Kristin A. Linn

Kristin A. Linn

Auto-generated profile, based on publicly available information

Kristin A. Linn

Penn Statistics in Imaging and Visualization Center, Department of Biostatistics, Epidemiology, and Informatics, University of Pennsylvania, Philadelphia, PA, USA; Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

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

Representative Publications

View All Publications
19 June 2020 in Nature Neuroscience
Verified
Journal Article
Open access
PDF file
JDJordan D. Dworkin
Jordan D. Dworkin
KLKristin A. Linn
Kristin A. Linn
ETErin G. Teich
Erin G. Teich
PZPerry Zurn
Perry Zurn
RSRussell T. Shinohara
Russell T. Shinohara
DBDani S Bassett
Dani S Bassett
Similarly to many scientific disciplines, neuroscience has increasingly attempted to confront pervasive gender imbalances. Although publishing and conference participation are often highlighted, recent research has called attention to the prevalence of gender imbalance in citations. Because of the downstream effects of citations on visibility and career advancement, understanding the role of gender in citation practices is vital for addressing scientific inequity. Here, we investigate whether gendered patterns are present in neuroscience citations. Using data from five top neuroscience journals, we find that reference lists tend to include more papers with men as first and last author than would be expected if gender were unrelated to referencing. Importantly, we show that this imbalance is driven largely by the citation practices of men and is increasing over time as the field diversifies. We assess and discuss possible mechanisms and consider how researchers might approach these issues in their own work.

1 October 2020 in NeuroImage
Verified
Journal Article
Open access
JBJoanne C. Beer
Joanne C. Beer
NTNicholas Tustison
Nicholas Tustison
PCPhilip A. Cook
Philip A. Cook
CDChristos Davatzikos
Christos Davatzikos
YSYvette I. Sheline
Yvette I. Sheline
KLKristin A. Linn
Kristin A. Linn
While aggregation of neuroimaging datasets from multiple sites and scanners can yield increased statistical power, it also presents challenges due to systematic scanner effects. This unwanted technical variability can introduce noise and bias into estimation of biological variability of interest. We propose a method for harmonizing longitudinal multi-scanner imaging data based on ComBat, a method originally developed for genomics and later adapted to cross-sectional neuroimaging data. Using longitudinal cortical thickness measurements from 663 participants in the Alzheimer's Disease Neuroimaging Initiative (ADNI) study, we demonstrate the presence of additive and multiplicative scanner effects in various brain regions. We compare estimates of the association between diagnosis and change in cortical thickness over time using three versions of the ADNI data: unharmonized data, data harmonized using cross-sectional ComBat, and data harmonized using longitudinal ComBat. In simulation studies, we show that longitudinal ComBat is more powerful for detecting longitudinal change than cross-sectional ComBat and controls the type I error rate better than unharmonized data with scanner included as a covariate. The proposed method would be useful for other types of longitudinal data requiring harmonization, such as genomic data, or neuroimaging studies of neurodevelopment, psychiatric disorders, or other neurological diseases.

22 June 2017 in American Journal of Neuroradiology
Verified
Journal Article
Open access
PDF file
RSR.T. Shinohara
R.T. Shinohara
JOJ. Oh
J. Oh
GNGovind Nair
Govind Nair
PCPeter A Calabresi
Peter A Calabresi
CDChristos Davatzikos
Christos Davatzikos
TCthe NAIMS Cooperative
the NAIMS Cooperative
BACKGROUND AND PURPOSE: MR imaging can be used to measure structural changes in the brains of individuals with multiple sclerosis and is essential for diagnosis, longitudinal monitoring, and therapy evaluation. The North American Imaging in Multiple Sclerosis Cooperative steering committee developed a uniform high-resolution 3T MR imaging protocol relevant to the quantification of cerebral lesions and atrophy and implemented it at 7 sites across the United States. To assess intersite variability in scan data, we imaged a volunteer with relapsing-remitting MS with a scan-rescan at each site. MATERIALS AND METHODS: All imaging was acquired on Siemens scanners (4 Skyra, 2 Tim Trio, and 1 Verio). Expert segmentations were manually obtained for T1-hypointense and T2 (FLAIR) hyperintense lesions. Several automated lesion-detection and whole-brain, cortical, and deep gray matter volumetric pipelines were applied. Statistical analyses were conducted to assess variability across sites, as well as systematic biases in the volumetric measurements that were site-related. RESULTS: Systematic biases due to site differences in expert-traced lesion measurements were significant (P < .01 for both T1 and T2 lesion volumes), with site explaining >90% of the variation (range, 13.0–16.4 mL in T1 and 15.9–20.1 mL in T2) in lesion volumes. Site also explained >80% of the variation in most automated volumetric measurements. Output measures clustered according to scanner models, with similar results from the Skyra versus the other 2 units. CONCLUSIONS: Even in multicenter studies with consistent scanner field strength and manufacturer after protocol harmonization, systematic differences can lead to severe biases in volumetric analyses.

1 November 2023 in JAMA Network Open
Verified
Journal Article
Open access
PDF file
CYChenxi Yuan
Chenxi Yuan
KLKristin A. Linn
Kristin A. Linn
RHRebecca A. Hubbard
Rebecca A. Hubbard
The development and use of machine learning (ML) algorithms in health care has received a surge of attention in recent years.1-12 Although ML algorithms can inform clinical decision-making and are potentially associated with improved population health,13 there is growing concern that ML may inadvertently introduce bias into decision-making processes, which may be associated with unintended discrimination against underrepresented and disadvantaged populations.14-16 Given that algorithms are vulnerable to biases that render their decisions unfair, fairness, in the context of decision making, is the absence of any prejudice or favoritism toward an individual or group based on that groups’ inherent or acquired characteristics. An unfair algorithm, also referred to as algorithmic bias, skews benefits toward a particular group of people with respect to protected attributes.17 Protected attributes are features that may not be used as the basis for decisions. There is no one universal set of protected attributes. They are determined based on laws, regulations, or other policies governing a particular application domain in a particular jurisdiction. Attributes such as race and ethnicity, color, age, gender and sex, national origin, religion, and marital status are commonly considered protected attributes.17-19

9 December 2022 in Epidemiology
Verified
Journal Article
Closed access
KGKylie Getz
Kylie Getz
RHRebecca A. Hubbard
Rebecca A. Hubbard
KLKristin A. Linn
Kristin A. Linn
Background: Missing data are common in studies using electronic health records (EHRs)-derived data. Missingness in EHR data is related to healthcare utilization patterns, resulting in complex and potentially missing not at random missingness mechanisms. Prior research has suggested that machine learning-based multiple imputation methods may outperform traditional methods and may perform well even in settings of missing not at random missingness. Methods: We used plasmode simulations based on a nationwide EHR-derived de-identified database for patients with metastatic urothelial carcinoma to compare the performance of multiple imputation using chained equations, random forests, and denoising autoencoders in terms of bias and precision of hazard ratio estimates under varying proportions of observations with missing values and missingness mechanisms (missing completely at random, missing at random, and missing not at random). Results: Multiple imputation by chained equations and random forest methods had low bias and similar standard errors for parameter estimates under missingness completely at random. Under missingness at random, denoising autoencoders had higher bias than multiple imputation by chained equations and random forests. Contrary to results of prior studies of denoising autoencoders, all methods exhibited substantial bias under missingness not at random, with bias increasing in direct proportion to the amount of missing data. Conclusions: We found no advantage of denoising autoencoders for multiple imputation in the setting of an epidemiologic study conducted using EHR data. Results suggested that denoising autoencoders may overfit the data leading to poor confounder control. Use of more flexible imputation approaches does not mitigate bias induced by missingness not at random and can produce estimates with spurious precision.

22 February 2018 in American Journal of Neuroradiology
Verified
Journal Article
Open access
PDF file
JDJ.D. Dworkin
J.D. Dworkin
KLK.A. Linn
K.A. Linn
IOI. Oguz
I. Oguz
GFG.M. Fleishman
G.M. Fleishman
RBR. Bakshi
R. Bakshi
RSR.T. Shinohara
R.T. Shinohara
BACKGROUND AND PURPOSE: Lesion load is a common biomarker in multiple sclerosis, yet it has historically shown modest association with clinical outcome. Lesion count, which encapsulates the natural history of lesion formation and is thought to provide complementary information, is difficult to assess in patients with confluent (ie, spatially overlapping) lesions. We introduce a statistical technique for cross-sectionally counting pathologically distinct lesions. MATERIALS AND METHODS: MR imaging was used to assess the probability of a lesion at each location. The texture of this map was quantified using a novel technique, and clusters resembling the center of a lesion were counted. Validity compared with a criterion standard count was demonstrated in 60 subjects observed longitudinally, and reliability was determined using 14 scans of a clinically stable subject acquired at 7 sites. RESULTS: The proposed count and the criterion standard count were highly correlated (r = 0.97, P < .001) and not significantly different (t59 = −.83, P = .41), and the variability of the proposed count across repeat scans was equivalent to that of lesion load. After accounting for lesion load and age, lesion count was negatively associated (t58 = −2.73, P < .01) with the Expanded Disability Status Scale. Average lesion size had a higher association with the Expanded Disability Status Scale (r = 0.35, P < .01) than lesion load (r = 0.10, P = .44) or lesion count (r = −.12, P = .36) alone. CONCLUSIONS: This study introduces a novel technique for counting pathologically distinct lesions using cross-sectional data and demonstrates its ability to recover obscured longitudinal information. The proposed count allows more accurate estimation of lesion size, which correlated more closely with disability scores than either lesion load or lesion count alone.