Microsatellite instability (MSI) detection using tumor tissue is a well-established prognostic and predictive biomarker for certain types of cancers. However, tumor tissue samples are less convenient to obtain than blood plasma samples. The main challenge facing next-generation sequencing-based MSI detection in blood plasma samples is the ultralow signal/noise ratio in plasma cell-free DNA (cfDNA). To address the challenge, plasmaMSI, a highly accurate cfDNA MSI detection method, is introduced with three novel performance-improving features: i) a set of stringent locus selection criteria to select loci with high robustness and compatibility across sequencing platforms; ii) a new deduplication strategy that greatly improves the signal/noise ratio for MSI detection; and iii) an MSI calling algorithm that customizes the baseline for each test sample based on its duplication rate. Through analytical validation in diluted cell line samples, the limit of detection of plasmaMSI was determined to be 0.15%. Furthermore, in analyzing 95 evaluable cfDNA samples from patients with gastrointestinal cancers, plasmaMSI exhibited a positive percentage agreement of 92.9% (39/42) and a negative percentage agreement of 100% (53/53) with tissue MSI-PCR. plasmaMSI provides novel solutions to key challenges in cfDNA MSI detection that have not been addressed by existing methods. It has also been systematically validated and is already used in clinical testing for patients with cancer.Copyright © 2025 Association for Molecular Pathology and American Society for Investigative Pathology. Published by Elsevier Inc. All rights reserved.