Automated Counting of Colony Forming Units Using Deep Transfer Learning From a Model for Congested Scenes Analysis

Abstract

Reliable quantification of cellular treatment effects in many bioassays depends on the accuracy of cell colony counting. However, colony counting processes tend to be tedious, slow, and error-prone. Thus, pursuing an effective colony counting technique is ongoing, and varies from manual approaches to partly automated and fully automated techniques. Most fully automated techniques were developed using deep learning (DL). A significant problem in applying DL to this task is the lack of sizeable collections of annotated plate images. For this reason, here we propose an application of Transfer Learning to cell colony counting that can overcome this problem by exploiting models trained for other tasks. To demonstrate this idea's feasibility, we show how a small dataset can be used to transform a DL model designed for counting objects in congested scenes into a specialized cell colony counting model and achieve better performance than existing, more widely-used models.

A comprehensive review of image analysis methods for microorganism counting: from classical image processing to deep learning approaches

“…In Albaradei et al. ( 2020 ), deep transfer learning is applied for automatic pluripotent stem cell colony counting. First, the RGB image is converted to a binary image by using thresholding.…”

“…( 2019 ) Microorganism Histogram peak searching Particle swarm optimization, breadth-first search and exponential entropy Albaradei et al. ( 2020 ) Stem cell Augmentation techniques include color jitter to randomly alter brightness, contrast, saturation, and hue of each image, horizontal/vertical flip, and random rotation. Thresholding SRNetDL Rolke and Lenz ( 1984 ) Zooplankton Mean grey-level selection Quantimet 720 image analysis system (Leica Cambridge Ltd., Cambridge, United Kingdom) Bloem et al.…”

“…( 2020 ) cCNN NVIDIA Quadro K620 – 3.8 times faster 86% Albaradei et al. ( 2020 ) SRNet – – – RMSE = 22.38 …”

Fractal dimension analysis: A new tool for analyzing colony-forming units

“…Some high-throughput methods have been suggested for CFU counting, such as fluorescent labeling, genome probing microarrays, quantitative PCR, but these methods require specific equipment and a very detailed protocol with a lot of inputs [4] . Some artificial intelligence tools, more specifically machine learning, have been proposed to improve the performance of the CFU count, such as the Convolutional Neural Network proposed by Ferrari et al.…”

“… [5] . However, this method is not suitable for high bacterial loads [4] . Thus, other machine learning methods that can evaluate a large number of colonies were also carried out, such as Transfer Learning that was first designed to count objects in a crowded scene [4] .…”