Didong Li

Mr. Didong Li

orcid

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
1 January 2024
Verified
Conference Paper
Closed access
JCJiawen Chen
Jiawen Chen
DLDidong Li
Didong Li
YLYun Li
Yun Li
WWWenrong Wu
Wenrong Wu
JZJinwei Zhang
Jinwei Zhang
MZMuqing Zhou
Muqing Zhou
Recent advances in multi-modal algorithms have driven and been driven by the increasing availability of large image-text datasets, leading to significant strides in various fields, including computational pathology. However, in most existing medical image-text datasets, the text typically provides high-level summaries that may not sufficiently describe sub-tile regions within a large pathology image. For example, an image might cover an extensive tissue area containing cancerous and healthy regions, but the accompanying text might only specify that this image is a cancer slide, lacking the nuanced details needed for in-depth analysis. In this study, we introduce STimage-1K4M, a novel dataset designed to bridge this gap by providing genomic features for sub-tile images. STimage-1K4M contains 1,149 images derived from spatial transcriptomics data, which captures gene expression information at the level of individual spatial spots within a pathology image. Specifically, each image in the dataset is broken down into smaller sub-image tiles, with each tile paired with 15,000 − 30,000 dimensional gene expressions. With 4,293,195 pairs of sub-tile images and gene expressions, STimage-1K4M offers unprecedented granularity, paving the way for a wide range of advanced research in multi-modal data analysis an innovative applications in computational pathology, and beyond. 1

2 December 2024 in PLOS Medicine
Verified
Journal Article
Open access
PDF file
MDMatthew R Dunn
Matthew R Dunn
DLDidong Li
Didong Li
MEMarc A Emerson
Marc A Emerson
CTCaroline A Thompson
Caroline A Thompson
HNHazel B Nichols
Hazel B Nichols
MTMelissa A Troester
Melissa A Troester
Background: Delays in breast cancer diagnosis and treatment lead to worse survival and quality of life. Racial disparities in care timeliness have been reported, but few studies have examined access at multiple points along the care continuum (diagnosis, treatment initiation, treatment duration, and genomic testing). Methods and findings: The Carolina Breast Cancer Study (CBCS) Phase 3 is a population-based, case-only cohort (n = 2,998, 50% black) of patients with invasive breast cancer diagnoses (2008 to 2013). We used latent class analysis (LCA) to group participants based on patterns of factors within 3 separate domains: socioeconomic status (“SES”), “care barriers,” and “care use.” These classes were evaluated in association with delayed diagnosis (approximated with stages III–IV at diagnosis), delayed treatment initiation (more than 30 days between diagnosis and first treatment), prolonged treatment duration (time between first and last treatment–by treatment modality), and receipt of OncotypeDx genomic testing (evaluated among patients with early stage, ER+ (estrogen receptor-positive), HER2- (human epidermal growth factor receptor 2-negative) disease). Associations were evaluated using adjusted linear-risk regression to estimate relative frequency differences (RFDs) with 95% confidence intervals (CIs). Delayed diagnosis models were adjusted for age; delayed and prolonged treatment models were adjusted for age and tumor size, stage, and grade at diagnosis; and OncotypeDx models were adjusted for age and tumor size and grade. Overall, 18% of CBCS participants had late stage/delayed diagnosis, 35% had delayed treatment initiation, 48% had prolonged treatment duration, and 62% were not OncotypeDx tested. Black women had higher prevalence for each outcome. We identified 3 latent classes for SES (“high SES,” “moderate SES,” and “low SES”), 2 classes for care barriers (“few barriers,” “more barriers”), and 5 classes for care use (“short travel/high preventive care,” “short travel/low preventive care,” “medium travel,” “variable travel,” and “long travel”) in which travel is defined by estimated road driving time. Low SES and more barriers to care were associated with greater frequency of delayed diagnosis (RFDadj = 5.5%, 95% CI [2.4, 8.5]; RFDadj = 6.7%, 95% CI [2.8,10.7], respectively) and prolonged treatment (RFDadj = 9.7%, 95% CI [4.8 to 14.6]; RFDadj = 7.3%, 95% CI [2.4 to 12.2], respectively). Variable travel (short travel to diagnosis but long travel to surgery) was associated with delayed treatment in the entire study population (RFDadj = 10.7%, 95% CI [2.7 to 18.8]) compared to the short travel, high use referent group. Long travel to both diagnosis and surgery was associated with delayed treatment only among black women. The main limitations of this work were inability to make inferences about causal effects of individual variables that formed the latent classes, reliance on self-reported socioeconomic and healthcare history information, and generalizability outside of North Carolina, United States of America. Conclusions: Black patients face more frequent delays throughout the care continuum, likely stemming from different types of access barriers at key junctures. Improving breast cancer care access will require intervention on multiple aspects of SES and healthcare access.

1 September 2024 in The Annals of Applied Statistics
Verified
Journal Article
Closed access
PDF file
DLDidong Li
Didong Li
AJAndrew Jones
Andrew Jones
BEBarbara Engelhardt
Barbara Engelhardt
Case-control experiments are essential to the scientific method, as they allow researchers to test biological hypotheses by looking for differences in outcome between cases and controls. It is then of interest to characterize variation that is enriched in a “foreground” (case) dataset relative to a “background” (control) dataset. For example, in a genomics context, the goal is to identify low-dimensional transcriptional structure unique to patients with certain disease (cases) vs. those without that disease (controls). In this work we propose probabilistic contrastive principal component analysis (PCPCA), a probabilistic dimension reduction method designed for case-control data. We describe inference in PCPCA through a contrastive likelihood and show that our model generalizes PCA, probabilistic PCA, and contrastive PCA. We discuss how to set the tuning parameter in theory and in practice, and we show several of PCPCA’s advantages in the analysis of case-control data over related methods, including greater interpretability, uncertainty quantification and principled inference, robustness to noise and missing data, and the ability to generate “foreground-enriched” data from the model. We demonstrate PCPCA’s performance on case-control data through a series of simulations, and we successfully identify variation specific to case data in genomic case-control experiments with data modalities, including gene expression, protein expression, and images.

15 October 2025 in PLOS Medicine
Verified
Journal Article
Open access
PDF file
AWAndrew E. Walker
Andrew E. Walker
XGXiaohua Gao
Xiaohua Gao
QWQichen Wang
Qichen Wang
GCGabriela De la Cruz
Gabriela De la Cruz
DLDidong Li
Didong Li
MTMelissa A. Troester
Melissa A. Troester
Background: Tumor infiltrating lymphocytes (TILs) are prognostic in triple-negative breast cancer, but not estrogen receptor (ER) positive cancers which comprise 70%–80% of breast cancers. This is due to the relatively low immune infiltration in ER-positive tumors. However, few studies have explored the prognostic impact of lower abundance TILs evaluated using spatial methods. The objective of this study was to explore whether the distribution of lymphocytes with respect to tumor cells predicts prognosis. Methods and findings: In this retrospective cohort study, we used multiplex immunofluorescent (IF)-stained images of tissue microarray cores (stained for cytokeratin [Ck], CD8, and FoxP3) obtained from 1,467 study participants to compute distance-based visual morphometry for epithelial and immune cells, including two new metrics, proximity and consistency . Proximity and consistency are defined as functions of the mean and variance of nearest neighbor distances between Ck+ tumor cells and CD8+ T cells. Prognostic significance of proximity and consistency were compared to lymphocyte counts using log-rank tests of differences in Kaplan–Meier survival curves and Cox proportional hazards models. Better recurrence-free survival (RFS) was observed for both ER+ and ER− breast cancers with high proximity and consistency of CD8+ T cells. Among ER− breast cancers, proximity had the highest RFS hazard ratio (HR 1.84, 95% CI [1.18, 2.87]; p = 0.0069) compared to count and consistency. Among ER-positive participants, RFS hazard ratios for proximity and consistency were 2.04 (95% CI [1.39, 2.98]; p = 0.0003) and 1.82 (95% CI [1.23, 2.69]; p = 0.0026), respectively. These associations were stronger than those observed for lymphocyte count (HR 1.35, 95% CI [0.92, 1.98]; p = 0.1289). Independent prognostic value was demonstrated by controlling clinical and demographic variables such as age, tumor grade, stage, ER status, progesterone receptor status, and human epidermal growth factor receptor 2 status. These IF-derived spatial metrics were also associated with established TILs metrics and RNA expression-based measures of tumor adaptive immune response. Though these results are promising, our exploration of the tumor immune microenvironment was limited by the small number of immune markers available for our data. Conclusions: Spatial characteristics described by proximity and consistency are frequently associated with recurrence irrespective of ER status. The prognostic significance of proximity in ER+ breast cancer implies that spatial parameters may identify individuals who would benefit from immune therapy; up to 75% of breast cancers experience T cell proximity suggestive of immune susceptibility.

11 April 2025 in Proceedings of the AAAI Conference on Artificial Intelligence
Verified
Journal Article
Open access
PDF file
EZEric Zhang
Eric Zhang
DLDidong Li
Didong Li
As functional data assumes a central role in contemporary data analysis, the search for meaningful dimension reduction becomes critical due to its inherent infinite-dimensional structure. Traditional methods, such as Functional Principal Component Analysis (FPCA), adeptly explore the overarching structures within the functional data. However, these methods may not sufficiently identify low-dimensional representations that are specific or enriched in a foreground dataset (case or treatment group) relative to a background dataset (control group). This limitation becomes critical in scenarios where the foreground dataset, such as a specific treatment group in biomedical applications, contains unique patterns or trends that are not as pronounced in the background dataset. Addressing this gap, we propose Contrastive Functional Principal Component Analysis (CFPCA), a method designed to spotlight low-dimensional structures unique to or enriched in the foreground dataset relative to the background counterpart. We supplement our method with theoretical guarantees on CFPCA estimates supported by multiple simulations. Through a series of applications, CFPCA successfully identifies these foreground-specific structures, thereby revealing distinct patterns and trends that traditional FPCA overlooks.

14 July 2020 in Biometrika
Verified
Journal Article
Open access
DLDidong Li
Didong Li
DDDavid B Dunson
David B Dunson
Summary: Classifiers label data as belonging to one of a set of groups based on input features. It is challenging to achieve accurate classification when the feature distributions in the different classes are complex, with nonlinear, overlapping and intersecting supports. This is particularly true when training data are limited. To address this problem, we propose a new type of classifier based on obtaining a local approximation to the support of the data within each class in a neighbourhood of the feature to be classified, and assigning the feature to the class having the closest support. This general algorithm is referred to as local manifold approximation classification. As a simple and theoretically supported special case, which is shown to have excellent performance across a broad variety of examples, we use spheres for local approximation, obtaining a spherical approximation classifier.