End-to-end plaque counting and virus titration from laboratory plate images with deep learning
Plaque assays are the gold standard for quantifying infectious virus, yet plaque enumeration is still routinely performed manually, making virus titration labor-intensive, subjective, and difficult to standardize across analysts and laboratories. Existing automated methods primarily address individual tasks, such as plaque segmentation or counting, but do not provide an integrated workflow from…
In a groundbreaking development, researchers have unveiled Titra, an end-to-end automated system for quantifying viral particles using images of plaque assay plates. This cutting-edge workflow streamlines the process of virus titration, traditionally a manual and labor-intensive task prone to subjectivity and inconsistencies between analysts and laboratories.
Titra integrates a series of automated steps from well detection to plaque counting, ultimately estimating the number of virus-forming units per millilitre (PFU/mL). This comprehensive solution is accessible through a user-friendly web-based platform, allowing for experiment management and expert review to ensure accuracy and reliability.
The researchers tested Titra using images from three viral species—Mayaro virus, Coxsackievirus B3, and vaccinia virus—acquired with two different plate formats (6- and 12-well) under various image acquisition conditions. These tests included both a newly curated dataset and the publicly available VACVPlaque dataset.
The results demonstrated exceptional performance of the Titra system. Automated plaque counts exhibited a high degree of agreement with manual annotations, with Pearson correlation coefficients of 0.98 for the Mayaro virus and Coxsackievirus B3 datasets, as well as 0.88 for the vaccinia virus dataset. Furthermore, PFU/mL estimates closely matched manual calculations for the Mayaro virus and Coxsackievirus B3 datasets, with a Pearson correlation coefficient of 0.975.
Comparative experiments against other state-of-the-art methods, such as U-Net, StarDist, HSD-WBR, and PyPlaque, showcased Titra's competitive performance in terms of segmentation and counting accuracy. The proposed approach emerged as the top choice on the public VACVPlaque dataset, achieving the highest Dice and mean average precision (mAP) scores.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
