CmIF: A Python Library for Scalable Multiplex Imaging Pipelines

Abstract

Histological staining and analysis of tissue sections is integral to diagnosis and treatment of many diseases, including cancer. Multiplex imaging technologies (e.g., cyclic immunostaining) have dramatically increased capabilities for assessing prognostic biomarkers in situ, enabling new insights into complex diseases. However, high-resolution, multiplex image data can be terabytes (TB) in size, and traditional pipelines for image analysis are not suited for these rich datasets. While much software development effort goes towards improving image processing tools such as stitching, registration, and segmentation; integration of these tools into a pipeline is often manual, which is highly laborious, error-prone and lacks reproducibility and scalability. Therefore, we developed a Python3 library, cmIF, a free and open-source tool to handle our high-throughput multiplex image processing pipeline. cmIF enables analysis of full-slide pathology tissue sections and tissue microarrays (TMAs), facilitating processing from raw image files through registration, segmentation, feature extraction, manual thresholding, and spatial pattern analysis. Our cmIF library includes functionality for image handling, quality control, metadata extraction, and subtraction of background images (i.e., autofluorescence subtraction). Additionally, it includes a Jupyter notebook for efficient generation and visualization of manual thresholds. Compared to a manual pipeline, use of cmIF reduces errors and improves processing time of datasets from weeks to hours, while documenting processing steps for reproducibility. All code is available on https://gitlab.com/engje/cmif. While our library is specific to our pipeline elements, it is a blueprint for types of functions needed for high throughput analysis. In the future, we will continue developing this open-source tool, and with input from the wider community, adapt it to a range of multiplex image pipelines.