: Increases in the number of cell therapies in the preclinical and clinical phases have prompted the need for reliable and noninvasive assays to validate transplant function in clinical biomanufacturing. We developed a robust characterization methodology composed of quantitative bright-field absorbance microscopy (QBAM) and deep neural networks (DNNs) to noninvasively predict tissue function and cellular donor identity. The methodology was validated using clinical-grade induced pluripotent stem cell-derived retinal pigment epithelial cells (iPSC-RPE). QBAM images of iPSC-RPE were used to train DNNs that predicted iPSC-RPE monolayer transepithelial resistance, predicted polarized vascular endothelial growth factor (VEGF) secretion, and matched iPSC-RPE monolayers to the stem cell donors. DNN predictions were supplemented with traditional machine-learning algorithms that identified shape and texture features of single cells that were used to predict tissue function and iPSC donor identity. These results demonstrate noninvasive cell therapy characterization can be achieved with QBAM and machine learning.

Approaches to training multiclass semantic image segmentation of damage in concrete.

Abstract: This paper addresses the problem of creating a large quantity of high-quality training segmentation masks from scanning electron microscopy (SEM) images. The images are acquired from concrete samples that exhibit progressive amounts of degradation resulting from alkali-silica reaction (ASR), a leading cause of deterioration, cracking and loss of capacity in much of the nation's infrastructure. The target damage classes in concrete SEM images are defined as paste damage, aggregate damage, air voids and no damage. We approached the SEM segmentation problem by applying convolutional neural network (CNN)-based methods to predict the damage classes due to ASR for each image pixel. The challenges in using the CNN-based methods lie in preparing large numbers of high-quality training labelled images while having limited human resources. To address these challenges, we designed damage- and context-assisted approaches to lower the requirements on human resources. We then evaluated the accuracy of CNN-based segmentation methods using the datasets prepared with these two approaches. LAY DESCRIPTION: This work is about automated segmentation of Scanning Electron Microscopy (SEM) images taken from core and prism samples of concrete. The segmentation must detect several damage classes in each image in order to understand properties of concrete-made structures over time. The segmentation problem is approached with an artificial network (AI) based model. The training data for the AI model are created using damage- and context-assisted approaches to lower the requirements on human resources. The access to all training data and to a web-based validation system for scoring segmented images is available at https://isg.nist.gov/deepzoomweb/data/concreteScoring.

Interoperability of web computational plugins for large microscopy image analyses

Abstract: This document summarizes conclusions from a workshop focused on Interoperability of Web Computational Plugins for Large Microscopy Image Analyses. The workshop conclusions are classified as practical recommendations and agreements on development and future research related to (1) containerization of execution code, (2) data storage, (3) interoperability requirements of workflow engines for running containerized plugins, (4) standard packaging of web user interface modules, and (5) security of container-based distribution.

Scientific Calculator for Designing Trojan Detectors in Neural Networks

Abstract: This work presents a web-based interactive neural network (NN) calculator and a NN inefficiency measurement that has been investigated for the purpose of detecting trojans embedded in NN models. This NN Calculator is designed on top of TensorFlow Playground with in-memory storage of data and NN graphs plus coefficients. It is "like a scientific calculator" with analytical, visualization, and output operations performed on training datasets and NN architectures. The prototype is aaccessible at this https URL. The analytical capabilities include a novel measurement of NN inefficiency using modified Kullback-Liebler (KL) divergence applied to histograms of NN model states, as well as a quantification of the sensitivity to variables related to data and NNs. Both NN Calculator and KL divergence are used to devise a trojan detector approach for a variety of trojan embeddings. Experimental results document desirable properties of the KL divergence measurement with respect to NN architectures and dataset perturbations, as well as inferences about embedded trojans.

Neural Network Calculator for Designing Trojan Detectors.

Abstract: This work presents a web-based interactive neural network (NN) calculator and a NN inefficiency measurement that has been investigated for the purpose of detecting trojans embedded in NN models. This NN Calculator is designed on top of TensorFlow Playground with in-memory storage of data and NN coefficients. Its been extended with additional analytical, visualization, and output operations performed on training datasets and NN architectures. The analytical capabilities include a novel measurement of NN inefficiency using modified Kullback-Liebler (KL) divergence applied to histograms of NN model states, as well as a quantification of the sensitivity to variables related to data and NNs. Both NN Calculator and KL divergence are used to devise a trojan detector approach for a variety of trojan embeddings. Experimental results document desirable properties of the KL divergence measurement with respect to NN architectures and dataset perturbations, as well as inferences about embedded trojans.