Filtering by
- All Subjects: Machine Learning
- All Subjects: Biomaterials
- Creators: Harrington Bioengineering Program
Colorimetric assays are an important tool in point-of-care testing that offers several advantages to traditional testing methods such as rapid response times and inexpensive costs. A factor that currently limits the portability and accessibility of these assays are methods that can objectively determine the results of these assays. Current solutions consist of creating a test reader that standardizes the conditions the strip is under before being measured in some way. However, this increases the cost and decreases the portability of these assays. The focus of this study is to create a machine learning algorithm that can objectively determine results of colorimetric assays under varying conditions. To ensure the flexibility of a model to several types of colorimetric assays, three models were trained on the same convolutional neural network with different datasets. The images these models are trained on consist of positive and negative images of ETG, fentanyl, and HPV Antibodies test strips taken under different lighting and background conditions. A fourth model is trained on an image set composed of all three strip types. The results from these models show it is able to predict positive and negative results to a high level of accuracy.
The goal of this research project is to create a Mathcad template file capable of statistically modelling the effects of mean and standard deviation on a microparticle batch characterized by the log normal distribution model. Such a file can be applied during manufacturing to explore tolerances and increase cost and time effectiveness. Theoretical data for the time to 60% drug release and the slope and intercept of the log-log plot were collected and subjected to statistical analysis in JMP. Since the scope of this project focuses on microparticle surface degradation drug release with no drug diffusion, the characteristic variables relating to the slope (n = diffusional release exponent) and the intercept (k = kinetic constant) do not directly apply to the distribution model within the scope of the research. However, these variables are useful for analysis when the Mathcad template is applied to other types of drug release models.
Polymeric nanoparticles (NP) consisting of Poly Lactic-co-lactic acid - methyl polyethylene glycol (PLLA-mPEG) or Poly Lactic-co-Glycolic Acid (PLGA) are an emerging field of study for therapeutic and diagnostic applications. NPs have a variety of tunable physical characteristics like size, morphology, and surface topography. They can be loaded with therapeutic and/or diagnostic agents, either on the surface or within the core. NP size is an important characteristic as it directly impacts clearance and where the particles can travel and bind in the body. To that end, the typical target size for NPs is 30-200 nm for the majority of applications. Fabricating NPs using the typical techniques such as drop emulsion, microfluidics, or traditional nanoprecipitation can be expensive and may not yield the appropriate particle size. Therefore, a need has emerged for low-cost fabrication methods that allow customization of NP physical characteristics with high reproducibility. In this study we manufactured a low-cost (<$210), open-source syringe pump that can be used in nanoprecipitation. A design of experiments was utilized to find the relationship between the independent variables: polymer concentration (mg/mL), agitation rate of aqueous solution (rpm), and injection rate of the polymer solution (mL/min) and the dependent variables: size (nm), zeta potential, and polydispersity index (PDI). The quarter factorial design consisted of 4 experiments, each of which was manufactured in batches of three. Each sample of each batch was measured three times via dynamic light scattering. The particles were made with PLLA-mPEG dissolved in a 50% dichloromethane and 50% acetone solution. The polymer solution was dispensed into the aqueous solution containing 0.3% polyvinyl alcohol (PVA). Data suggests that none of the factors had a statistically significant effect on NP size. However, all interactions and relationships showed that there was a negative correlation between the above defined input parameters and the NP size. The NP sizes ranged from 276.144 ± 14.710 nm at the largest to 185.611 ± 15.634 nm at the smallest. In conclusion, the low-cost syringe pump nanoprecipitation method can achieve small sizes like the ones reported with drop emulsion or microfluidics. While there are trends suggesting predictable tuning of physical characteristics, significant control over the customization has not yet been achieved.
Although relatively new technology, machine learning has rapidly demonstrated its many uses. One potential application of machine learning is the diagnosis of ailments in medical imaging. Ideally, through classification methods, a computer program would be able to identify different medical conditions when provided with an X-ray or other such scan. This would be very beneficial for overworked doctors, and could act as a potential crutch to aid in giving accurate diagnoses. For this thesis project, five different machine-learning algorithms were tested on two datasets containing 5,856 lung X-ray scans labeled as either “Pneumonia” or “Normal”. The goal was to determine which algorithm achieved the highest accuracy, as well as how preprocessing the data affected the accuracy of the models. The following supervised-learning methods were tested: support vector machines, logistic regression, decision trees, random forest, and a convolutional neural network. Each model was adjusted independently in order to achieve maximum performance before accuracy metrics were generated to pit the models against each other. Additionally, the effect of resizing images on model performance was investigated. Overall, a convolutional neural network proved to be the superior model for pneumonia detection, with a 91% accuracy. After resizing to 28x28, CNN accuracy decreased to 85%. The random forest model performed second best. The 28x28 PneumoniaMNIST dataset achieved higher accuracy using traditional machine learning models than the HD Chest X-Ray dataset. Resizing the Chest X-ray images had minimal effect on traditional model performance when resized to 28x28 or larger.