Authors:

Xinwan Hu

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Functional magnetic resonance imaging (fMRI) is a non-invasive technique that measures brain activity by detecting changes in blood flow. Time series of fMRI signals is widely used to estimate functional connectomes (FCs), which are commonly represented as correlation matrices. One characteristic of FCs is "fingerprinting" that can identify individuals based on a population of FCs, often measured by identification rates (ID rates). It has been showed that tangent space projections of FCs achieved higher ID rates, while only the initial scan volumes of the time series were evaluated. This study investigates the robustness and reliability of high ID rates when varying the starting points and scan lengths using different distance metrics (euclidean distance and correlation distance). We segmented the time series data into chunks of different sizes, both overlapping and non-overlapping. Using unrelated subjects, we compared ID rates for matching chunks (same starting time and length) and non-matching chunks (same length but different starting times).Our results showed consistently high ID rates using correlation distance for tangent space projections, whereas significant variability was observed with Euclidean distance. These findings suggest that while tangent space projections with correlation distance provide robust identification across varying intervals, the choice of distance metric is critical. The high variability with Euclidean distance implies a need for careful selection of analytical methods in fMRI studies. This study underscores the importance of using appropriate distance metrics and considering different scan intervals to enhance the reliability and applicability of individual identification techniques in connectome research, thus advancing the precision and robustness of neuroimaging analyzes. Keywords: fMRI Scan Length; Functional Connectome; Fingerprinting; Tangent Space Projection

Source:

Purdue University / 2024

Topics:

No topics listed

Co-authors:

Xinwan Hu