BioPII 0.2.4

BioPII (Biology Parallel Integral Image) is a Python package for performing sliding window analysis (SWA) on biological images.

BioPII

BioPII (Biology Parallel Integral Image) is a Python package for performing sliding window analysis (SWA) on biological images. This package makes use of the integral image technique to drastically speedup SWA, enabling SWA analysis of images that would have been too large for previous SWA implementations.

To use the package include the following statement at the top of your file: from BioPII import PII

To cite this package, cite our IEEE CIBCB 2023 short paper.

This code and package is under the MIT License - copyright (c) Seth Ockerman

The package name is prounounced bio-pie :)

Performance Numbers

We tested our summing SWA to determine its performance relative to past methods. We created a C++ SWA (referred to as DP-Naive) script which used dynamic programming to reduce the number of additions needed to be performed. DP-Naive SWA was already 400x faster than a naive approach (a simple four for loop approach) on small images and is used as the baseline for comparison to our integral-image-based methods.

	30k by 30k	30k by 30k Speedup	40k by 50k	40k by 50k Speedup	60k by 60k	60k by 60k Speedup
DP-Naive	211,381,433	N/A	491,374,150	N/A	943,858,845	N/A
Integral Image	56,693	3728x	253,082	1942x	319,666	2953x
Parallel Integral Image	20,137	10,497x	46,532	10,559x	86,583	10,901x

Note: SWA runtime in milliseconds on different image sizes; speedups relative to DP-Naive Solution

Features

• Efficient computation for various biological image analysis tasks.
• Supports different types of SWA (Sliding Window Analysis) algorithms.
• Flexible window size selection for analyzing different cellular features.
• GPU acceleration option for faster computations.
• Works with numpy arrays for easy integration into existing workflows.

Installation

You can install BioPII using pip:

pip3 install BioPII

Documentation

Documentation can be found internally within our code and in the documentation file.

Note

The average-SWA-algorithm and std-deviation-SWA algorithm have not been as tested as extensively as the sum-SWA-algorithm. If you discover a bug, please reach out to [e-mail].