Agenda

  • Coordinating discussion on the analysis of Region of Interest (ROI) quality in mesoscope imaging data.

Meeting Recording

Meeting Notes

Mesoscope Data Validation and Quality Control Metrics: Jerome led a discussion with Ali, Farzaneh, Nicholas, Shiella, and others on validating mesoscope data for the upcoming data release paper, focusing on quality control metrics, segmentation methods, and the need for robust analysis across sessions and planes.

Quality Control Metrics Discussion: Jerome initiated a focused discussion on mesoscope data validation, aiming to compile a list of important metrics for manuscript inclusion. The group considered metrics such as ROI quality, trace stability, and cross-session alignment, with input from Ali and Farzaneh on their lab's practices and the need for comprehensive QC figures.

Segmentation Methods Comparison: Farzaneh asked about segmentation software used for Allen data, and Jerome clarified that a variant of Suite 2P is used, with QC files generated during processing. Ali added that cell pose is also used, and Nicholas confirmed the classifier was trained by David Tang's team using ROI cat, with probability labels stored in NWB files.

ROI and Functional Metrics: Ali described metrics such as ROI count per imaging plane, soma probability, ROI area, and functional measures like DFF variability, robust range, activation frequency, and drift rate. Farzaneh and Nicholas probed the methods for calculating these metrics, including the use of classifiers and thresholding for spike detection.

Motion Artifacts and Drift Analysis: Nicholas raised concerns about drift measurement methods, prompting Jerome to explain the use of Z drift metrics from QC files and the impact of motion artifacts, bleaching, and cell damage. Ali and Jerome discussed how motion correction algorithms are applied, and Shiella detailed the team's approach to laser damage detection and exclusion of compromised sessions.

Group Organization for Analysis: Jerome, Ali, and Nicholas discussed organizing the team into groups focusing on motion correction stability, ROI extraction and trace quality, and cross-session stability. The plan includes developing shared notebooks for loading and analyzing data, with responsibilities divided among participants based on interest and expertise.

Development and Sharing of Analysis Tools: Ali and Nicholas proposed creating and sharing Jupyter or Google Colab notebooks for loading and analyzing mesoscope sessions, enabling all team members to contribute QC analyses and generate figures efficiently.

Notebook Functionality: Nicholas suggested developing a notebook capable of loading individual or multiple sessions, facilitating figure generation for QC analysis. Ali confirmed that his code supports loading multiple sessions and saving results in dictionaries, and both agreed to share and improve the code for broader use.

Data Access Methods: Jerome clarified that analysis can be performed using NWB files or directly from AWS buckets, with the latter providing access to raw movies for more detailed analysis. Ali explained his approach of loading NWB files directly without downloading, and Jerome highlighted the potential for cross-mouse analysis using AWS.

Documentation and Collaboration: Ali committed to sharing well-documented code, and Jerome emphasized the importance of reproducibility and collaborative review. The team agreed to use GitHub discussions to organize metrics, share figures, and avoid redundant work, with a checklist to track analysis progress.

Team Roles and Workflow for QC Analysis: Jerome, Farzaneh, Ali, and Nicholas discussed dividing responsibilities among team members for QC analysis, with plans to start collectively and then split into groups focusing on specific metrics or aspects of the data.

Initial Collective Approach: The team agreed to begin with a shared notebook and collective analysis, allowing everyone to familiarize themselves with the metrics and tools before dividing tasks. Ali suggested that after the initial stage, participants should take responsibility for specific analyses and develop their own metrics and figures.

Group Division by Analysis Type: Jerome proposed organizing the team into three groups: one for motion correction stability, one for ROI extraction and trace quality, and one for cross-session stability. Farzaneh suggested event extraction as a possible fourth group, depending on the team's preferences.

Assignment and Coordination: Farzaneh explained that multiple students from her lab would contribute, dividing the workload based on their familiarity with 2P data analysis. Jerome and Nicholas emphasized the need for clear documentation and coordination via GitHub discussions to ensure efficient progress and avoid duplication.

💬 Join the discussion for this page on GitHub