Work

Automated monitoring of Mycobacterium population growth in time-lapse microscopy with Deep Learning pseudo-supervised algorithm

Year: 2025

Type: article

Abstract: The understanding of bacterial population viability through time is a main step in the preclinical development of antitubercular treatments. Tuberculosis, caused by Mycobacterium tuberculosis (Mtb), is one of the leading causes of death worldwide. Observation of Mycobacterium under the microscope applying different technologies provides valuable information to assess the effect of treatments on bacteria. Nevertheless, manual analysis of the images is a time-consuming and difficult process. In this paper, we propose a novel method to automatically assess bacterial populations and infer the growth pattern and growth rate over Time-lapse microscopy (TLM) sequences. This method is based on a weakly-supervised Deep Learning technique based on a light U-Net. To avoid bias and reduce efforts of database construction we deployed and exploited a mix of manually and synthetically annotated databases including two different Mycobacterium species.The experiments demonstrated the efficiency of the detection of both bacterial replication rate and bacterial death rates (in the presence of the antimicrobial isoniazid) with the Phase-Contrast microscopy (PhC) channels as a unique input. Reaching a segmentation accuracy of 96.3%, recall of 78.2%, and precision of 85.5% on Mtb detection. The experiments validate as well the synthetically label generation technique, which provides more consistence and accurate results than the manually segmented techniques employed. The mask generation workflow developed opens new opportunities for microbial evaluation in the Artificial intelligence (AI) field. Overall, the system proposed for bacterial population characterization offers new research avenues for general microbiology, drug development, diagnosis, and treatment, speeding up laboratory processes in these areas.

Source: Engineering Applications of Artificial Intelligence

DOI: 10.1016/j.engappai.2025.112503

Authors: Lara Visuña, Javier Garcia-Blas, Santiago Ferrer-Bazaga, Patricio Lopez-Exposito, Jesus Carretero

Institution: Universidad Carlos III de Madrid

Language: English

Topic: Cell Image Analysis Techniques

Subfield: Biophysics

Field: Biochemistry, Genetics and Molecular Biology

Domain: Life Sciences

Open Access status: hybrid

Estimated APC paid: $3,200