Optimizing Efficiency of a Clinical Molecular Diagnostic Assay using an Integrated, Fully- Automated Image Analysis Workflow
Dr Tom Nifong
Vice President of Clinical Operations Metamark Genetics, Cambridge, MA
Efficiently incorporating quantitative image analysis into a clinical molecular laboratory is challenging. Quantitative image analysis requires a combination of traditional anatomic pathology tools for managing and staining tissue slides, clinical pathology systems for handling quantitative patient and quality control data, and robust image acquisition and analysis platforms. Image file management, manual selection of regions of interest (ROI) for quantitation, and handling of image-based quantitative data can be time consuming and labor intensive, adding significant costs to each test.
We used CRi Vectra, Definiens Developer XD, Orchard Harvest & Pathology laboratory information systems (LIS), and proprietary Metamark-developed scripting to create a fully integrated laboratory system that minimizes human intervention, improves reproducibility and throughput, and reduces errors and labor costs. We have recently identified a set of protein biomarkers capable of predicting the risk of progression and aggressiveness of newly diagnosed prostate cancer. We used these biomarkers to create a multiplex immunofluorescence assay for in situ tumor identification and biomarker measurement in early stage prostate cancer biopsies. Patient and quality control samples are tracked in our LIS from order entry through automated risk score calculation and report generation. Our system automatically detects and transfers image files from the acquisition computers to our server, manages image file conversions, creates Definiens workspaces, and launches image analysis. Our Developer XD algorithm not only detects ROI and performs quantitative biomarker analysis, but it also incorporates logic for determining tissue and image quality, detecting and eliminating artifacts, mapping cell-line and tissue control slides, and exporting QC and patient data directly into our LIS.