Posters

  • Deep Learning Methods to Predict Cognitive Performance in Midlife

    August 29, 2020

    Deep Learning Methods to Predict Cognitive Performance in Midlife S Koorathota, P Sajda, G Liu, M Lachman, RP Sloan Psychosomatic Medicine 82 (2)

  • Impact of Reference Standard, Data Augmentation, and OCT Input on Glaucoma Detection Accuracy by CNNs on a New Test Set

    August 21, 2020

    Thakoor, K.A., Tsamis, E.M., De Moraes, C.G., Sajda, P., Hood, D.C. Impact of Reference Standard, Data Augmentation, and OCT Input on Glaucoma Detection Accuracy by CNNs on a New Test Set. Investigative Ophthalmology and Visual Science, 61(7), pp. 4540-4540, 2020.

  • Assessing the Ability of Convolutional Neural Networks to Detect Glaucoma from OCT Probability Maps

    February 28, 2020

    Thakoor, K.A., Zheng, Q., Nan, L., Li, X., Tsamis, E.M., Rajshekhar, R., Dwivedi, I., Drori, I., Sajda, P. and Hood, D.C., Assessing the Ability of Convolutional Neural Networks to Detect Glaucoma from OCT Probability Maps. Investigative Ophthalmology and Visual Science, 60(9), pp.1464-1464, 2019.

  • Simulation of 3D architectural and mechanical changes in human trabecular bone during menopause

    January 17, 2017

    The increase in bone remodeling after menopause is responsible for both reduced bone mass and deterioration of the trabecular bone network, leading to an increased susceptibility to osteoporosis. In this study, we simulated the dynamic and physiological process of trabecular bone remodeling, using parameters derived from recently published clinical data [1]. These simulations considered three types of microscopic bone loss in the pathophysiology of osteoporosis: perforation, breakage, and disconnection of trabeculae. The simulations were done using rigorous 3D digital topological analysis (DTA) [2]. The simulation included a bone remodeling cycle corresponded to a 200-day period (40 days resorption/160 days formation). Resorption cavities (42~tm deep and 126~tm in diameter) were created in 40 days according to the current activation frequency [1] and distributed randomly over the bone surface. Every resorption cavity was refilled in 160 days unless it caused a perforation, or breakage of a trabecula while disconnected trabeculae removed [2]. New resorption cavities were continuously created in every 40-days. The simulation was started 5 years before and ended 15 years after menopause and applied to twelve human trabecular bone samples. The time course of the averaged bone volume fraction (BVF) of samples from two anatomical sites (spine and femoral neck) showed great agreement with the corresponding clinical data [1]. Bone with higher initial BVF showed a slower bone loss time course, which was consistent with the fact that peak bone mass is one of the important risk factors in osteoporosis. The structural transition of each sample from plate-like to rod-like during menopause can be qualitatively observed and also quantitatively confirmed by the change of plate fraction. Furthermore, the results of this study suggest that the trabecular plate perforation accounts for more than 70% of bone loss during menopause.

  • Decoding fMRI with temporal integration: Learning the hemodynamical response function

    January 17, 2017

    There is a growing interest in employing multivariate methods for analyzing fMRI data, specifically as a way to exploit spatially distributed correlations linked to events/conditions of interest. Such approaches typically focus on learning spatial decompositions which optimize either a supervised or unsupervised objective function. However, fMRI is inherently a spatio-temporal signal and a principled approach should simultaneously find the spatial and temporal filters which optimize the objective of interest. [1] Bilinear logistic regression (BLR) has previously been applied for simultaneous learning of topographies and temporal envelopes in event-related EEG. [2] Here we present a version of BLR suitable for fMRI. The goal is to extract a spatial map of discriminating voxels and an associated hemodynamical integral for optimal inference about the experimental events (i.e. decoding).

  • Extraclassical Responses in V1 Modeled via Modulated Cortical Conductances

    January 17, 2017

  • Using Transcranial Magnetic Stimulation to Elucidate Interactions Between Top-down and Bottom-up Brain Networks in Visual Discrimination

    January 12, 2017