Yifei Wang

Yifei Wang

Auto-generated profile, based on publicly available information

Yifei Wang

Department of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, CA, USA.

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
9 November 2021 in Radiology
Verified
Journal Article
Closed access
RSRebecca Smith-Bindman
Rebecca Smith-Bindman
SYSophronia Yu
Sophronia Yu
YWYifei Wang
Yifei Wang
MKMarc D. Kohli
Marc D. Kohli
PCPhilip Chu
Philip Chu
ABAndrew Bindman
Andrew Bindman
A framework that assigns CT examinations into broad categories based on body region and image quality requirements simplifies assessment of radiation dose for benchmarking and for developing CT pro... Background Lack of standardization in CT protocol choice contributes to radiation dose variation. Purpose To create a framework to assess radiation doses within broad CT categories defined accordin...

30 August 2023 in European Radiology
Verified
Journal Article
Open access
PDF file
YWYifei Wang
Yifei Wang
PCPhilip Chu
Philip Chu
TSTimothy P. Szczykutowicz
Timothy P. Szczykutowicz
CSCarly Stewart
Carly Stewart
RSRebecca Smith-Bindman
Rebecca Smith-Bindman
Objective: Quantify the relationship between CT acquisition parameters and radiation dose, how often parameters are adjusted in real-world practice, and their degree of contribution to real-world dose distribution. Identify discrepancies between parameters that are impactful in theory and impactful in practice. Methods: This study analyses 1.3 million consecutive adult routine abdomen exams performed between November 2015 and Jan 2021 included in the University of California, San Francisco International CT Dose Registry of 155 institutions. We calculated geometric standard deviation (gSD) for five parameters (kV, mAs, spiral pitch, number of phases, scan length) to assess variation in practice. A Gaussian mixed regression model was performed to predict the radiation dose-length product (DLP) using the parameters. Three conceptualizations of “impact” were computed for each parameter. To reflect the theoretical impact, we predict the increase in DLP per 10% (and 15%) increase in the parameter. To reflect the real-world practical impact, we predict the increase in DLP per gSD increase in the parameter. Results: Among studied examinations, mAs, number of phases, and scan length were frequently manipulated (gSD 1.52–1.70); kV was rarely manipulated (gSD 1.07). Theoretically, kV is the most impactful parameter (29% increase in DLP per 10% increase in kV, versus 5–9% increase for other parameters). In real-world practice, kV is less impactful; for each gSD increase in kV, the DLP increases by 20%, versus 22–69% for other parameters. Conclusion: Despite the potential impact of kV on radiation dose, this parameter is rarely manipulated in common practice and this potential remains untapped. Clinical relevance statement: CT beam energy (kV) modulation has the potential to strongly reduce radiation over-dosage to the patient, theoretically more so than similar degrees of modulation in other CT acquisition parameters. Despite this, beam energy modulation rarely occurs in practice, leaving its potential untapped. Key Points: • The relationship between CT acquisition parameter selection and radiation dose roughly coincided with established theoretical understanding. • CT acquisition parameters differ from each other in frequency and magnitude of manipulation, with beam energy (kV) being rarely manipulated. • Beam energy (kV) has the potential to substantially impact radiation dose, but because it is rarely manipulated, it is the least impactful CT acquisition parameter affecting radiation dose in practice.

11 March 2026 in Journal of Imaging Informatics in Medicine
Verified
Journal Article
Open access
YWYifei Wang
Yifei Wang
ACAlejandro Alejandrez Cisneros
Alejandro Alejandrez Cisneros
CSCarly Stewart
Carly Stewart
MMMatthew Malekhedayat
Matthew Malekhedayat
JLJason Luong
Jason Luong
JMJohn Mongan
John Mongan
Prompt engineering techniques which aid in the use of generative artificial intelligence to address classification tasks have expanded considerably in the last 2 years. The success of such methods varies depending on context, and their efficacy in extracting structured data from unstructured medical text is not well understood. In this paper, five large language prompting strategies were evaluated on a structured categorical question about unstructured radiology reports. Three categories were typically explicit in the text, while one required extrapolation from medical knowledge. The five prompting strategies each contained one or more of the following prompting techniques: external knowledge source, recursive criticism and improvement, chain-of-thought. The efficacy of each strategy was assessed by measuring accuracy and rate of non-determinism. Accuracy was measured by overall correctness and sensitivity to the non-explicit category. Non-determinism was assessed by running each strategy multiple times per exam, tracking both the frequency of exams with non-deterministic outputs and the number of runs needed before stabilization. The presence of an external knowledge source increased sensitivity to the non-explicit category from 10 to 66%, with minimal impact on overall accuracy. Sensitivity further increased to 78% with the introduction of recursive criticism and improvement and chain-of-thought but at the cost of increased non-determinism, with the proportion of exams with non-deterministic results increasing from 7% with only external knowledge source to 14%.

29 October 2018 in Journal of the American Statistical Association
Verified
Journal Article
Closed access
YWYifei Wang
Yifei Wang
DTDaniel J. Tancredi
Daniel J. Tancredi
DMDiana L. Miglioretti
Diana L. Miglioretti
It is a common interest in medicine to determine whether a hospital meets a benchmark created from an aggregate reference population, after accounting for differences in distributions of multiple covariates. Due to the difficulties of collecting individual-level data, however, it is often the case that only marginal distributions of the covariates are available, making covariate-adjusted comparison challenging. We propose and evaluate a novel approach for conducting indirect standardization when only marginal covariate distributions of the studied hospital are known, but complete information is available for the reference hospitals. We do this with the aid of two existing methods: iterative proportional fit, which estimates the cells of a contingency table when only marginal sums are known, and synthetic control methods, which create a counterfactual control group using a weighted combination of potential control groups. The proper application of these existing methods for indirect standardization would require accounting for the statistical uncertainties induced by a situation where no individual-level data is collected from the studied population. We address this need with a novel method which uses a random Dirichlet parametrization of the synthetic control weights to estimate uncertainty intervals for the standard incidence ratio. We demonstrate our novel methods by estimating hospital-level standardized incidence ratios for comparing the adjusted probability of computed tomography examinations with high radiations doses, relative to a reference standard and we evalauate out methods in a simulation study.

2 November 2022 in Research Square
Verified
Preprint
Open access
PDF file
YWYifei Wang
Yifei Wang
PCPhilip Chu
Philip Chu
Indirect standardization, and its associated parameter the Standardized Incidence Ratio, is a commonly-used tool in hospital profiling for comparing the incidence of negative outcomes between an index hospital and a larger population of reference hospitals, while adjusting for confounding factors. For hypothesis testing and statistical inference, the denominator of the Standardized Incidence Ratio, whose computation requires the distribution of confounding covariates in the index hospital, is assumed to be known, and is thus fixed. This assumption severely compromises one's ability to compute required sample sizes for high-powered indirect standardization, as in the contexts where sample size calculation is desired, it's difficult to even estimate such a covariate distribution, much less know it outright. This paper presents novel statistical methodology to perform hypothesis testing and sample size calculation for the Standardized Incidence Ratio while not knowing the covariate distribution of the index hospital, accounting for the mathematical uncertainty this lack of knowledge induces. We apply our methods to example data, as well as to simulation studies, to assess both its capabilities in a vacuum and in comparison to traditional methods of indirect standardization.

5 December 2021 in Statistics in Medicine
Verified
Journal Article
Closed access
PDF file
YWYifei Wang
Yifei Wang
DTDaniel J. Tancredi
Daniel J. Tancredi
DMDiana L. Miglioretti
Diana L. Miglioretti