Non-destructive label-free automated identification of bacterial colonies at the species level directly on agar media using digital holography and convolutional neural network algorithms.

UNLABELLED: This study aimed to establish a thorough proof of concept for an innovative, fully automated, non-destructive, and label-free approach for the identification of bacterial colonies directly on agar plates, integrating digital holography with artificial intelligence, and to evaluate its performance. High-resolution holographic images of individual bacterial colonies grown on translucent brain-heart agar plates were captured every 30 min over an 18-h incubation period (530 MPx per full plate). Imaging was performed using a large field 1× magnification system, a partially coherent LED light source, and a high-resolution complementary metal-oxide-semiconductor (CMOS) sensor. A database containing 49,490 digital holograms of individual colonies from 276 clinical strains belonging to 10 of the most prevalent pathogenic bacterial species was used to train the convolutional neural network (CNN). Prediction accuracy was further enhanced by incorporating information across different phylogenetic levels. The performance of the BAIO-DX system was evaluated on 232 strains from the 10 species included in the training data set, as well as 64 strains from 8 species not included in the training data set. For the species included in the training data set, this new method identified 86.6% of the strains at the species level, with a positive-percent agreement of 96.5%. An additional 48% of the strains not identified at the species level could be correctly classified at the genus level through phylogenetic interpretation. These results validate this innovative approach as a candidate for a fully automated, non-destructive, and label-free solution for bacterial identification in clinical microbiology laboratories. IMPORTANCE: Identification of pathogenic bacteria in clinical laboratories is traditionally performed using culture-based methods, such as matrix-assisted laser desorption/ionization-time-of-flight (MALDI-TOF) mass spectrometry or biochemical assays. Although recent advances in automation and artificial intelligence (AI) applied to agar plate imaging have reduced manual workload, bacterial identification remains time-consuming and labor-intensive. Here, we present a fully automated, non-destructive, and label-free identification method of bacterial colonies directly on agar plates, combining digital holography and convolutional neural network (CNN) algorithms. After training the system with 276 strains belonging to 10 of the most frequent pathogenic species, the BAIO-DX solution was able to identify 86.6% of new strains at the species level, on a 232-sample test set, achieving a positive-percent agreement of 96.5%. These thorough proof of concept shows that advanced imaging methods with AI-driven analysis can enable reliable, fully automated identification of a substantial proportion of clinically relevant bacteria, offering significant potential to streamline and enhance diagnostic workflows in clinical microbiology.