AWARDED Grants

AwardED Gratns (FACULTY)

Brain Imaging Network Data Repository (BINDR)

Received award: $14,547
From: WKU Office of  Sponsored Programs

Dr. Zhuhadar was awarded an RCAP fund ($14,547) to cover the hardware expenses of Stage I of BINDR. The ultimate aim of BINDR is to develop an App that assists radiologists in classifying and detecting the type of brain disease associated with newly scanned (X-ray, CT, or MRI) brain images by using artificial intelligence (AI) algorithms. This fund will help Dr. Zhuhadar build a prototype that forms the necessary groundwork to establish regional, multidisciplinary research between the WKU Center for Applied Data Analytics and the University of Louisville Knowledge Discovery & Web Mining Lab. 

The proposed research collaboration between institutions (refer to Stage III in Figure 1) will enable the growth of a multi-tiered and multidisciplinary network of researchers, radiologists, and computer scientists with the end goal of facilitating integrative and interactive research. This network will match the data sources, data structures, and analysis needs of the radiologist community with the current advances in artificial intelligence algorithms from data mining research, and biomedical research to advance radiology research to meet the increasing challenges of understanding human brain disease. The proposed research can be characterized as research in computational neuroscience – work highly fundable by the NSF. Success in this project will enable us to bring external contracts and funds through research in IIBR , BDI , and technologies at a larger scale. 

Background: In 2012, Dr. Zhuhadar and collaborators from WKU & UofL researched Brain MRI classifications. While the research findings were published in the IEEE Conferences [1, 2], the results were limited. The state-of-the-art of the AI was inefficient at that time to conduct accurate image recognition in real time and on a vast number of images. However, in recent years, the breakthrough of AI helped in solving scientific challenges. One of these algorithms that captured the scientific community is the advancement of AI – Deep Learning. Deep Learning (DL) is viewed to be computationally and statistically efficient to deal with a vast number of images [3]. AI (DL) has also had a significant impact in historically difficult areas of research, such as in self-driving cars, fraud detections, and other applications that provide digital assistants (e.g., Amazon's Alexa, Apple's Siri, Google's Now).

AI (DL) has been used for several decades, but some key elements (software and hardware) that permit learning to persist had not been discovered until recently. From the hardware perspective, the recent development of a graphics processing unit (GPU) has become one of the most crucial computing technologies.  GPU is designed for parallel processing and used in various applications, including graphics and video rendering. Although they’re best known for their capabilities in gaming, GPUs are becoming more popular for use in artificial intelligence. More notably, AI (DL) is viewed to be computationally efficient in classifying images accurately. 

Significance: Since 2018, the Society for Imaging Informatics in Medicine (SIIM), the Society of Thoracic Radiology (STR), and the Radiological Society of North America (RSNA) are collaborating on advancing the field of AI in medicine. It is predicted [4] that "AI-enabled medical imaging solutions market to reach 4,720 Million by 2027."

Currently, there are some platforms similar to BINDR available in the market. The most widely used platforms today are Alzheimer’s Disease Neuroimaging Initiative  (ADNI), The Cancer Imaging Archive Net  (TCIA) repository, Medical Image Database , Brain Imaging Data Structure (BIDS) , and MR brain . These platforms host a vast archive of medical images for public downloads. However, what makes BINDR unique is the use of AI algorithms to identify to which collection (disease) this specific (X-ray, CT, or MRI) image belongs and automatically classify it. BINDR will provide a diagnostic record (disease type) associated with the image.  

Intellectual Merit:  Although numerous pioneering works have applied AI (Deep Learning) in detecting brain tumors, classifying a cancer's type, and segmenting MRI images [5-8],  to date, there is no real-time Artificial Intelligence application to classify and detect the disease type based on medical images. 

The plan is to program an application to classify the brain (mix of X-ray, CT, and MRI) images collected in Stage-I. It uses a convolutional neural networks (CNNs) algorithm, which in recent years proved to be the best method for image classification [17, 18]. 

Each brain image (2D image) will be imported into the program, as shown in Figure 2. The input layer consists of the raw pixel values of the brain images. Let the image be denoted by A, the 2D filter of size m× n  be denoted by K, and the 2D feature map be denoted by F. The next step in this program is to convolute each image A with a specific filter K and creates the feature map F. This convolution operation is denoted by A*K and is mathematically given as follows,

F(i,j )= (A * K)(i,j )=∑_m▒∑_n▒〖A(m,n)K(i - m,j - n)〗…(eq1).

The convolution operation is commutative, so we can write Eq.2 [19] as follows, 

F(i,j )= (A * K)(i,j )=∑_m▒∑_n▒〖A(i - m,j - n)K(m,n)〗…(eq2).

The program will flip the kernel K relative to the input. Next, it will compute the inner product of the filter at every location in the image (eq.2.) This is followed by operating a sequence of different layers to accomplish various tasks. Finally, the program will feed the output to a classifier. She plans to experiment with softmax and support vector machine classifiers and test the program with vast numbers of images while changing the number of layers convolution, pooling, and fully connected [19], as shown in Figure 2. 

Here are some previous successes in similar platforms:

Between 2016 and 2019, Dr. Zhuhadar developed a cloud-based semantically enriched decision support system to investigate the Offshore Panama Papers leaks databases. She used Amazon Web Services to host GraphDB server and linked it to ~2.7 billion DBPedia/GeoNames nodes/entities (refer to [20, 21]).

Between 2005 and 2015, she developed HyperManyMedia (HMM) platform. HMM is the 1st WKU MOOC platform. Besides, it is considered as the first Worldwide recommender system for semantically enriched open learning resources. By the end of 2015, there were more than 800,000 students enrolled in HMM. HMM contributed to over 30 articles, most of them are indexed in ACM and IEEE digital libraries [22-35], more details are provided in Bio-sketch.

References or Works Cited

[1] L. Zhuhadar and G. C. Nutakki, "Hybrid appearance based disease recognition of human brains," in 16th International Conference on Information Visualisation (IV), 2012: IEEE, pp. 588-598. 

[2] G. C. Nutakki, L. Zhuhadar, and R. Wyatt, "Appearance based Disease Recognition of Human Brains," in BIOINFORMATICS, 2012, pp. 351-355. 

[3] Y. Bengio, I. Goodfellow, and A. Courville, Deep learning. Citeseer, 2017.

[4] BioSpace. "AI-Enabled Medical Imaging Solutions Market." BioSpace. https://www.biospace.com/article/ai-enabled-medical-imaging-solutions-market-to-reach-usd-4-720-6-million-by-2027-growing-at-a-cagr-of-31-3-percent-says-emergen-research/?keywords=%22COVID-19%22+OR+Coronavirus 

[5] A. H. Marblestone, G. Wayne, and K. P. Kording, "Toward an Integration of Deep Learning and Neuroscience," (in English), Frontiers in Computational Neuroscience, Hypothesis and Theory vol. 10, no. 94, 2016-September-14 2016, doi: 10.3389/fncom.2016.00094.

[6] M. Havaei et al., "Brain tumor segmentation with deep neural networks," Medical image analysis, vol. 35, pp. 18-31, 2017.

[7] Y. Han, L. Sunwoo, and J. C. Ye, "k-space deep learning for accelerated MRI," IEEE transactions on medical imaging, 2019.

[8] R. Anderson, H. Li, Y. Ji, P. Liu, and M. L. Giger, "Evaluating deep learning techniques for dynamic contrast-enhanced MRI in the diagnosis of breast cancer," in Medical Imaging 2019: Computer-Aided Diagnosis, 2019, vol. 10950: International Society for Optics and Photonics, p. 1095006. 

[9] P. Riva-Posse et al., "A connectomic approach for subcallosal cingulate deep brain stimulation surgery: prospective targeting in treatment-resistant depression," Molecular psychiatry, vol. 23, no. 4, p. 843, 2018.

[10] S. Chen et al., "Near-infrared deep brain stimulation via upconversion nanoparticle–mediated optogenetics," Science, vol. 359, no. 6376, pp. 679-684, 2018.

[11] R. P. Duncan, L. R. Van Dillen, J. M. Garbutt, G. M. Earhart, and J. S. Perlmutter, "Physical therapy and deep brain stimulation in Parkinson’s Disease: protocol for a pilot randomized controlled trial," Pilot and feasibility studies, vol. 4, no. 1, p. 54, 2018.

[12] T. Kunath et al., "Are PARKIN patients ideal candidates for dopaminergic cell replacement therapies?," European Journal of Neuroscience, vol. 49, no. 4, pp. 453-462, 2019.

[13] M. V. Cherkasova et al., "Dopamine replacement remediates risk aversion in Parkinson's disease in a value-independent manner," Parkinsonism & related disorders, 2019.

[14] M. Stacy, "Treatment of Dopamine Dysregulation Syndrome and Impulse Control Disorders in Parkinson’s Disease," in Therapy of Movement Disorders: Springer, 2019, pp. 125-127.

[15] T. Morita, M. Asada, and E. Naito, "Contribution of neuroimaging studies to understanding development of human cognitive brain functions," Frontiers in Human Neuroscience, vol. 10, p. 464, 2016.

[16] P. Tuite, "Brain magnetic resonance imaging (MRI) as a potential biomarker for Parkinson’s disease (PD)," Brain sciences, vol. 7, no. 6, p. 68, 2017.

[17] X. Lv, D. Ming, Y. Chen, and M. Wang, "Very high resolution remote sensing image classification with SEEDS-CNN and scale effect analysis for superpixel CNN classification," International journal of remote sensing, vol. 40, no. 2, pp. 506-531, 2019.

[18] A. S. Lundervold and A. Lundervold, "An overview of deep learning in medical imaging focusing on MRI," Zeitschrift für Medizinische Physik, vol. 29, no. 2, pp. 102-127, 2019.

[19] M. A. Wani, F. A. Bhat, S. Afzal, and A. I. Khan, "Introduction to Deep Learning," in Advances in Deep Learning: Springer, 2020, pp. 1-11.

[20] L. P. Zhuhadar and M. Ciampa, "Novel findings of hidden relationships in offshore tax-sheltered firms: 

a semantically enriched decision support system," Journal of Ambient Intelligence and Humanized Computing, journal article August 10 2019, doi: 10.1007/s12652-019-01392-1.

[21] L. Zhuhadar and M. Ciampa, "Leveraging learning innovations in cognitive computing with massive data sets: Using the offshore Panama papers leak to discover patterns," Computers in Human Behavior, 2017/12/28/ 2017, doi: https://doi.org/10.1016/j.chb.2017.12.013.



The Impact of AmeriCorps Members on Invasive Species and Wildfire Fuels Mitigation 

Received award: $156,000 

From: the Corporation for National  & Community Service 

Millions of acres of public lands in Montana, Idaho, North Dakota, South Dakota, and Wyoming are imperiled by invasive species and wildfires, which reduce ecosystem health, decrease the productivity of public lands for wildlife, diminish recreation assets, and threaten the safety of communities. There is a need to manage public lands to sustain ecosystem services and enhances human safety. Agencies responsible for managing these lands, including the Forest Service, National Park Service, BLM, State Parks, and others, lack the capacity to complete such work and rely on the intervention of AmeriCorps members to mitigate the spread of invasive weeds and reduce wildfire fuel loads that threaten communities and habitat. For this evaluation, Montana Conservation Corps (MCC) is collaborating with a cohort of conservation corps (Public Lands Service Coalition (PLSC) to conduct a national evaluation. 

Dr. Allie McCreary (PI) and Dr. Lily Popova Zhuhadar (CoPI) from Western Kentucky University were awarded ($156,000) from the Corporation for National and Community Service to conduct this research. 

On August 22, 2022, Co-PI Dr. Zhuhadar presented this research at the 31st International Scientific Conference in Burgas, Bulgaria. In this presentation, Dr. Zhuhadar introduced the method (Before-After-Control-Impact Quasi-experimental Design) used to assess the impact of AmeriCorps members on natural and recreation resource enhancement. These results demonstrate members’ efficacy in improving trails and habitats on public lands, enhancing recreational access and experiences, reducing fire risk, and decreasing invasive plant cover resulting in improved ecological conditions for native species. 

AwardED Gratns (FACULTY - StuDENT)  Collaboration

Think Pandemic-Related Fraud Is Going Away? Think Again! 

Abstract: According to the Association of Certified Fraud Examiners (ACFE) and Grant Thornton, 51% of anti-fraud experts surveyed higher levels of fraud appeared since the pandemic began, with a 20% significant increase.  71% envision the level of fraud impacting businesses and employers to increase over the next year.

Financial fraud and identity theft are two significant risks all consumers face when they use an electronic transaction to make a purchase. This research investigates irregularities in historical data to prove that the enterprise is not maintaining its accounting system in agreement with its procedures. Jonah plans to present a research study utilizing machine learning algorithms to detect fraud in transactional datasets within this context. To this end, he acquired a dataset from SAS Institution. This dataset consists of 2,562 transactional records. For privacy and security reasons, the dataset he obtained from SAS Institution is anonymous data.

FUSE Award #: 22-FA238

Improving Ecosystem Quality Through Data Mining Application

Lily Zhuhadar, Kirk Atkinson, Albert Meier, and Ouida Meier 

Abstract. In this presentation, we describe how the WKU Initiative for Applied Data Analytics (ADA) can play a crucial role in helping YOU as a business organization or an environmental research entity make sense of YOUR data and how to best utilize it. For example, we will present our first grant proposal, “Designing and Implementing a Cloud-based Repository for the WKU Green River Preserve: Moving from Entrenched Data Structure to Semantic Web.” The ADA Initiative recently offered its Data Mining and Predictive Analytics research services to WKU Green River Preserve (GRP). Dr. Albert Meier (Executive Director of the GRP) and Dr. Ouida Meier have decades of experience and dedicated efforts to host numerous projects focused on the Preserve or include it as a study site in a larger project. Through this grant proposal, we worked with Dr. Meier to capture, organize, store, and release various datasets accumulated and growing at an accelerated rate at the Green River Preserve (GRP). In this presentation, we will describe what Data Analytics can do for GRP and or for YOU!    

OSP Award #: SP240      

Risk Analysis: Profiling Customers 

Abstract: Lending Club is a peer-to-peer marketplace where borrowers and investors are matched together. The goal of the Lending Club is to reduce the costs associated with these banking transactions and make borrowing less expensive and investment more engaging. Lending Club provides data on loans approved and rejected since 2007, including the assigned interest rate and the type of loan. To help the Lending Club produce effective matches between borrowers and investors, we looked at what factors are weighed heavier than others when determining if the borrowers have been accepted or rejected for loans. In determining which factors impact the loan status, the Lending Club could narrow down its borrowers by these factors to provide the best matches cost-effectively. We used the K-means clustering algorithm and regression models to model the results and determine these characteristics. 

FUSE Award #: 20-SP250              

Target Marketing: Predicting Pregnant Customers Using Analytics

Abstract: Retailmart wanted to know which of their sampled customers are most likely to be predicted to be pregnant. Through preliminary data mining research,  I could indicate which customers in the given sample data are most likely to be pregnant based on their current purchases. The attributes that proved to be the top predictors are Folic Acid, Prenatal Yoga, Prenatal Vitamins, Birth Control, Maternity Clothes, and Ginger Ale. This analysis can save Retailmart thousands of dollars in advertising costs. Instead of sending out coupons to thousands of people, they can now narrow their Marketing Campaign to the most likely to respond. Predictive analytics can go so much further than predicting pregnancy; it can be used to maximize the marketing campaign’s effectiveness while simultaneously aiding in increasing profits and saving the company money. Using the customers’ purchase history dataset, I first used the Logistic Regression model to get a Lift Chart. Also, I determined which variables were unnecessary to the prediction and excluded them from my determining model. Finally, I presented the most influential predicting attributes in this case study. 

FUSE Award #: 20-SP250              

Using Machine Learning Algorithms to Detect Financial Fraud

Abstract: 

New fraud detection methods are constantly evolving using big data, stream computing, and machine learning technologies (Vona, 2017). Companies such as SAS and IBM are heavily investing in building real-time fraud detection solutions for organizations, banks, and governments. While IBM announced that its solution could detect online fraud incidents before they occur, this type of solution is costly (Scuotto, Ferraris, & Bresciani, 2016). In our proposed research study, I do not consider a real-time solution. Still, I examine historical transaction data from SAS Institute to build a model that can detect fraudulent patterns. Hence, this research investigates irregularities in historical data to prove that the enterprise is not maintaining its accounting system in agreement with its procedures.

More specifically, the research process starts with the theoretical consideration known as Benford’s law (the law of anomalous numbers). This law was introduced in 1938 by Frank Benford to detect some types of patterns in numbers. In short, Benford’s law assumes a relationship between the frequency of a specific digit and its magnitude. This phenomenon can be used in accounting data to detect fraud. Using Benford’s law, I can see whether or not a customer (or a sales manager) exhibits a behavior (over time) that corresponds with the normal pattern of behavior.

FUSE Award #: 19-SP263

Determining the Effect and Detriment through Sentiment Analysis

Abstract: 

Computerized text analysis can often pick up these linguistic features better than humans. This project proposes utilizing sentiment analysis better to understand the relationship between language and school shootings. This research aims to analyze the semantics surrounding terrorism, including nonpolitical attacks, such as school shootings. I believe additional research is necessary for looking at the people’s sentiment following and preceding an attack. GDP does not measure the isolation, insecurity, and despair that occurs after an attack immediately; it only tracks the movement of capital. The effects of terrorism are much larger than that. In this research, I plan to (1) perform sentiment analyses on textual data surrounding terrorist attacks (i.e., Twitter, Facebook) and (2) determine the affective state of the sentences: whether or not the messages contain positive, neutral, or negative sentiment concerning the attacks.

FUSE Award #: 18-SP280              

Discovering pathways to success in business majors

Kyler Hart was awarded a FUSE Grant ($3,000). Dr. Zhuhadar mentored Kyker on this research. 

FUSE Award #: 18-SP265              

Too big to fail? Or too blind to see? Profit Analysis

Zach Ross was awarded a FUSE grant ($3,000). Dr. Zhuhadar mentored Zach on this research. 

FUSE Award #: 16-SP280              

Using predictive analytics to better enhance consumer lending 

Corey Travis  was awarded a FUSE grant ($3,000). Dr. Zhuhadar mentored Corey on this research. 

FUSE Award #: 16-SP274              

Who Wants My Product? Affinity-Based Marketing

Cody Kirk was awarded a FUSE grant ($3,000). Dr. Zhuhadar mentored Cody on this research. 

FUSE Award #: 15-SP278