MISA++: A standardized interface for automated bioimage analysis - Original data

Original data for the publication "MISA++: A standardized interface for automated bioimage analysis" These archives contain the complete output of all benchmark experiments reported in the paper and its supplementary material. Each experiment was run across multiple implementations (C++/MISA++, Java/ImgLib2, Python/Snakemake) under different threading and optimization configurations. Naming convention {experiment}_{language}_{framework}_{threading}_{optimization}.7z Experiment Tag Description glomeruli Glomeruli segmentation in 3D LSFM images of whole murine kidneys (23 samples) cells Cell (conidia) segmentation in 2D phagocytosis assay images deconv Image deconvolution via regularized inverse filtering microbench Single-operation benchmarks (canny, fft/ifft, io, median, morphology, otsu, percentile, wiener2) Language & framework Tag Implementation cxx_misaxx C++ using MISA++ with OpenCV java_imglib2 Java using ImgLib2 / ImageJ java_deconvlab2 Java using DeconvolutionLab2 (deconvolution only) python_snakemake Python using Snakemake with scikit-image / NumPy python_custom Original Python implementation without Snakemake (glomeruli only; the ~3-day baseline) Threading Tag Meaning dag30 Parallelized via DAG with 30 threads noMT Single-threaded (no multi-threading) custom Original implementation’s own parallelization (glomeruli Python custom only) Optimization Tag Meaning opMT / optMT Additional library optimizations enabled (OpenCV/NumPy parallelization, etc.) noopMT / nooptMT Additional library optimizations disabled Note: Glomeruli and cells use opMT/noopMT, while deconv and microbench use optMT/nooptMT. Java deconvolution archives use noopMT. The meaning is the same in all cases. File listing Input data File Size Description input_glomeruli.7z 33 GB LSFM input images of 23 whole murine kidneys (OME TIFF) used as input for all glomeruli segmentation runs input_deconv_microbench_cells.7z 4.7 GB Input TIFF images shared by the deconvolution, microbenchmark, and cell segmentation experiments Analysis results File Size Description analysis.7z 14 MB Extracted and aggregated results from all experiments: CSV summaries, runtime logs, statistical test outputs (Bootstrap KS test), shift plots (SVG/PNG), and the R/Python notebooks used to generate figures and tables for the paper Glomeruli segmentation File Size Description glomeruli_cxx_misaxx_dag30_noopMT.7z 1.5 GB MISA++ C++, 30 threads, optimizations disabled glomeruli_cxx_misaxx_dag30_opMT.7z 1.5 GB MISA++ C++, 30 threads, optimizations enabled glomeruli_cxx_misaxx_noMT_nooptMT.7z 730 MB MISA++ C++, single-threaded, optimizations disabled glomeruli_cxx_misaxx_noMT_optMT.7z 730 MB MISA++ C++, single-threaded, optimizations enabled glomeruli_java_imglib2_dag30_noopMT.7z 633 MB Java/ImgLib2, 30 threads, optimizations disabled glomeruli_python_custom_custom_optMT.7z 753 MB Original Python implementation (no Snakemake), optimizations enabled — the ~3-day baseline glomeruli_python_snakemake_dag30_noopMT.7z 449 MB Python/Snakemake, 30 threads, optimizations disabled glomeruli_python_snakemake_dag30_opMT.7z 445 MB Python/Snakemake, 30 threads, optimizations enabled Cell segmentation File Size Description cells_cxx_misaxx_dag30_noopMT.7z 5.9 MB MISA++ C++, 30 threads, optimizations disabled cells_cxx_misaxx_dag30_opMT.7z 6.0 MB MISA++ C++, 30 threads, optimizations enabled cells_cxx_misaxx_noMT_noopMT.7z 5.9 MB MISA++ C++, single-threaded, optimizations disabled cells_cxx_misaxx_noMT_opMT.7z 6.0 MB MISA++ C++, single-threaded, optimizations enabled cells_java_imglib2_dag30_noopMT.7z 134 MB Java/ImgLib2, 30 threads, optimizations disabled cells_java_imglib2_noMT_noopMT.7z 133 MB Java/ImgLib2, single-threaded, optimizations disabled cells_python_snakemake_dag30_noopMT.7z 4.1 MB Python/Snakemake, 30 threads, optimizations disabled cells_python_snakemake_dag30_opMT.7z 4.1 MB Python/Snakemake, 30 threads, optimizations enabled cells_python_snakemake_noMT_noopMT.7z 4.1 MB Python/Snakemake, single-threaded, optimizations disabled cells_python_snakemake_noMT_opMT.7z 4.1 MB Python/Snakemake, single-threaded, optimizations enabled Deconvolution File Size Description deconv_cxx_misaxx_dag30_nooptMT.7z 5.5 GB MISA++ C++, 30 threads, optimizations disabled deconv_cxx_misaxx_dag30_optMT.7z 5.5 GB MISA++ C++, 30 threads, optimizations enabled deconv_cxx_misaxx_noMT_nooptMT.7z 5.5 GB MISA++ C++, single-threaded, optimizations disabled deconv_cxx_misaxx_noMT_optMT.7z 5.5 GB MISA++ C++, single-threaded, optimizations enabled deconv_java_deconvlab2_dag30_noopMT.7z 4.8 GB Java/DeconvolutionLab2, 30 threads, optimizations disabled deconv_java_deconvlab2_noMT_noopMT.7z 4.8 GB Java/DeconvolutionLab2, single-threaded, optimizations disabled deconv_python_snakemake_dag30_nooptMT.7z 3.0 GB Python/Snakemake, 30 threads, optimizations disabled deconv_python_snakemake_dag30_optMT.7z 3.0 GB Python/Snakemake, 30 threads, optimizations enabled deconv_python_snakemake_noMT_nooptMT.7z 3.0 GB Python/Snakemake, single-threaded, optimizations disabled deconv_python_snakemake_noMT_optMT.7z 3.0 GB Python/Snakemake, single-threaded, optimizations enabled Single-operation microbenchmarks File Size Description microbench_cxx_misaxx_nooptMT.7z 6.1 GB MISA++ C++, optimizations disabled microbench_cxx_misaxx_optMT.7z 6.1 GB MISA++ C++, optimizations enabled microbench_java_imglib2_nooptMT.7z 6.9 GB Java/ImgLib2, optimizations disabled microbench_python_snakemake_nooptMT.7z 3.5 GB Python/Snakemake, optimizations disabled microbench_python_snakemake_optMT.7z 3.5 GB Python/Snakemake, optimizations enabled Archive contents Each experiment output archive contains an output-docker/ (MISA++) or output/ (Java/Python) directory with per-sample subfolders. Depending on the experiment, these include: Glomeruli: segmented tissue masks, 2D glomeruli segmentations, 3D reconstructed glomeruli, and quantification results (glomerular count, tissue volume, diameter) as JSON. Cells: segmented cell masks and quantification results (cell counts) as JSON. Deconvolution: convolved and deconvolved output images (TIFF) and runtime logs. Microbenchmarks: per-operation output images (canny, fft/ifft, io, median, morphology, otsu, percentile, wiener2) and runtime logs. The analysis.7z archive contains the extracted CSV summaries, R scripts for statistical testing (Bootstrap Kolmogorov-Smirnov test, shift plots), and generated figures (SVG/PNG) that were used to produce Tables S3–S4 and Figures S4–S10 in the supplementary material. Hardware All experiments were executed on a server with an Intel Xeon E7-8894 CPU, 1 TB of memory, and 30 threads.