-
Artificial Intelligence in Fetal Resting-State Functional MRI Brain Segmentation: A Comparative Analysis of 3D UNet, VNet, and HighRes-Net Models
Authors:
Farzan Vahedifard,
Xuchu Liu,
Mehmet Kocak,
H. Asher Ai,
Mark Supanich,
Christopher Sica.,
Kranthi K Marathu,
Seth Adler,
Maysam Orouskhani,
Sharon Byrd
Abstract:
Introduction: Fetal resting-state functional magnetic resonance imaging (rs-fMRI) is a rapidly evolving field that provides valuable insight into brain development before birth. Accurate segmentation of the fetal brain from the surrounding tissue in nonstationary 3D brain volumes poses a significant challenge in this domain. Current available tools have 0.15 accuracy. Aim: This study introduced a…
▽ More
Introduction: Fetal resting-state functional magnetic resonance imaging (rs-fMRI) is a rapidly evolving field that provides valuable insight into brain development before birth. Accurate segmentation of the fetal brain from the surrounding tissue in nonstationary 3D brain volumes poses a significant challenge in this domain. Current available tools have 0.15 accuracy. Aim: This study introduced a novel application of artificial intelligence (AI) for automated brain segmentation in fetal brain fMRI, magnetic resonance imaging (fMRI). Open datasets were employed to train AI models, assess their performance, and analyze their capabilities and limitations in addressing the specific challenges associated with fetal brain fMRI segmentation. Method: We utilized an open-source fetal functional MRI (fMRI) dataset consisting of 160 cases (reference: fetal-fMRI - OpenNeuro). An AI model for fMRI segmentation was developed using a 5-fold cross-validation methodology. Three AI models were employed: 3D UNet, VNet, and HighResNet. Optuna, an automated hyperparameter-tuning tool, was used to optimize these models. Results and Discussion: The Dice scores of the three AI models (VNet, UNet, and HighRes-net) were compared, including a comparison between manually tuned and automatically tuned models using Optuna. Our findings shed light on the performance of different AI models for fetal resting-state fMRI brain segmentation. Although the VNet model showed promise in this application, further investigation is required to fully explore the potential and limitations of each model, including the HighRes-net model. This study serves as a foundation for further extensive research into the applications of AI in fetal brain fMRI segmentation.
△ Less
Submitted 17 November, 2023;
originally announced November 2023.
-
Sylvester-Gallai type theorems for approximate collinearity
Authors:
Albert Ai,
Zeev Dvir,
Shubhangi Saraf,
Avi Wigderson
Abstract:
We study questions in incidence geometry where the precise position of points is `blurry' (e.g. due to noise, inaccuracy or error). Thus lines are replaced by narrow tubes, and more generally affine subspaces are replaced by their small neighborhood. We show that the presence of a sufficiently large number of approximately collinear triples in a set of points in d dimensional complex space implies…
▽ More
We study questions in incidence geometry where the precise position of points is `blurry' (e.g. due to noise, inaccuracy or error). Thus lines are replaced by narrow tubes, and more generally affine subspaces are replaced by their small neighborhood. We show that the presence of a sufficiently large number of approximately collinear triples in a set of points in d dimensional complex space implies that the points are close to a low dimensional affine subspace. This can be viewed as a stable variant of the Sylvester-Gallai theorem and its extensions.
Building on the recently found connection between Sylvester-Gallai type theorems and complex Locally Correctable Codes (LCCs), we define the new notion of stable LCCs, in which the (local) correction procedure can also handle small perturbations in the euclidean metric. We prove that such stable codes with constant query complexity do not exist. No impossibility results were known in any such local setting for more than 2 queries.
△ Less
Submitted 1 November, 2012;
originally announced November 2012.
-
One-bit compressed sensing with non-Gaussian measurements
Authors:
Albert Ai,
Alex Lapanowski,
Yaniv Plan,
Roman Vershynin
Abstract:
In one-bit compressed sensing, previous results state that sparse signals may be robustly recovered when the measurements are taken using Gaussian random vectors. In contrast to standard compressed sensing, these results are not extendable to natural non-Gaussian distributions without further assumptions, as can be demonstrated by simple counter-examples. We show that approximately sparse signals…
▽ More
In one-bit compressed sensing, previous results state that sparse signals may be robustly recovered when the measurements are taken using Gaussian random vectors. In contrast to standard compressed sensing, these results are not extendable to natural non-Gaussian distributions without further assumptions, as can be demonstrated by simple counter-examples. We show that approximately sparse signals that are not extremely sparse can be accurately reconstructed from single-bit measurements sampled according to a sub-gaussian distribution, and the reconstruction comes as the solution to a convex program.
△ Less
Submitted 8 April, 2013; v1 submitted 30 August, 2012;
originally announced August 2012.
