A5071349804- Cancer Immunotherapy and Biomarkers
- Cutaneous Melanoma Detection and Management
- Blind Source Separation Techniques
- Melanoma and MAPK Pathways
- Colorectal Cancer Treatments and Studies
- CAR-T cell therapy research
- EEG and Brain-Computer Interfaces
- Sparse and Compressive Sensing Techniques
- Functional Brain Connectivity Studies
- Immunotherapy and Immune Responses
- Dermatology and Skin Diseases
- Advanced Statistical Methods and Models
- Cell Image Analysis Techniques
- Neural dynamics and brain function
- Immune Cell Function and Interaction
- Autoimmune Bullous Skin Diseases
- AI in cancer detection
- Fault Detection and Control Systems
- Single-cell and spatial transcriptomics
- Nonmelanoma Skin Cancer Studies
- Spectroscopy and Chemometric Analyses
- Graph Theory and Algorithms
- Complex Network Analysis Techniques
- Advanced Clustering Algorithms Research
- Urticaria and Related Conditions
Massachusetts General Hospital
2021-2025
Harvard University
2021-2025
Center for Systems Biology
2022-2024
China Pharmaceutical University
2024
Beijing Language and Culture University
2024
Harvard University Press
2022-2024
Mass General Brigham
2022
The University of Texas at Dallas
2019-2021
Electrophysiological source imaging (ESI) refers to the process of reconstructing underlying activated sources on cortex given brain signal measured by Electroencephalography (EEG) or Magnetoencephalography (MEG). Due ill-posed nature ESI, solving ESI requires design neurophysiologically plausible regularization priors guarantee a unique solution. Recovering focally extended is more challenging, and traditionally uses total variation promote spatial continuity sources. In this paper, we...
Monitoring the depth of unconsciousness during anesthesia is beneficial in both clinical settings and neuroscience investigations to understand brain mechanisms. Electroencephalogram (EEG) has been used as an objective means characterizing altered arousal and/or cognition states induced by anesthetics real-time. Different general affect cerebral electrical activities different ways. However, performance conventional machine learning models on EEG data unsatisfactory due low Signal Noise...
Prognostic analysis for early-stage (stage I/II) melanomas is of paramount importance customized surveillance and treatment plans. Since immune checkpoint inhibitors have recently been approved stage IIB IIC melanomas, prognostic tools to identify patients at high risk recurrence become even more critical. This study aims assess the effectiveness machine-learning algorithms in predicting melanoma using clinical histopathologic features from Electronic Health Records (EHRs). We collected 1720...
Abstract Background Cutaneous immune-related adverse events (cirAEs) are the most common toxicities to occur in setting of immune checkpoint inhibitor (ICI) therapy. Identifying patients who at increased risk developing cirAEs may improve quality life and outcomes. Objectives To investigate influence cancer type histology on development ICI therapy survival Methods This retrospective cohort study included recruited between 1 December 2011 30 October 2020. They received from 2020 with...
Dupilumab has been added to National Cancer Comprehensive Network (NCCN) guidelines as a therapeutic strategy for managing certain cutaneous immune-related adverse events (cirAEs) from immune checkpoint inhibitor (ICI) therapy. However, little is known about the implications of dupilumab cancer outcomes in this population. In multi-institutional study, we evaluate impact treatment on survival among ICI recipients. We conducted muti-institutional retrospective cohort study recipients Mass...
Computational pathology foundation models (CPathFMs) have emerged as a powerful approach for analyzing histopathological data, leveraging self-supervised learning to extract robust feature representations from unlabeled whole-slide images. These models, categorized into uni-modal and multi-modal frameworks, demonstrated promise in automating complex tasks such segmentation, classification, biomarker discovery. However, the development of CPathFMs presents significant challenges, limited data...
Given a data matrix, unsupervised column subset selection refers to the problem of identifying columns that can be used linearly approximate original matrix. This has many applications, such as feature and representative selection, but solving it optimally is known NP-hard. We consider multi-view which extends concept (single-view) represented in multiple views or modalities. introduce combinatorial search algorithm for this generalized problem. One variant guaranteed compute an optimal...
Cutaneous immune-related adverse events (cirAEs) affect nearly one-third of patients treated with immune checkpoint inhibitors (ICIs). Identifying at high risk for cirAE can help guide monitoring and therapeutic management. Whether tumor somatic mutations are associated the development remains unknown. We analyzed single nucleotide variants (SNVs) from a discovery cohort 733 stage III/IV melanoma receiving ICIs Dana Farber Cancer Institute Massachusetts General Hospital. CirAE was determined...
Abstract Background: Several studies have outlined whole genome-doubled (WGD) tumor tissue characterized by elevated genome ploidy as an important biomarker positively associated with immunotherapy response and overall survival of cutaneous melanoma patients (Liu NatMed2019, Tarantino BioRXiv2022), well conferring therapeutic vulnerabilities (Quinton Nat2021). A robust cost-effective method for detecting this would significantly improve patient risk assessment treatment optimization....
Abstract Cutaneous melanoma develops from precursor cells that progress to in situ (MIS) and eventually invasive disease. In metastatic tumors, switch a melanocytic an undifferentiated, neural crest (NC)-like state, which is associated with metastasis therapeutic resistance. However, it unclear whether these cell states or gene expression changes occur at the premetastatic stage if they are features of local tumor microenvironment (TME). We analyzed cohort 62 patients primary cutaneous...
Although indications for immune checkpoint inhibitors (ICIs) have dramatically increased in the past decade, ICIs been associated with autoinflammatory immune-related adverse events, which can resemble autoimmune diseases (ADs). Little is known about impact of baseline AD on mortality cancer patients treated ICIs. Here, we identified 17 497 preexisting diagnoses prior to treatment antiprogrammed cell death receptor-1 or ligand-1 therapy and matched controls through TriNetX Diamond network...
The following are two classical approaches to dimensionality reduction: 1. Approximating the data with a small number of features that exist in (feature selection). 2. arbitrary extraction). We study generalization approximates both selected and extracted features. show an optimal solution this hybrid problem involves combinatorial search, cannot be trivially obtained even if one can solve optimally separate problems selection extraction. Our approach gives approximate solutions uses “best...
Spectral clustering is a powerful technique. It leverages the spectral properties of graphs to partition data points into meaningful clusters. The most common criterion for evaluating multi-way NCut. Column Subset Selection an important optimization technique in domain feature selection and dimension reduction which aims identify subset columns given matrix that can be used approximate entire matrix. We show column compute use this obtain new graph algorithms.
The common criteria for evaluating spectral clustering are NCut and RatioCut. seemingly unrelated column subset selection (CSS) problem aims to compute a that linearly approximates the entire matrix. A criterion is approximation error in Frobenius norm (ApproxErr). We show any algorithm CSS can be viewed as minimizes by applying it matrix formed from graph edges. Conversely, seen identifying In both cases, ApproxErr have same value. Analogous results hold RatioCut with slightly different...
Abstract Cancer is a complex cellular ecosystem where malignant cells coexist and interact with immune, stromal other within the tumor microenvironment (TME). Recent technological advancements in spatially resolved multiplexed imaging at single-cell resolution have led to generation of large-scale high-dimensional datasets from biological specimens. This underscores necessity for automated methodologies that can effectively characterize molecular, spatial properties TMEs various...