Skip to main content

Friday, April 9th, 2021

Philip Onffroy

Ray’cycle Initiative: Characterizing and Productizing Community-Sourced Plastics

Certain everyday plastic product waste ranging from grocery bags to bottle caps and single-use coffee pods cannot typically be recycled in the United States due to municipality and recycling plant regulations. Driven partially by the COVID-19 pandemic, this campus sustainability initiative establishes a new community means of collecting plastic waste materials and reprocessing them into products in an innovative fashion. Post-consumer plastics made of high-density polyethylene (HDPE), low-density polyethylene (LDPE), and polypropylene (PP) are processed using the novel solid-state/melt extrusion (SSME) technique, which has previously been proven to compatibilize polymer blends and commingled plastic waste. Post-consumer plastic materials are often contaminated, non-uniform, and therefore lower quality than virgin plastics. However, SSME has the potential to yield recycled plastic materials with properties comparable to relevant virgin plastic pellets. The mechanical property characterization of these recycled HDPE, LDPE, and PP materials by way of tensile testing and thermal characterization, such as thermogravimetric analysis and differential scanning calorimetry, are benchmarked against as-received plastics as well as virgin analogs. Additionally, these community-sourced plastic materials are made into useable tools and Bucknell memorabilia through injection molding as a sustainable end-use for the polymer material. This project showcases the actual recyclability of “difficult-to-recycle” plastic waste products while also making a broader impact to the local community through plastics recycling education and public sustainability awareness.

Continue reading »

Friday, April 9th, 2021

Nicholas Passantino

Domain Adaptation in Machine Learning for Medical Imaging​

Machine learning models do well in learning to classify images, but can undesirably learn features unique to the specific dataset they were trained on, and not to the desired content of the images themselves. In our work, we use various machine learning techniques to develop models that learn domain-invariant features between two popular datasets in medical imaging. These models take in images of the heart created via echocardiogram and output a segmentation of the images into the different components of the human heart. The desired end-goal is to increase the model’s accuracy on the secondary dataset while minimizing the decrease in accuracy on the initial dataset.​

Continue reading »

Friday, April 9th, 2021

Tung Tran

Bayesian Optimization of 2D Echocardiography Segmentation
Bayesian Optimization (BO) is a well-studied hyperparameter tuning technique that is more efficient than grid search for high-cost, high-parameter machine learning problems. Echocardiography is a ubiquitous modality for evaluating heart structure and function in cardiology. In this work, we use BO to optimize the architectural and training-related hyperparameters of a previously published deep fully convolutional neural network model for multi-structure segmentation in echocardiography. In a fair comparison, the resulting model outperforms this recent state-of-the-art on the annotated CAMUS dataset in both apical two- and four-chamber echo views. We report mean Dice overlaps of 0.95, 0.96, and 0.93 on left ventricular (LV) endocardium, LV epicardium, and left atrium respectively. We also observe significant improvement in derived clinical indices, including smaller median absolute errors for LV end-diastolic volume (4.9mL vs. 6.7), end-systolic volume (3.1mL vs. 5.2), and ejection fraction (2.6% vs. 3.7); and much tighter limits of agreement, which were already within inter-rater variability for non-contrast echo. These results demonstrate the benefits of BO for echocardiography segmentation over a recent state-of-the-art framework, although validation using large-scale independent clinical data is required.

Continue reading »

Friday, April 9th, 2021

Lucas Rankin

Establishing Independent Tunability of the Mechanical and Transport Properties of Polymer Gels

Polymer gels can be used in the fabrication of materials for filtering liquid and gaseous media, solid-state electrolytes, and transdermal medical patches. This diverse range of applications primarily relies on the transport and mechanical properties of polymer gels. Both sets of properties have shown excellent tunability, but typically in a coupled fashion. Establishing the independent tunability of the transport and mechanical properties of polymer gels (using simple, cost-effective methods) is paramount if polymer gels are to be used to their full potential. Specifically, block copolymer gels self-assemble into organized nanoscale networks within the gel solvent, which allows for facile control of material properties. Mechanical properties can be tuned by altering gel network connectivity, which does not have an effect on solute transport rate. Solute transport rate is affected by polymer concentration and solvent choice. Two formulation methods were used in this work to independently tune the mechanical and transport properties of block copolymer gels. Gel mechanical behavior was tuned independently of solute transport rate via exchanging triblock and diblock copolymers (to change network connectivity) at constant polymer concentration. Solute transport rate was tuned independently of mechanical behavior by editing solvent viscosity.

Continue reading »

Friday, April 9th, 2021

Yida Chen

Assessing the generalizability of temporally-coherent echocardiography video segmentation

Existing deep-learning methods achieve state-of-art segmentation of multiple heart substructures from 2D echocardiography videos, an important step in the diagnosis and management of cardiovascular disease. However, these methods generally perform frame-level segmentation, ignoring the temporal coherence in heart motion between frames, which is a useful signal in clinical protocols. In this study, we implement temporally consistent video segmentation, which has recently been shown to improve performance on the multi-structure annotated CAMUS dataset. We show that data augmentation further improves results, which are consistent with prior state-of-art works. Our 10-fold cross-validation shows that video segmentation improves the automatic comparison to clinical indices including smaller median absolute errors for left ventricular end-diastolic volume (6.4 ml), end-systolic volume (4.2 ml), and ejection fraction (EF) (3.5%). In segmenting key cardiac structures, video segmentation achieves mean Dice overlap of 0.93 on left ventricular endocardium, 0.95 on left ventricular epicardium, and 0.88 on left atrium. To assess clinical generalizability, we further apply the CAMUS-trained video segmentation models, without tuning, to a larger, recently published EchoNet-Dynamic clinical dataset. On 1274 patients in the test set, we obtain absolute errors of 6.3% ± 5.4 in EF, confirming the reliability of this scheme. In that the EchoNet-Dynamic videos contain limited annotation only for left ventricle endocardium, this effort extends at little cost generalizable, multi-structure video segmentation to a large clinical dataset.

Continue reading »