Publications by S F Pugh

Publications by authors named "S F Pugh"

J Pathol

February 2025

Machine learning, particularly deep learning with convolutional neural networks (CNNs), is being used to detect prostate cancer in tissue slides, but sample type differences affect model accuracy.
Research tested whether CNNs trained on one type of sample (biopsy or radical prostatectomy) could effectively analyze the other type, revealing a significant drop in performance across sample types.
Results indicated that models performed well on their own sample but poorly on the alternative type, highlighting the need to consider morphological differences in training to improve cancer detection accuracy in clinical settings.*

BMC Public Health

December 2024

The study examined the awareness, sanitation practices, and education efforts regarding zoonotic disease risks among employees at pet and animal feed stores that sell live animals.* -
Surveys of 206 employees revealed that handwashing and perceived disease risk were significantly higher in workplaces with specific hygiene policies and training.* -
Key factors influencing employees' likelihood to educate customers included supervisory roles, higher perceived disease risks, and the frequency of sanitation practices.*

JAMA Netw Open

November 2024

The study investigates how using small datasets to select an optimal cutoff score for the Patient Health Questionnaire-9 (PHQ-9) can lead to inaccurate results.
Researchers evaluated whether data-driven methods for cutoff selection resulted in scores that were significantly different from the true population optimal score and if these methods produced biased accuracy estimates.
Findings showed that many small studies frequently failed to identify the correct optimal cutoff score, particularly in smaller samples, leading to an overestimation of test sensitivity.

Chemphyschem

November 2024

Carbon dioxide capture technologies are crucial for addressing climate change, and solid-state O NMR spectroscopy can enhance the development of effective sorbent materials.
Researchers conducted static density functional theory NMR calculations to differentiate between bicarbonate, carbonate, and water species in hydroxide-based CO capture systems.
They propose a new workflow utilizing machine-learning force fields for dynamic modeling, which better aligns computational results with experimental findings, leading to insights into the binding mechanisms in metal-organic frameworks.

Neuro Oncol

November 2024

Glioblastoma (GBM) is a severe brain cancer that can lead to toxic side effects during treatment, prompting this study to explore genetic and clinical factors associated with vascular toxicities such as thrombosis and hypertension in patients.
A total of 591 Non-Hispanic White GBM patients were analyzed, with 62 experiencing thrombosis and 59 hypertension, revealing that hypertensive patients had improved survival rates compared to those without hypertension.
The study found that genetic factors significantly predicted hypertension better than clinical data alone, while corticosteroid use was identified as a notable risk factor for thrombosis, suggesting a need for further research into these associations in cancer treatments.