Ensuring Excellence: The QC Workflow of Indica Labs’ Pharma Services
In our last post, we introduced the image analysis and data visualization experts of Indica Labs’ Pharma Services team, along with their procedures and tools. In this post, we are going to provide a detailed look into their QC processes, which we consider to be some of the most important stages in completing a project.
Our QC team typically consists of an Image Analyst, a Project Lead, and a consulting Pathologist. A project starts with the Image QC step, where our team looks at 100% of your images. Whether it’s a small proof-of-concept study with a handful of images, or a large study with hundreds, our team will have an eye on every image in the set.
During Image QC we ensure that each image is suitable for analysis by looking at:
- Tissue quality – Tissue damage, folds/tears, necrosis
- Scanning quality – Artifacts, blurry regions
- Staining quality – Signal to noise, signal, dynamic range
This allows us to capture inconsistencies early on in the study, enabling the customer to either correct the discrepancies or be aware of them and work around them as we continue the analysis.
ROI QC, the next step in our QC process, goes hand-in-hand with Image QC. During this step we ensure that we are capturing the correct regions of interest based on your study goals.
Our team has experience working with a variety of tissue types, so we often lead the selection of tissue regions for analysis. If requested, we can also collaborate with pathologists to ensure the correct selection of the ROI.
Manual annotations are created to capture the entire region of interest, and we will simultaneously create exclusion annotations to remove artifacts, tissue damage, and adjacent tissue that you are not interested in analyzing. HALO AI also gives us the option to leverage classifiers to automate this process. Indica Labs’ new Slide QC classifier allows us to detect the smallest exclusion regions, and we can develop custom AI classifiers to both capture ROI tissue and exclusion regions.
Once we are confident in image quality and ROI selection, we move on to the cellular analysis. This generally consists of AI classifier development and HALO module optimization. Each image in the study will be QC’d after completion of these milestones. During Classifier QC, we look at the ability of the classifier to delineate the tissue. During Cellular Analysis QC, we ensure that the HALO module selected for the study is working at an optimal level. This ensures proper nuclear segmentation and biomarker thresholding, and also allows us to make additional annotations to include/exclude tissue if necessary. Pathologists are often involved with these steps as well, assuring accurate tissue segmentation and biomarker detection.
The final steps of any Pharma Services project are the Final QC and Secondary QC processes. During Final QC, we finalize all analysis and export the summary data. From the raw summary data, we put together data visualizations depending on the deliverables selected for the study. We also ensure that the data in visualizations matches the data that HALO reports.
Secondary QC is then performed by another member of the team, where they review 25% of the images. This is essentially a second pair of eyes on the classifier markup, cellular analysis, and summary data, to ensure we have fulfilled your goals to the best of our ability.
Throughout the study, you will be able to track each of these QC checkpoints through Indica Labs’ cloud-based, collaborative platform, HALO Link. Then, at completion of the study, you will receive an archive folder that contains all images, markups, classifier and analysis settings, summary data files, project review meeting presentations, and study documents. The most important document you will receive in this folder is the Study Tracker. This document outlines every QC step that was performed. Each image in the study will contain one of the following labels: PASS, SUSPECT, or FAIL.
A “PASS” is straightforward – the image was suitable for analysis and the analysis markup looks accurate. A “SUSPECT” label can point to issues with the image that we feel may affect the analysis, or the analysis itself is not completely representative of the image. Images that warrant a “FAIL” were either not suitable for analysis, or the analysis itself was not acceptable. This information allows you to interpret your data accordingly.
Each of these QC steps in the Pharma Services workflow will give you the ultimate confidence in your data. We look forward to sharing future posts that will detail more about how we provide image analysis services using HALO, HALO AI, and HALO Link. In the meantime, read our white paper to learn how Pharma Services collaborates with customers from project start to finish using HALO Link, and reach out to Pharma Services at email@example.com to discuss how we can advance your biomarker studies.