Debarshi Datta, PhD

When I started this project, vaccination was not available in the market. Front-line workers were engaged in saving the lives of patients. The trend of death was widespread throughout the world. Those who survived underwent critical events: they needed ventilation, ICU, or other measures. I endeavored to comprehend which features were essential for giving them proper management before they became severely sick. Even I tried to understand the best management plans for patients with COVID-19… Show more

When I started this project, vaccination was not available in the market. Front-line workers were engaged in saving the lives of patients. The trend of death was widespread throughout the world. Those who survived underwent critical events: they needed ventilation, ICU, or other measures. I endeavored to comprehend which features were essential for giving them proper management before they became severely sick. Even I tried to understand the best management plans for patients with COVID-19 before laboratory tests were done, which was difficult when the surge of patients was high. This research provides a shed of light to the clinician on how to properly manage patients during the pandemic. This research applies artificial intelligence and statistics to analyze hospital data and understand predictors of patients undergoing critical clinical events. Show less

I witnessed a death march in the USA in 2020. Schools, colleges, offices, and markets were shut down. I finished my Ph.D. during that crucial moment. I started this research last year—during my tenure as a postdoctoral researcher; I got this messy, unformatted, and unstructured dataset—having no direction on what I was supposed to do. Before vaccination became available, doctors were occupied with saving patients' lives rather than writing detailed medical information.
After understanding… Show more

I witnessed a death march in the USA in 2020. Schools, colleges, offices, and markets were shut down. I finished my Ph.D. during that crucial moment. I started this research last year—during my tenure as a postdoctoral researcher; I got this messy, unformatted, and unstructured dataset—having no direction on what I was supposed to do. Before vaccination became available, doctors were occupied with saving patients' lives rather than writing detailed medical information.
After understanding each variable, I developed the research question to find out how I could provide insight to frontline workers to gain advanced knowledge about patients who are susceptible to dying. As well, I thought I would find key elements that could contribute to COVID-related patient deaths. In addition, this study provides a solid insight to clinicians that they may be less inclined to rely on the patients' pathological reports—which makes decision-making and patient management more efficient when facilities are limited.
This is the first kind of study to examine the population of South Florida. This research is dedicated to all COVID patients who are no more and those who have experienced one or many episodes of COVID. This paper is also dedicated to those who have not responded to the scientific discovery of vaccine development.
I would appreciate it if all of you could read this paper. There are no scientific equations that are obscured to reach a larger audience. Show less

Patient care in emergency rooms can utilize urgency labeling to facilitate resource allocation. With COVID-19 care, one of the most important indicators of care urgency is the severity of respiratory illness. We present an early analysis of 5,584 patient records, of whom 5,371 (96.2%) have returned a positive COVID-19 test, to understand how well we can predict the severity of a respiratory illness given other features describing a patient using Deep Learning methods. The goal of our work is to… Show more

Patient care in emergency rooms can utilize urgency labeling to facilitate resource allocation. With COVID-19 care, one of the most important indicators of care urgency is the severity of respiratory illness. We present an early analysis of 5,584 patient records, of whom 5,371 (96.2%) have returned a positive COVID-19 test, to understand how well we can predict the severity of a respiratory illness given other features describing a patient using Deep Learning methods. The goal of our work is to illustrate the connection of our COVID-19 patient dataset with Deep Learning techniques, setting the stage for future work. The features in our dataset include when COVID-19 symptoms began, age, height, weight, demographics, and pre-existing conditions, to give a quick preview. We report train-test performance of a Deep Multi-Layer Perceptron (MLP) to predict the severity of respiratory analysis on a one-hot encoded scale of 5 labels. This 5-level scale is a truncation of our available labels, which we plan to extend and include in future work. We utilize a high-level of Dropout in order to avoid overfitting with our Deep Learning model. Further , we particularly study the impact of class imbalance on this dataset (Johnson and Khoshgoftaar 2019). We find that Random Oversampling (ROS) is an effective solution for decreasing minority class false negatives, as well as increasing overall accuracy. Readers will understand the performance of Deep Learning, with Dropout and ROS, to predict the severity of a COVID-19 pa-tient's respiratory illness in which patients are described with Tabular Electronic Health Records (EHR). Show less

Classic Gestalt examples of perceptual grouping entail arrays of disconnected surfaces that
are grouped on the basis of the surfaces’ relative similarity or proximity. However, most natural
environments contain multiple objects, each with multiple, connected surfaces. Moreover,
an object in a scene is likely to partially occlude other objects in the 2-dimensional
retinal projection of the scene. A central question, therefore, is how the visual system forms
a 3-dimensional… Show more

Classic Gestalt examples of perceptual grouping entail arrays of disconnected surfaces that
are grouped on the basis of the surfaces’ relative similarity or proximity. However, most natural
environments contain multiple objects, each with multiple, connected surfaces. Moreover,
an object in a scene is likely to partially occlude other objects in the 2-dimensional
retinal projection of the scene. A central question, therefore, is how the visual system forms
a 3-dimensional representation of multi-object scenes by determining which surfaces belong
to which objects. To this end, a recently developed dynamic grouping methodology determines
whether pairs of surfaces are grouped together on the basis of the direction in which
motion is perceived across a surface when its luminance is perturbed. It is shown using this
method that the visible surfaces of a partially occluded object are perceptually grouped
when they are plausibly connected and represented in a depth plane behind the occluding
object. Invisible connectivity (amodal completion) as well as connectivity established by a
visible surface have a powerful influence on the grouping of surfaces. However, for neither
kind of connectivity is grouping affected by the distance between the surfaces. This absence
of a distance/proximity effect on grouping is obtained when the space between to-begrouped
surfaces is filled with other surfaces. It contrasts with the strong effect of distance/
proximity on the grouping of disconnected surfaces, and on the clarity of illusory contours
formed between disconnected contours. It is concluded that distance/proximity is an operative
grouping variable only when there is empty space between the to-be-grouped surfaces. Show less

Objects in a scene are likely to occlude other objects partially and are itself likely
to be partially occluded. A central question, therefore, is how the visual system resolves
the resulting surface correspondence problem by successfully determining which surfaces
belong to which objects. To this end, a recently developed dynamic grouping
methodology has determined whether pairs of adjacent surfaces are grouped (Hock &
Nichols, 2012). The grouping of adjacent surfaces, which… Show more

Objects in a scene are likely to occlude other objects partially and are itself likely
to be partially occluded. A central question, therefore, is how the visual system resolves
the resulting surface correspondence problem by successfully determining which surfaces
belong to which objects. To this end, a recently developed dynamic grouping
methodology has determined whether pairs of adjacent surfaces are grouped (Hock &
Nichols, 2012). The grouping of adjacent surfaces, which depends on their affinity state,
is indicated by the direction of perceived motion across one surface when its luminance is
perturbed. In the current stimuli, which consists of a horizontal surface partially occluded
by a vertical bar, dynamic grouping also can occur for nonadjacent surfaces, providing
they are linked in two-dimensions by a connecting surface. Results indicate that the
dynamic grouping motion is stronger for amodal completion entailing the perceptual
grouping of nonadjacent surfaces behind an occluder. Show less

We present Language-Interfaced Fine-Tuning
(LIFT) in application to COVID-19 patient survival classification.
LIFT describes translating tabular Electronic Health Records
(EHRs) into text inputs for transformer neural networks.
We study LIFT with a dataset of 5,371 COVID-19 patients.
We focus on the predictive task of survival classification
utilizing demographic and medical history features. We begin
by presenting information about our dataset. We preface
our… Show more

We present Language-Interfaced Fine-Tuning
(LIFT) in application to COVID-19 patient survival classification.
LIFT describes translating tabular Electronic Health Records
(EHRs) into text inputs for transformer neural networks.
We study LIFT with a dataset of 5,371 COVID-19 patients.
We focus on the predictive task of survival classification
utilizing demographic and medical history features. We begin
by presenting information about our dataset. We preface
our investigation in text-based transformers by reporting the
performances of conventional machine learning models such
as Logistic Regression and Random Forest classifiers. We
also present the results of a few configurations of tabular
input-based Deep Multilayer Perceptron (MLP) networks. 86%
of the patients in our database survived in the measured time
window. Thus, predictive models are heavily biased to predict
that a patient will survive. We emphasize that this problem
of Class Imbalance was a major challenge in developing these
models. Our balanced sampling strategy from examples in
the majority and minority classes is crucial to achieving even
reasonable predictive performance. For this reason, we also
report performance based on Precision, Recall, and F-score
metrics, in addition to Accuracy. Having established baselines
with tabular inputs, we then shift our focus to the prompts
for translating from tabular to text inputs. We report the
performance of 5 prompts. The LIFT model achieves an F-score
on the held-out test set of 0.21, slightly behind the Deep MLP
with Tabular Features score of 0.23. Both models outperform
the Random Forest with Tabular Features at 0.15. We believe
that LIFT is a very exciting direction for machine learning
in healthcare applications because text-based inputs enables
us to take advantage of recent advances in Transfer Learning
and Retrieval-Augmented Learning. Show less

Datta, D., & Hock, H. (2018). The Perceptual Grouping of Non-Adjacent Surfaces: Invisible (amodal) and Visible Connectivity. Manuscript submitted for publication.