Electrical and Computer Engineering Research Open House Presentation Schedule

Poster ID: 50

2:00 - 3:00

Anjum, Md Fahim



Using Linear Predictive Coding for real-time detection of Parkinson’s Disease

Authors: Anjum, Md Fahim; Haug, Joshua; Alberico, Stephanie L.; Dasgupta, Soura; Mudumbai, Raghuraman; Kennedy, Morgan A.; Singh, Arun; Narayanan, Nandakumar S.; Cavanagh, James F.

Parkinson's disease (PD) is a neurodegenerative disorder that causes profound changes in cortical and subcortical brain activity. These changes can be measured by electroencephalography (EEG) and local field potentials (LFP). Advanced PD therapies such as adaptive deep-brain stimulation provide neural stimulations by analyzing such signals in real time. However, traditional methods for detecting PD with EEG and LFP are not very accurate and computationally expensive. We propose a novel feature based on Linear Predictive Coding (LPC) which is very efficient in capturing PD-related spectral characteristics in EEG and LFP. This LPC-based feature can be used to detect PD in real time.

Three-Minute Video: https://youtu.be/ToJ8sOlxWkw

Presentation: anjum_ece_vroh_2020.pdf


Poster ID: 51

11:00 - 12:00

Brzus, Michal



Miniature Pig Brain Segmentation Using Transfer Learning

Authors: Michal Brzus

Embedded commonly throughout biomedical imaging studies focusing on the human brain, machine learning algorithms have proven to be extremely powerful. Considering the extensive use of animals in medical research, there is surprisingly little work in the field for non-human species.

In this study, to achieve mini pig brain segmentation, I designed a pipeline that uses manually annotated landmarks on both human and mini pig data to transform the mini pig MR images into human brain space. Then pre-trained on human data Deep CNN is applied. Finally, the output is converted to the original mini pig space to obtain the segmentation.

Three-Minute Video: https://youtu.be/A5DDhau8raE

Presentation: brzus_ece_vroh_2020.pdf


Poster ID: 52

10:00 - 11:00

Finley, Matthew



Two-channel image-based compression of 3D range data using virtual plane encoding

Authors: Finley, Matthew G.; Bell, Tyler

Modern computing technology allows 3D range data to be acquired at speeds much faster than real-time, with sub-millimeter precision. However, this speed and precision results in an increased quantity of 3D data being generated, potentially limiting target applications. One approach to compressing 3D range data is to encode it within the three color channels of a traditional 24-bit RGB image. This paper presents a novel method for the modification and compression of 3D range data such that the original depth information can be stored within, and recovered from, only two channels of a traditional 2D RGB image.

Three-Minute Video: https://www.youtube.com/watch?v=HAClNXwNN-U

Presentation: finley_ece_vroh_2020.pdf


Poster ID: 53

10:00 - 11:00

Johnson, Chase



Preparing the PREDICT-HD dataset for public release through landmark detection and deidentifying data

Authors: Johnson, Charles

The PREDICT-HD study collected MRI brain scans from participants at risk for Huntington’s Disease from 1999-2016 with the purpose of identifying the disease with medical imaging before symptom expression. Nearing data publication, we have deidentified the data by shifting dates, replacing identifying words/phrases, and defacing images in the dataset to prevent a future researcher from identifying participants. We identified landmarks on each scan using our custom landmark detection software then visually inspected and adjusted any landmarks needing correction. Once published, this neuroimaging dataset will be available to the research community for use in further studies.

Three-Minute Video: https://youtu.be/zdq5D_gHHRI

Presentation: johnson_ece_vroh_2020.pdf


Poster ID: 54

12:00 - 1:00

Keefe, Daniel



Rapid and Inexpensive Biosensor for Sensitive and Selective Detection of COVID-19

Authors: Keefe, Daniel; Gao, Bingtao; Rojas Chavez, Anthony; Smith, Rasheid; Haim, Hillel; Salem, Aliasger; Toor, Fatima

Current methods for diagnosing infection of COVID-19 are too expensive (~$1500/test) and take too much time (~24 hours/test). In this presentation, we will present our most recent results on an inexpensive biosensor from vertically-oriented silicon nanowires (vSiNWs) for rapid (~10 min/test) COVID-19 infection detection. The vSiNWs are fabricated and turned into an electronic transducer utilizing nanofabrication processes. The vSiNWs are then conjugated with angiotensin converting enzyme 2 (ACE2) which binds with high affinity to the SARS-CoV-2 spike protein. Our testing shows that the biosensor is able to detect the spike protein in complex mixtures sensitively and specifically.

Three-Minute Video: https://youtu.be/-U1X0xjESsc

Presentation: keefe_ece_vroh_2020.pdf


Poster ID: 55

11:00 - 12:00

Kubicek, Bernice



Sonar target feature representation and classification using a two-dimensional Gabor wavelet

Authors: Kubicek, Bernice; Sen Gupta, Ananya; Kirsteins, Ivars

Sonar target recognition suffers from feature uncertainties due to unpredictable oceanic parameters and unknown target geometries. These effects may combine in a nonlinear fashion, making automatic target classification near impossible. We propose a feature extraction algorithm and representation using a two-dimensional Gabor wavelet as a kernel filter. The feature representation is validated by comparing the overall classification accuracy of a support vector machine, random forest tree, and neural network – trained and tested on unfiltered and filtered data. Results from experimental field data are presented.

This research is funded by the Office of Naval Research grant number N00014-19-1-2436.

Three-Minute Video: https://youtu.be/0dGVpkyTelk

Presentation: kubicek_ece_vroh_2020.pdf


Poster ID: 56

9:00 - 10:00

Le, Nam Hoang



LOGISMOS-JEI: Segmentation of Vessel Walls Associated with Brain Aneurysms

Authors: Le, Nam Hoang; Zhang, Honghai; Hasan, David; Samaniego, Edgar; Derdeyn, Colin; Koscik, Tim; Bathla, Girish; Roa Loor, Jorge; Sonka, Milan

Intracranial aneurysms are swollen parts of blood vessels filled with blood in the brain. In this project, we developed segmentation workflow for the detection of vessel walls which implements the LOGISMOS method for detection of vessel walls. LOGISMOS is a graph search method that optimally segments multiple n-D surfaces which can have mutual geometric relationships. This is achieved by first representing the interested structures by a set of columns filled with graph nodes, then the algorithm tries to find a set of node per columns whose total cost is minimal. The performance of the method and associated software was validated on CT perfusion brain images.

Three-Minute Video: https://youtu.be/uS5eReQUBE0

Presentation: le_ece_vroh_2020.pdf


Poster ID: 57

11:00 - 12:00

Liu, Xingxing



U-Net Based Spine Segmentation: In this project, we used a modified U-Net to get  good  spine segmentation results

Authors: Liu, Xingxing; Quang Tri; Askari Karchegani, Maziyar; Liu, Yang; etc

Deep learning has played an important role in medical image processing, especially in segmentation. Good and robust segmentation of medical image helps doctors to diagnosis diseases and conduct surgery. We used the well-known U-Net to segment spine images and achieve good performance.

Three-Minute Video: https://www.youtube.com/watch?v=-q5uLzo6TJ0

Presentation: liu_ece_vroh_2020.pdf


Poster ID: 58

12:00 - 1:00

McCarthy, Ryan



Tracking Multipath Structures in Shallow Water Acoustic Channels using Nonlinear Feature Representations

Authors: McCarthy, Ryan; Sen Gupta, Ananya

We investigate the limitations of geometric braids as a connected nonlinear feature representation to capture time-varying multipath activity for shallow water acoustic channels. These feature representations keep the identity of individual multipaths separate providing real-time continuous determination of dominant multipath within the channel delay spread. While braids do not offer unique representations for individual delay taps, they can track and adapt non-uniformly across estimated acoustic channels. Results will be shown and discussed through a few example underwater acoustic channels simulated using the BELLHOP model and a portion of the estimated channel from experimental field data collected during the SPACE08 experiment.

Three-Minute Video: https://youtu.be/keHiZtaYnXA

Presentation: mccarthy_ece_vroh_2020.pdf


Poster ID: 59

12:00 - 1:00

Pramanik, Aniket




Authors: Pramanik, Aniket; Jacob, Mathews

We introduce a fast model-based deep-learning (DL) approach for calibrationless Parallel Magnetic Resonance Image (PMRI) reconstruction. It is a non-linear extension of recent calibrationless structured low-rank methods for PMRI, called PSLR, that rely on linear relations in Fourier domain. The proposed scheme pre-learns non-linear relations in the Fourier domain from exemplar data. It is about three orders of magnitude faster than PSLR methods. A challenge with calibration-based methods is the potential for motion artifacts in images due to mismatches between the calibration and main scans. The proposed calibrationless strategy out-performs the calibrated model-based DL approach MoDL while avoiding mismatches.

Three-Minute Video: https://www.youtube.com/watch?v=hivfmGjwArU&feature=youtu.be

Presentation: pramanik_ece_vroh_2020.pptx


Poster ID: 60

1:00 - 2:00

Rendleman, Michael



Novel Feature Engineering Exploration of Cancer -Omics Data: Identifying Potential Biomarkers for Head and Neck Squamous Cell Carcinoma

Authors: Rendleman, Michael; Nwakama, Chibuzo; Braun, Terry; Casavant, Tom

Modern oncology has a growing wealth of -omics data. From DNA and RNA sequencing to methylation and somatic mutations, in aggregate these high-dimensional data types present a complex analysis challenge that requires computational methods to tease out useful oncological decision support information. In this poster, work towards a comparison of a diverse set of both novel and common feature engineering techniques for cancer biomarker discovery is presented. Machine learning techniques are integral to this analysis, applied at multiple stages for feature construction, selection, and evaluation. Ultimately, this work will produce a biomarker-based prognostic model for Head and Neck Squamous Cell Carcinoma.

Three-Minute Video: https://www.youtube.com/watch?v=ltIj1oe73Qk

Presentation: rendleman_cbcb_vroh_2020.pdf


Poster ID: 61

2:00 - 3:00

Rouabhi, Kawther



Autonomous detection and tracing of ion trails in the Martian ionosphere by exploiting spectral morphology and spatial geometry

Authors: Rouabhi, Kawther; Sen Gupta, Ananya; Halekas, Jasper

Detection, tracking, and isolation of ion trails in the Martian ionosphere against solar wind is a computational challenge. This is due to the varying shape of ion trails, overlap between the trails themselves, and trails buried within signatures of solar wind. We present an isolation application that autonomously detects and tracks ion trails in the Martian ionosphere using solar wind ion analyzer (SWIA) data from NASA’s MAVEN mission. Our technique involves analysis of energy spectrograms to extract trails exhibiting high signal-to-noise ratio levels. We provide results of our algorithm over energy spectrograms and metrics on the evolution of individual ion trails.

Three-Minute Video: https://youtu.be/t-VXVt-3Ois

Presentation: rouabhi_ece_vroh_2020.pdf


Poster ID: 62

2:00 - 3:00

Schwartz, Broderick



V3CS: Virtual 3D Capture Suite is a software platform for the simulation of 3D capture systems

Authors: Schwartz, Broderick; Yang, Bingdi; Bell, Tyler

Modern 3D sensing devices are used to provide accurate 3D measurements in many fields, such as forensics and security. The engineering processes required to develop or modify such 3D imaging systems are often quite time-consuming, and this is only exacerbated by COVID-19. Thus, we propose the Virtual 3D Capture Suite (V3CS) that will allow for the accurate simulation of real-world 3D capture pipelines and the rapid development of new 3D capture pipelines. Further, with its representative simulation, V3CS will also be able to automatically generate customized, labelled training data for the development and validation of new 3D-based machine learning algorithms.

Three-Minute Video: https://youtu.be/ogOuqF1FZs8

Presentation: schwartz_ece_vroh_2020.pdf


Poster ID: 63

12:00 - 1:00

Siemonsma, Stephen



Magnetic Resonance Fingerprinting Using Model-Based Deep Learning: Exploiting physics-based and deep-learned priors in a novel iterative reconstruction and quantification algorithm applied to simulated brain MRI data

Authors: Siemonsma, Stephen; Eldar, Yonina; Jacob, Mathews

Since its introduction, magnetic resonance fingerprinting (MRF) has proven itself to be a versatile and increasingly important quantitative MRI method. Traditional MRF dictionary matching techniques are quickly being superseded with convolutional neural network (CNN) approaches. However, most CNN approaches simply use the network as a black box. In this work, although we include a CNN in our algorithm, the data consistency steps ensure that our results remain reliable even at very high acceleration factors. Overall, we are proposing a novel model-based, unrolled algorithm that simultaneously restores the temporal profiles of the data and accurately estimates the underlying tissue parameter maps.

Three-Minute Video: https://youtu.be/_D_V6bRjt4g

Presentation: siemonsma_ece_vroh_2020.pdf


Poster ID: 64

11:00 - 12:00

Sindt, Jacob



Ray Optics Modeling to Determine the Effect of Light Trapping Coatings on Solar Cell Efficiency

Authors: Sindt, Jacob; Toor, Fatima

Abstract: Sunlight reflection off of solar cells inhibits maximum energy production, but with the use of light trapping coatings, an increase in energy production can be achieved. We will present results on 3D ray optics modeling performed using COMSOL to design textured light trapping coatings for solar cells. The modeling of the textured structures enables analysis of light concentration on the solar cell surface due to the coatings. Based on our modeling results, we will select the highest performing light concentrator design, 3D-print it in our lab, and test it with a silicon solar cell.

Three-Minute Video: https://youtu.be/mpUD4oA4EvU

Presentation: sindt_ece_vroh_2020.pdf


Poster ID: 65

1:00 - 2:00

Wang, Di



Fully-UNet: Unsupervised 3D End-to-End Medical Image Registration for Large Deformation

Authors: Wang, Di; Durumeric, Oguz; Reinhardt, Joseph; Christensen, Gary

We present a learning-based method for deformable image registration (DIR). Many published learning-based methods have shown promising results to predict the deformation vector field (DVF). However, these methods limit to small deformation tasks. To address this shortcoming, we develop lung DIR method using U-Net-like architecture, namely Fully-UNet, to capture large lung motion between inhale-exhale pulmonary CT images. The network is trained end-to-end by optimization of loss metric between pairs of 3D images. Evaluation was performed on public DIRLAB datasets. The results demonstrate that Fully-UNet has achieved excellent performance in terms of TRE and plausibility of DVF among learning-based methods.

Three-Minute Video: https://youtu.be/OCuablwgvZ8

Presentation: wang_ece_vroh_2020.pdf


Poster ID: 66

2:00 - 3:00

Zainab, Hunza



Improving collection dynamics by monotonic filtering

Authors: Zainab, Hunza, Audrito, Giorgio,  Dasgupta, Soura, Beal, Jacob

A key coordination problems in distributed open systems is distributed sensing, as achieved by cooperation and interaction among individual devices. An archetypal operation of distributed sensing is data summarization over a region of space, by which many higher level problems can be addressed, including counting items, measuring space, averaging environmental values, etc. A typical coordination strategy to perform data summarization in a peer-to-peer scenario, where devices can communicate only with a neighborhood, is to progressively accumulate information towards one or more collector devices, though this typically exhibits problems of reactivity and fragility. In this paper, we present a monotonic filtering strategy for improving the dynamics of single path collection algorithms. The strategy consists of inhibiting communication across devices whose distance towards the collector device is not decreasing. We prove that single path collection in a line graph results in quadratic overestimates after a source change and that these overestimates disappear with the application of monotonic filtering. These preliminary results suggest that monotonic filtering is likely to improve the dynamics of single path collection algorithms, by preventing excessive overestimates.

Three-Minute Video:

Presentation: zainab_ece_vroh_2020.pdf


Poster ID: 67

10:00 - 11:00

Zelenski, Sasha



Using SNR to aid in peak-cognizant signal processing to quantify environmental weathering of contaminants from the Deepwater Horizon spill

Authors: Zelenski, Sasha

As part of the Sen Gupta research group, I worked on developing an algorithm to analyze GC-MS data from the Deepwater Horizon Oil Spill and how various oil compounds break down. The algorithm is able to autonomously align the samples and analyze their peak profiles, allowing researchers to easily visualize data and identify which compounds weather more than others within the raw signal. This could help minimize the usage of harmful dispersants in future oil spills. My role in this project was helping with development of the peak filtering code using SNR and implementing techniques to identify and discard co-eluting peaks.

Three-Minute Video: https://youtu.be/wz8MSGfwGjQ

Presentation: zelenski_ece_vroh_2020.pdf


Poster ID: 68

10:00 - 11:00

Zhang, Lichun



Active Learning with FilterNet for Calf Muscle Compartment Segmentation

Authors: Zhang, Lichun; Zhihui, Guo; Honghai, Zhang; Eric Axelson; Daniel Thedens; Ellen van der Plas; Peg Nopoulos; Sonka, Milan

With the limited cost for annotation and computation, what instances should be traced to obtain the best performance? We address the question and present a deep active learning framework that combines a novel fully convolutional network (FCN), called FilterNet and active learning to try significantly reducing annotation effort but remains the best performance meanwhile.

Three-Minute Video: https://youtu.be/SIFVEnubbPU

Presentation: zhang_lichun_ece_vroh_2020.pdf


Poster ID: 69

10:00 - 11:00

Zhang, Xiaoliu



CT-Based Characterization of Transverse and Longitudinal Trabeculae and Its Applications

Authors: Zhang, Xiaoliu; Letuchy, Elena; Levy, Steven; Torner, James; Saha, Punam

Osteoporosis is characterized by reduced bone mineral density (BMD), micro-structural deterioration, and enhanced fracture-risk. There are compelling evidences suggesting that bone micro-structural quality is a strong determinant of bone strength and fracture-risk. Trabecular bone (Tb) consists of transverse and longitudinal microstructures, and there is a hypothesis that transverse trabeculae improve bone strength by arresting the buckling of longitudinal trabeculae. We present a new in vivo CT-based method for characterizing transverse and longitudinal trabeculae, evaluate their repeat CT scan reproducibility, and examine their links with gender, height, weight, and BMI.

Three-Minute Video: https://youtu.be/_3PjSACbCcE

Presentation: zhang_xiaoliu_ece_vroh_2020.pdf