arXiv subjects

Joseph Cox

Publications and source records attributed to Joseph Cox.

6 recordsLinked to original sources

arXiv2024

BrainSegFounder: Towards 3D Foundation Models for Neuroimage Segmentation

The burgeoning field of brain health research increasingly leverages artificial intelligence (AI) to interpret and analyze neurological data. This study introduces a novel approach towards the creation of medical foundation models by integrating a large-scale multi-modal magnetic resonance imaging (MRI) dataset derived from 41,400 participants in its own. Our method involves a novel two-stage pretraining approach using vision transformers. The first stage is dedicated to encoding anatomical structures in generally healthy brains, identifying key features such as shapes and sizes of different brain regions. The second stage concentrates on spatial information, encompassing aspects like location and the relative positioning of brain structures. We rigorously evaluate our model, BrainFounder, using the Brain Tumor Segmentation (BraTS) challenge and Anatomical Tracings of Lesions After Stroke v2.0 (ATLAS v2.0) datasets. BrainFounder demonstrates a significant performance gain, surpassing the achievements of the previous winning solutions using fully supervised learning. Our findings underscore the impact of scaling up both the complexity of the model and the volume of unlabeled training data derived from generally healthy brains, which enhances the accuracy and predictive capabilities of the model in complex neuroimaging tasks with MRI. The implications of this research provide transformative insights and practical applications in healthcare and make substantial steps towards the creation of foundation models for Medical AI. Our pretrained models and training code can be found at https://github.com/lab-smile/GatorBrain.

eess.IV↗

arXiv2024

Estimating hidden population size from a single respondent-driven sampling survey

This work is concerned with the estimation of hard-to-reach population sizes using a single respondent-driven sampling (RDS) survey, a variant of chain-referral sampling that leverages social relationships to reach members of a hidden population. The popularity of RDS as a standard approach for surveying hidden populations brings theoretical and methodological challenges regarding the estimation of population sizes, mainly for public health purposes. This paper proposes a frequentist, model-based framework for estimating the size of a hidden population using a network-based approach. An optimization algorithm is proposed for obtaining the identification region of the target parameter when model assumptions are violated. We characterize the asymptotic behavior of our proposed methodology and assess its finite sample performance under departures from model assumptions.

stat.ME↗

arXiv2021

Neighbourhood Bootstrap for Respondent-Driven Sampling

Respondent-Driven Sampling (RDS) is a form of link-tracing sampling, a sampling technique used for `hard-to-reach' populations that aims to leverage individuals' social relationships to reach potential participants. While the methodological focus has been restricted to the estimation of population proportions, there is a growing interest in the estimation of uncertainty for RDS as recent findings suggest that most variance estimators underestimate variability. Recently, Baraff et al. (2016) proposed the \textit{tree bootstrap} method based on resampling the RDS recruitment tree, and empirically showed that this method outperforms current bootstrap methods. However, some findings suggest that the tree bootstrap (severely) overestimates uncertainty. In this paper, we propose the \textit{neighbourhood} bootstrap method for quantifiying uncertainty in RDS. We prove the consistency of our method under some conditions and investigate its finite sample performance, through a simulation study, under realistic RDS sampling assumptions.

stat.ME↗

arXiv2020

General Regression Methods for Respondent-Driven Sampling Data

Respondent-Driven Sampling (RDS) is a variant of link-tracing sampling techniques that aim to recruit hard-to-reach populations by leveraging individuals' social relationships. As such, an RDS sample has a graphical component which represents a partially observed network of unknown structure. Moreover, it is common to observe homophily, or the tendency to form connections with individuals who share similar traits. Currently, there is a lack of principled guidance on multivariate modeling strategies for RDS to address homophilic covariates and the dependence between observations within the network. In this work, we propose a methodology for general regression techniques using RDS data. This is used to study the socio-demographic predictors of HIV treatment optimism (about the value of antiretroviral therapy) among gay, bisexual and other men who have sex with men, recruited into an RDS study in Montreal, Canada.

stat.ME↗

arXiv2020

Sampling from Networks: Respondent-Driven Sampling

Respondent-Driven Sampling (RDS) is a variant of link-tracing, a sampling technique for surveying hard-to-reach communities that takes advantage of community members' social networks to reach potential participants. As a network-based sampling method, RDS is faced with the fundamental problem of sampling from population networks where features such as homophily (the tendency for individuals with similar traits to share social ties) and differential activity (the ratio of the average number of connections by attribute) are sensitive to the choice of a sampling method. Though not clearly described in the RDS literature, many simple methods exist to generate simulated RDS data, with specific levels of network features, where the focus is on estimating simple estimands. However, the accuracy of these methods in their abilities to consistently recover those targeted network features remains unclear. This is also motivated by recent findings that some population network parameters (e.g.~homophily) cannot be consistently estimated from the RDS data alone \citep{Crawford17}. In this paper, we conduct a simulation study to assess the accuracy of existing RDS simulation methods, in terms of their abilities to generate RDS samples with the desired levels of two network parameters: homophily and differential activity. The results show that (1) homophily cannot be consistently estimated from simulated RDS samples and (2) differential activity estimates are more precise when groups, defined by traits, are equally active and equally represented in the population. We use this approach to mimic features of the Engage Study, an RDS sample of gay, bisexual and other men who have sex with men in Montreal.

stat.AP↗

arXiv2009

Spatial Localization Problem and the Circle of Apollonius

The Circle of Apollonius is named after the ancient geometrician Apollonius of Perga. This beautiful geometric construct can be helpful when solving some general problems of mathematical physics, optics and electricity. Here we discuss its applications to the 'source localization' problems, e.g. pinpointing a radioactive source using the set of Geiger counters. The Circles of Apollonius help analyze these problems in transparent and intuitive manner. This discussion can be useful for High School Physics and Math curriculum.

physics.ed-ph↗