zhe-ch / ACTEV

Accelerator for Calcium Trace Extraction from Video

Accelerator for Calcium Trace Extraction from Video (ACTEV)

1. FPGA and Embedded Software

The ACTEV was built on top of the Ultra96 Platform.

1.1 FPGA

Please refer to Xilinx Support Archive (XSA) files under ./FPGA/Vivado/. The precompiled FPGA image is under ./FPGA/SD/. You could simply rely on the image to boot the FPGA and run ACTEV under the SD mode. This directory contains two separate image files, located in ./Linear_Classifier and ./Tree-Based_Classifier sub folders.

1.2 Embedded Software

Source code for the embedded software is under ./FPGA/Vitis/. Embedded software with linear classifier and tree-based classifier functions are arranged in separate projects.

If you would like to rebuild the embedded software in Vitis, please remember to include the ./FPGA/Vitis/EmbeddedSw/ in your project local directory.

2. PCB

The interface PCB sits on top of the Ultra96 board. It provides Ethernet connection to a host PC and a 14-pin flywire connection to the Data Acquisition Box (DAQ). Fabrication files are under ./PCB/.

3. Simulation

Matlab and C programs can simulate ACTEV behaviors at high level accurately.

3.1 Matlab

We provide 3 sets of Matlab scripts for analyzing calcium image data offline:

1. Motion Correction: Frame-based rigid motion correction.
2. Trace Extraction: Separate files supporting cell and tile based trace extraction.
3. Linear Classifier Training: Provide training and inference code for the linear classifier.

3.2 C Program

We also provide following C programs as a behavioral simulation model for ACTEV:

1. Extract Classifier Parameter: Convert Matlab generated linear classifier parameters into loadable binary file for ACTEV.
2. Extract Template: Extract mean of recorded images as a template for the motion correction C simulation.
3. Linear Classifier Inference: Perform simulation of the position decoding based on linear classifier parameters.
4. Motion Correction: Frame-based rigid motion correction.
5. Trace Extraction: Support cell and tile based trace extraction, with provided example contour files.

4. RTI Interface

We developed real-time interface (RTI) in C# to support user interaction with ACTEV. Source code of the RTI development is under ./Software/. Example loadable files for the RTI are collected under ./Software/file/:

1. Cell Contour: cell/tile contour in text/binary and other parameter files in text format.
2. Filter Kernel: Constant 17x17 Gaussian contrast filter kernel in binary format.
3. Template: Example 128x128 template image for motion correction in text format.