|
Sleep quality prediction using heart rate variability
We employed machine learning techniques to enhance sleep quality prediction by selecting more effective features
and reducing the dependability to the expert knowledge in feature computation.
We designed several sleep quality methodologies by applying a great range of machine learning techniques over
electronic health records, heart rate variability (HRV) features, and the raw Electrocardiogram (ECG) signals.
In the process of predicting sleep quality from HRV,
we came up with a clinical decision support system that processes the raw electrocardiogram signals
independently from the prior knowledge of sleep experts.
This system employs a convolutional neural network (CNN) to predict sleep quality
based on heart activities during each night by analyzing images of two ECG signals during Polysomnography studies.
To our knowledge, this is one of the first studies to predict sleep quality using HRV.
The detailed report of this research is provided in “Predicting sleep quality in osteoporosis patients using electronic health records and heart rate variability”,
which is presented in 42nd IEEE Engineering in Medicine and Biology Society (EMBC2020).
|
|
Identifying the disease-specific risk factors of daytime sleepiness
Abnormal sleep quality or sleep quantity causes excessive daytime sleepiness (EDS),
which is a highly prevalent condition in the older adult society.
EDS is a symptom of several diseases, such as neurological disorders, e.g. dementia, and sleep breathing disorders,
e.g. apnea. Distinguishing disease-specific risk factors of EDS can both reveal underlying reasons of abnormal sleep quality
and enhance its prediction. To do so, we investigated EDS risk factors in two groups of patients one group with
and one group without severe sleep apnea. The report of this research has been reflected in “Sleep Propensity and Sleep Apnea-Specific Hypoxia Are Associated with Excessive Daytime Sleepiness”
and presented at Annals of the American Thoracic Society (ATS2020).
|
|
Sleep quality prediction in caregivers of people with dementia
Most caregivers of people with dementia (CPWD) experience a high degree of stress due to the demands of providing care,
especially when addressing unpredictable behavioral and psychological symptoms of dementia.
Such challenging responsibilities make caregivers susceptible to poor sleep quality with detrimental effects on their overall health.
Hence, monitoring caregivers’ sleep quality can provide important CPWD stress assessment.
Most current sleep studies are based on polysomnography, which is expensive and potentially disrupts the caregiving routine.
To address these issues, we propose a clinical decision support system to predict sleep quality based on
trends of physiological signals in the deep sleep stage. This research has been elaborated on “Sleep quality prediction in caregivers using physiological signals” (Presentation slides / Source code)
and published in Computers in Biology and Medicine.
It is also featured in Using the E4 to assess sleep in caregivers of people with dementia.
|
|
Alcohol withdrawal prediction
By the definition of the World Health Organization, alcohol use disorder refers to any form of alcohol consumption that causes
health problems. Alcohol Withdrawal Syndrome (AWS) occurs roughly 4 to 72 hours following cessation or reduction of prolonged,
heavy alcohol ingestion. Withdrawal delirium or Delirium Tremens is considered the most dangerous symptom of AWS which can lead to
the death. During the initial 8 hours after the last drink, patients face with anxiety, insomnia, nausea, and abdominal pain.
This condition is followed by high blood pressure, increased body temperature, unusual heart rate, and confusion.
If this syndrome does not receive any treatment, the patients will suffer from hallucinations, fever, seizures, and agitation.
As a result, there is an essential need to predict and treat this syndrome in the initial stages.
The outcomes of this research are described in “Predicting alcohol withdrawal in intensive care units”
and presented at The Symposium of Student Research, Scholarship, and Creative Activities 2020.
|
|
Mortality Prediction
Early hospital mortality prediction is critical as intensivists strive to make efficient medical decisions
about the severely ill patients staying in intensive care units. As a result, various methods have been developed
to address this problem based on clinical records. However, some of the laboratory test results are time-consuming
and need to be processed. To address this issue, we proposed a novel method to predict mortality using features extracted from
the heart signals of patients within the first hour of ICU admission. In order to predict the risk, quantitative features
have been computed based on the heart rate signals of ICU patients. Outcomes of the experiments indicates that heart rate signals
can be used for predicting mortality in patients in the ICU, achieving a comparable performance with existing predictions that rely
on high dimensional features from clinical records which need to be processed and may contain missing information.
The outcomes of this research are reported in “Early Hospital Mortality Prediction using Vital Signals” (Presentation slides / Source code)
and has been accepted in IEEE/ACM CHASE 2018 and published in the journal of Smart Health.
|
|
Tor structure characterization
Tor is among most well-known dark net in the world. It has noble uses, including as a platform for free speech and information
dissemination under the guise of true anonymity, but may be culturally better known as a conduit for
criminal activity and as a platform to market illicit goods and data.
Past studies on the content of Tor support this notion, but were carried out by targeting popular domains
likely to contain illicit content. A survey of past studies may thus not yield a complete evaluation of the content
and use of Tor. This work addresses this gap by presenting a broad evaluation of the content of the English Tor ecosystem.
The outcomes of this research are reported in “A Broad Evaluation of the Tor English Content Ecosystem”
and “Interaction of Structure and Information on Tor”
is presented in "10th ACM Conference on Web Science" and "COMPLEX NETWORKS 2020", respectively.
|
|
Web Robot Detection
The accurate detection of web robot sessions from a web server log is essential to take accurate traffic-
level measurements and to protect the performance and privacy of information on a Web server. Moreover,
the irrecoverable risks of visits from malicious robots that intentionally try to evade web server
intrusion detection systems, covering-up their visits with fabricated fields in their http r
est packets,
cannot be ignored. To separate both types of robots from humans in practice, analysts turn to heuristic
methods or state-of-the-art soft computing approaches that have only been tuned to the specification of
a kind of web server. Noting that the landscape of web robot agents is ever changing, and that behavioral
patterns and characteristics vary across different web servers, both options are lacking. To overcome this
challenge, my colleagues and I proposed several methods based on Fuzzy Rough Set, Markov Clustering, Self-organizing Map concepts.
The report of this research is reflected in “A soft computing approach for benign and malicious web robot detection” (Source code) and
“Detection of Web site visitors based on fuzzy rough sets” published in Expert Systems with Applications journals (Impact factor: 3.928) and Soft Computing (Impact factor: 2.472), respectively.
|
|
Outlier Detection
Event handlers have wide range of applications
such as medical assistant systems and fire suppression systems.
These systems try to provide accurate responses based
on the least information. Support vector data description
(SVDD) is one of the appropriate tools for such detections,
which should handle lack of information. Therefore, many
efforts have been done to improve SVDD. Unfortunately,
the existing descriptors suffer from weak data characteristic
in sparse data sets and their tuning parameters are organized
improperly. These issues cause reduction of accuracy in event
handlers when they are faced with data shortage. Therefore,
my colleagues and I proposed several methods based on Fuzzy Rough Set, Bat Algorithm, and Chaos theory.
The results of this research is presented in “Weighted support vector data description based on chaotic bat algorithm” and
“Automatic support vector data description” (Source code) published in Applied Soft Computing journals (Impact factor: 3.541) and Soft Computing (Impact factor: 2.472), respectively.
|
|
Dynamic Facility Location
Determination of facilities, such as factories or
warehouses, location and availability conditions is one of the
important and strategic decisions for an organization to make.
Transportation costs that form a major part of goods price are
dependent to this decision making. There are verity of methods
have been presented to achieve the optimal locations of these
facilities which are generally deterministic. In real world accurate
estimation of the effective parameters on this optimal location for
single or multiple time periods is difficult and merely impossible.In
this research, my colleagues and I tried to achieve an efficient model with
consideration of uncertainty demand over different time periods on
the basis of previously presented models that we call stochastic
dynamic facilities location problem. In order to do so we use
stochastic constrain programming which convert the stochastic
model to a deterministic one. The results of this research is presented in “Dynamic Facility Location with Stochastic Demand” and
Stochastic Facilities location Model by Using Stochastic Programming” published in Shiraz Journal of System Management.
|
|
Job Scheduling
Nowadays the requests of managers and other
persons who benefited by such projects for drop total
project’s cost have increased considerably. Besides,
the amount of changes exert on this factors can result
to variation in initial estimation of them. In this way,
however; by observing this types of changes in
different circumstances, the ideal quality for projects
is going to be considered. Moreover, generally in
realty world, either falling in time consuming or the
amount usage of resources for a task could not lead
to decline of task quality. As a result, assigning fixed
and deterministic values for assessing ideal quality
leading to unpredictable outcomes. I this study,
fuzzy logic developed remarkably to measure the
quality for wide range of tasks and activities in
various variation circumstances. At last, the
presented model apply to real case study and the
obtained values have proved the efficiency of
proposed model in comparison to others in
deterministic situations. The report of this research is reflected in
“Solving the equilibrium problem of time, cost, resource and quality of project network by using expanded fuzzy logic set”.
|