I am an Assistant Professor in the Department of Biomedical Engineering and the Institute for Computational Medicine at Johns Hopkins University. I sit at the Center for Imaging Science in Clark Hall at the Homewood campus. I also have an appointment in the Institute for Data Intensive Engineering and Sciences. My work focuses largely on big and wide data, especially neuroscience, focusing on statistics of brain graphs (connectomes). I co-founded the Open Connectome Project (OCP) with my brother R. Jacob Vogelstein and Randal Burns, Associated Professor in the Department of Computer Science at Johns Hopkins University. Recently it has spawned NeuroData which is the mother of several complementary projects, including OCP, Open Synaptome Project, and more. We run a very vertical group, with people working at all levels of analysis, ranging from data collection to analysis and interpretation. We are always looking for new collaborators and team members. Please inquire if you are interested.
We seek patterns in our physical worlds (e.g., our bodies, our brains) as well as our mental worlds (e.g., our perceptions, experiences, memories, thoughts, emotions, psychiatric conditions). More importantly, we seek to understand patterns in our mental worlds in terms of our physical worlds. Our hope and belief is that via developing a deeper understanding of the links between these worlds, we will be able to bring them into greater alignment. A primary motivating factor is that all humans/animals have brains and therefore, such ideas could directly benefit all of animalkind. Thus, all of our research products are freely available to all. For more details see my cv.
The fundamental driving force of science is the discovery of latent structure that converts myriad disparate data into understanding. In the 21st century, the growth of data acquisition has greatly outpaced the growth of data analysis, rendering current computational and statistical tools insufficient for extracting meaning from large datasets. Neuroscience is particularly susceptible to these big data challenges, as neuroexperimentalists have devised methods of collecting terabytes of data per hour. Without statistical and computational frameworks for analysis and organization of these data, the field will be unable to fully reap the benefits of this enormous potential. Our expertise in (i) computer science, (ii) data curation, (iii) statistical science, and (iv) neuroscience enables us to fill this gap. Below we elaborate on these four threads.
Statistical sciences have developed a beautiful corpus of knowledge on inference in various settings over the last 100 years. However, in recent years, new experimental and data collection methods have yielded datasets that no longer fit within the assumptions upon which that body of literature rests. We therefore focus on building statistical methods for 21st century data, including: time-series, graphs, shapes, vectors, and arbitrary metric-valued data.
Statistical methods require computational implementations. Moreover, the raw data of the 21st century is not typically in a form immediately amenable to statistical analysis; rather, various data wrangling/munging/pre-processing is typically beneficial. We therefore develop computational tools to enable principled statistical inference on such data. This includes: petascale data management, interactive visualizations, streaming image processing, distributed computer vision, and scalable machine learning libraries. We are adherents to the philosophies of open science, our code is always open source. Thus, my personal github account contains many repos including code from publications prior to launching NeuroData and FlashX (links contain all the relevant code repos).
To answer any of the below questions, using any of the above statistical methods and computational tools, fundamentally requires data. Much like all of our code is open source, all of our data is open access. Moreover, we provide it in a fashion immediately amenable to analysis for everyone, including electron microscopy, array tomography, expansion microscopy, x-ray microscopy, CLARITY, histology, as well as physiological images such as calcium imaging, massively parallel electrophysiology, and multi-modal mri, with the corresponding data derivatives including annotations, graphs with rich attributes, and more.
The above described tools, both statistical and computational, are designed and built in the service of answering fundamental questions in neuroscience, both basic and clinical. The projects we are have worked on, and continue to develop include anatomy, physiology, systems, and connectome coding. Connectome coding is homologous to neural coding, where we learn the rules by which memories & experiences are stored in patterns of connectivity, rather than patterns of activity.
jovo [at symbol] jhu [dot] edu
Joshua T. Vogelstein
Center for Imaging Science
Clark Hall, Rm 317C
3400 N. Charles St
Baltimore, MD 21218
NeuroData is always hiring exceptional individuals at all levels! In fact, even if we are not an optimal choice for your future development as a scientist, we endeavor to assist/support your growth in various ways, including short/extended visits to work with us. Unfortunately, because our time and funding are limited, we cannot work with everyone that we would love to; nonetheless, please don't hesitate to contact us in case we can provide some guidance :) We are currently particularly looking for people with their own ideas that they are super passionate about, that we might be able to contribute to, as well any of the above listed projects.
Please email your cv (and transcript if appropriate) to me. See below for more details: