Towards automated detection, semi-quantification and identification of microbial growth in clinical bacteriology: A proof of concept

Interpretation of Bacterial Smears and Cultures Using Artificial Intelligence

The use of artificial intelligence (AI) to aid in the diagnosis of infectious diseases is a growing area of interest. AI-based applications in the clinical microbiology laboratory have shown great prospect for the automated interpretation of smears and cultures through digitalization. As automation becomes more common and the algorithms become more accurate, these can optimize laboratory efficiency, workflows, and turnaround times. This review will focus on summarizing the performance of AI-based methods for the interpretation of bacteriology smears and cultures.

The Impact of Laboratory Automation on the Time to Urine Microbiological Results: A Five-Year Retrospective Study

Total laboratory automation (TLA) is a valuable component of microbiology laboratories and a growing number of publications suggest the potential impact of automation in terms of analysis standardization, streaking quality, and the turnaround time (TAT). The aim of this project was to perform a detailed investigation of the impact of TLA on the workflow of commonly treated specimens such as urine. This is a retrospective observational study comparing two time periods (pre TLA versus post TLA) for urine specimen culture processing. A total of 35,864 urine specimens were plated during the pre-TLA period and 47,283 were plated during the post-TLA period. The median time from streaking to identification decreased from 22.3 h pre TLA to 21.4 h post TLA (p < 0.001), and the median time from streaking to final validation of the report decreased from 24.3 h pre TLA to 23 h post TLA (p < 0.001). Further analysis revealed that the observed differences in TAT were mainly driven by the contaminated and positive samples. Our findings demonstrate that TLA has the potential to decrease turnaround times of samples in a laboratory. Nevertheless, changes in laboratory workflow (such as extended opening hours for plate reading and antibiotic susceptibility testing or decreased incubation times) might further maximize the efficiency of TLA and optimize TATs.

Development and evaluation of an artificial intelligence for bacterial growth monitoring in clinical bacteriology

In clinical bacteriology laboratories, reading and processing of sterile plates remain a significant part of the routine workload (30%-40% of the plates). Here, an algorithm was developed for bacterial growth detection starting with any type of specimens and using the most common media in bacteriology. The growth prediction performance of the algorithm for automatic processing of sterile plates was evaluated not only at 18-24 h and 48 h but also at earlier timepoints toward the development of an early growth monitoring system. A total of 3,844 plates inoculated with representative clinical specimens were used. The plates were imaged 15 times, and two different microbiologists read the images randomly and independently, creating 99,944 human ground truths. The algorithm was able, at 48 h, to discriminate growth from no growth with a sensitivity of 99.80% (five false-negative [FN] plates out of 3,844) and a specificity of 91.97%. At 24 h, sensitivity and specificity reached 99.08% and 93.37%, respectively. Interestingly, during human truth reading, growth was reported as early as 4 h, while at 6 h, half of the positive plates were already showing some growth. In this context, automated early growth monitoring in case of normally sterile samples is envisioned to provide added value to the microbiologists, enabling them to prioritize reading and to communicate early detection of bacterial growth to the clinicians.

Machine learning algorithms in microbial classification: a comparative analysis

This research paper presents an overview of contemporary machine learning methodologies and their utilization in the domain of healthcare and the prevention of infectious diseases, specifically focusing on the classification and identification of bacterial species. As deep learning techniques have gained prominence in the healthcare sector, a diverse array of architectural models has emerged. Through a comprehensive review of pertinent literature, multiple studies employing machine learning algorithms in the context of microbial diagnosis and classification are examined. Each investigation entails a tabulated presentation of data, encompassing details about the training and validation datasets, specifications of the machine learning and deep learning techniques employed, as well as the evaluation metrics utilized to gauge algorithmic performance. Notably, Convolutional Neural Networks have been the predominant selection for image classification tasks by machine learning practitioners over the last decade. This preference stems from their ability to autonomously extract pertinent and distinguishing features with minimal human intervention. A range of CNN architectures have been developed and effectively applied in the realm of image classification. However, addressing the considerable data requirements of deep learning, recent advancements encompass the application of pre-trained models using transfer learning for the identification of microbial entities. This method involves repurposing the knowledge gleaned from solving alternate image classification challenges to accurately classify microbial images. Consequently, the necessity for extensive and varied training data is significantly mitigated. This study undertakes a comparative assessment of various popular pre-trained CNN architectures for the classification of bacteria. The dataset employed is composed of approximately 660 images, representing 33 bacterial species. To enhance dataset diversity, data augmentation is implemented, followed by evaluation on multiple models including AlexNet, VGGNet, Inception networks, Residual Networks, and Densely Connected Convolutional Networks. The results indicate that the DenseNet-121 architecture yields the optimal performance, achieving a peak accuracy of 99.08%, precision of 99.06%, recall of 99.00%, and an F1-score of 98.99%. By demonstrating the proficiency of the DenseNet-121 model on a comparatively modest dataset, this study underscores the viability of transfer learning in the healthcare sector for precise and efficient microbial identification. These findings contribute to the ongoing endeavors aimed at harnessing machine learning techniques to enhance healthcare methodologies and bolster infectious disease prevention practices.

Rapid and automated screening of carbapenemase- and ESBL-producing Gram-negative bacteria from rectal swabs using chromogenic agar media and the ScanStation device

The ScanSation 100 device (Interscience, France) is an incubator allowing real-time detection of bacterial colony growth by frequently imaging agar plates over time, counting CFU, and detecting colony color. This study evaluated its performance for the early detection of carbapenemase-producing bacteria (CPB) and extended-spectrum β-Lactamase-producing bacteria (ESBL-PB) from rectal swabs inoculated on CHROMagar mSuperCARBA and ESBL media, respectively. Rectal screening ESwabs collected from patients admitted to Grenoble University Hospital between January and June 2021 were analyzed. After inoculation, chromogenic media were incubated for 24 h in the automaton, with image acquisition every 30 min. ScanStation results were compared to visual observations of the plates after 24 h of incubation. In total, 501 rectal swabs were tested. ScanStation showed 100% positive percent agreement (PPA) for the detection of CPB and ESBL-PB, whereas the PPA of color categorization ranged between 45% and 100%. Negative percent agreement (NPA) ranged between 70% and 98%. Negative predictive values (NPVs) were 100% for both bacterial groups, whereas positive predictive values (PPVs) were 70.3% for CPB and 81.0% for ESBL-PB. Importantly, real-time screening allowed detection of the first suspected colony within 10-14 h of growth, on average, whereas visual observation is usually only performed once a day after 18-24 h of incubation. Our study demonstrates the potential use of early images to accelerate the detection of CPB and ESBL-PB and implement effective and timely infection control measures. IMPORTANCE The ScanStation 100 device is an incubator able to follow the real-time growth of bacterial colonies on agar plates through digital imaging, allowing users to sort plates according to the presence or absence of colonies, and to distinguish their color using four numeric color filters. Real-time screening shows that first colony detection is possible much earlier (after 10-14 h of growth, on average), whereas visual observation is usually performed only once a day after 18-24 h of incubation. The ScanStation device, combined with chromogenic agar media, is an efficient automated screening method to accelerate the detection of Gram-negative multidrug-resistant bacteria in laboratories that do not have access to larger laboratory automation systems. Our study shows that setting the image acquisition to one or two early images may allow for the detection of positive samples that were inoculated in the morning, by the end of the working day.

Laboratory automation, informatics, and artificial intelligence: current and future perspectives in clinical microbiology

Clinical diagnostic laboratories produce one product-information-and for this to be valuable, the information must be clinically relevant, accurate, and timely. Although diagnostic information can clearly improve patient outcomes and decrease healthcare costs, technological challenges and laboratory workflow practices affect the timeliness and clinical value of diagnostics. This article will examine how prioritizing laboratory practices in a patient-oriented approach can be used to optimize technology advances for improved patient care.

Rapid, Low-Complexity, Simultaneous Bacterial Group Identification and Antimicrobial Susceptibility Testing Performed Directly on Positive Blood Culture Bottles Using Chromogenic Agar

The use of positive blood culture bottles for direct disk diffusion susceptibility testing (dDD), together with chromogenic culture limited to groups of pathogens for antimicrobial susceptibility testing interpretation may provide a means for laboratories-in-development to introduce rapid abbreviated blood culture testing. We assessed the performance of dDD on Chromatic MH agar using contrived positive blood culture bottles and compared findings with current standard practice. Furthermore, we characterized the growth of 24 bacterial and 3 yeast species on Chromatic MH agar with the aid of rapid spot tests for same-day identification. The coefficient of variation for reproducibility of dDD of four reference strains in 4 to 10 replicates (238 data points) ranged from 0% to 16.3%. Together with an additional 10 challenge isolates, the overall categorical agreement was 91.7% (351 data points). The following bacteria were readily identifiable: cream/white Staphylococcus aureus, coagulase-negative staphylococci, Streptococcus pyogenes; turquoise Streptococcus agalactiae, enterococci, Listeria monocytogenes; mauve Escherichia coli, Shigella sonnei, Citrobacter freundii; dark-blue Klebsiella and Enterobacter; green Pseudomonas aeruginosa; and brown Proteus. Clear colonies were seen with Salmonella, Acinetobacter, Burkholderia, and Yersinia enterocolitica (turns pink). Our study suggests that Chromatic MH for dDD may show promise as a rapid, clinically useful presumptive method for overnight simultaneous identification and antimicrobial susceptibility testing. However, there is a need to optimize the medium formulation to allow the recovery of Streptococcus pneumoniae and Haemophilus influenzae.

Molecular epidemiology of bovine tuberculosis in Northern Ghana identifies several uncharacterized bovine spoligotypes and suggests possible zoonotic transmission

Objective

We conducted an abattoir-based cross-sectional study in the five administrative regions of Northern Ghana to determine the distribution of bovine tuberculosis (BTB) among slaughtered carcasses and identify the possibility of zoonotic transmission.

Methods

Direct smear microscopy was done on 438 tuberculosis-like lesions from selected cattle organs and cultured on Lowenstein-Jensen media. Acid-fast bacilli (AFB) isolates were confirmed as members of the Mycobacterium tuberculosis complex (MTBC) by PCR amplification of IS6110 and rpoß. Characterization and assignment into MTBC lineage and sub-lineage were done by spoligotyping, with the aid of the SITVIT2, miruvntrplus and mbovis.org databases. Spoligotype data was compared to that of clinical M. bovis isolates from the same regions to identify similarities.

Results

A total of 319/438 (72.8%) lesion homogenates were smear positive out of which, 84.6% (270/319) had microscopic grade of at least 1+ for AFB. Two hundred and sixty-five samples (265/438; 60.5%) were culture positive, of which 212 (80.0%) were MTBC. Approximately 16.7% (34/203) of the isolates with correctly defined spoligotypes were negative for IS6110 PCR but were confirmed by rpoß. Spoligotyping characterized 203 isolates as M. bovis (198, 97.5%), M. caprae (3, 1.5%), M. tuberculosis (Mtbss) lineage (L) 4 Cameroon sub-lineage, (1, 0.5%), and M. africanum (Maf) L6 (1, 0.5%). A total of 53 unique spoligotype patterns were identified across the five administrative regions (33 and 28 were identified as orphan respectively by the SITVIT2 and mbovis.org databases), with the most dominant spoligotype being SIT1037/ SB0944 (77/203, 37.93%). Analysis of the bovine and human M. bovis isolates showed 75% (3/4) human M. bovis isolates sharing the same spoligotype pattern with the bovine isolates.

Conclusion

Our study identified that approximately 29% of M. bovis strains causing BTB in Northern Ghana are caused by uncharacterized spoligotypes. Our findings suggest possible zoonotic transmission and highlight the need for BTB disease control in Northern Ghana.

Current State of Laboratory Automation in Clinical Microbiology Laboratory

BACKGROUND

Although it has been 30 years since the first automation systems were introduced in the microbiology laboratory, total laboratory automation (TLA) has only recently been recognized as a valuable component of the laboratory. A growing number of publications illustrate the potential impact of automation. TLA can improve standardization, increase laboratory efficiency, increase workplace safety, and reduce long-term costs.

CONTENT

This review provides a preview of the current state of automation in clinical microbiology and covers the main developments during the last years. We describe the available hardware systems (that range from single function devices to multifunction workstations) and the challenging alterations on workflow and organization of the laboratory that have to be implemented to optimize automation.

SUMMARY

Despite the many advantages in efficiency, productivity, and timeliness that automation offers, it is not without new and unique challenges. For every advantage that laboratory automation provides, there are similar challenges that a laboratory must face. Change management strategies should be used to lead to a successful implementation. TLA represents, moreover, a substantial initial investment. Nevertheless, if properly approached, there are a number of important benefits that can be achieved through implementation of automation in the clinical microbiology laboratory. Future developments in the field of automation will likely focus on image analysis and artificial intelligence improvements. Patient care, however, should remain the epicenter of all future directions and there will always be a need for clinical microbiology expertise to interpret the complex clinical and laboratory information.

Efficient automated semi-quantitative urine culture analysis via BD Urine Culture App

We aimed to assess the clinical utility of BD Kiestra^TM Urine Culture App (UCA). High concordance rates were observed between the urine culture colony counts obtained by medical technologists and those produced using UCA. This application may increase the efficiency of obtaining semi-quantitative urine culture results.

State-of-the-Art Techniques for Diagnosis of Medical Parasites and Arthropods

Conventional methods such as microscopy have been used to diagnose parasitic diseases and medical conditions related to arthropods for many years. Some techniques are considered gold standard methods. However, their limited sensitivity, specificity, and accuracy, and the need for costly reagents and high-skilled technicians are critical problems. New tools are therefore continually being developed to reduce pitfalls. Recently, three state-of-the-art techniques have emerged: DNA barcoding, geometric morphometrics, and artificial intelligence. Here, data related to the three approaches are reviewed. DNA barcoding involves an analysis of a barcode sequence. It was used to diagnose medical parasites and arthropods with 95.0% accuracy. However, this technique still requires costly reagents and equipment. Geometric morphometric analysis is the statistical analysis of the patterns of shape change of an anatomical structure. Its accuracy is approximately 94.0-100.0%, and unlike DNA barcoding, costly reagents and equipment are not required. Artificial intelligence technology involves the analysis of pictures using well-trained algorithms. It showed 98.8-99.0% precision. All three approaches use computer programs instead of human interpretation. They also have the potential to be high-throughput technologies since many samples can be analyzed at once. However, the limitation of using these techniques in real settings is species coverage.

Digital imaging for reading of direct rapid antibiotic susceptibility tests from positive blood cultures

Delaying effective antibiotic therapy is a major cause of sepsis-associated mortality. The EUCAST rapid antibiotic susceptibility test (RAST) is performed from positive blood cultures to provide rapid results. Disc diffusion tests inoculated with positive blood culture broth are read at 4, 6, and 8 h and interpreted against species and time-specific criteria. Potential problems are the possibility of missing specific reading times for tests and slower growth in incubators that are frequently opened. The current study aimed to assess if digital visualization by the BD Kiestra™ total laboratory automation system is suitable for reading RASTs by capturing images at the correct times and retaining them for review. Utilizing the Kiestra™ InoqulA, 100 μl of positive blood culture broth was lawn-inoculated onto Mueller-Hinton agar and incubated at 35 °C for automated digital zone measurement at 4, 6, and 8 h. Aliquots from 135 positive blood cultures were tested against EUCAST-recommended and other drugs and assessed for readability of digital images. Microdilution MICs were determined in parallel to RASTs. All isolates except 7/10 enterococci yielded images of suitable quality for zone measurement. Of the 641 digitally read tests for other organisms, 207 (32.3%) were readable in 4 h, 555 (86.6%) in 6 h, and 641 (100%) in 8 h. For tests included in EUCAST criteria, 92.1% provided categorical agreement with microdilution MICs. Digital image reading of RASTs is a potentially viable, inexpensive tool for providing rapid susceptibility results which can help reduce sepsis-associated mortality.

Benefits Derived from Full Laboratory Automation in Microbiology: a Tale of Four Laboratories

Automation in clinical microbiology is starting to become more commonplace and reportedly offers several advantages over the manual laboratory. Most studies have reported on the rapid turnaround times for culture results, including times for identification of pathogens and their respective antimicrobial susceptibilities, but few have studied the benefits from a laboratory efficiency point of view. This is the first large, multicenter study in North America to report on the benefits derived from automation measured in full-time equivalents (FTE), FTE reallocation, productivity, cost per specimen, and cost avoidance. Pre- and post-full automation audits were done at 4 laboratories that have vastly different culture volumes, and results show that regardless of the size of the facility, improved efficiencies can be realized after implementation of full laboratory automation.

Use of artificial intelligence for tailored routine urine analyses

OBJECTIVES

Urine is the most common material tested in clinical microbiology laboratories. Automated analysis is already performed, permitting quicker results and decreasing the laboratory technologist's (LT) workload. These automatic systems have introduced digital imaging concepts. PhenoMATRIX (PHM) is an artificial intelligence software that merges picture algorithms and user rules to provide presumptive results. This study aimed at designing a tailored workflow using PHM, performing its validation and checking its performance in routine practice.

METHODS

Two data collections including 96 and 135 urine samples from nephrostomy/ureterostomy and artificial bladder (US), 948 and 1257 urine samples from catheter (UC) and 3251 and 2027 midstream urine (MSU) were used to compare LT results with those obtained using two versions of PHM. Another 19 US, 102 UC and 508 MSU were used to monitor performance level 3 months after routine implementation.

RESULTS

Before and after revisions, agreement between the first version of PHM and LT results were 83% (95% confidence interval [CI], 74.3-90.2) and 83% (95% CI, 75.3-90.9) (US), 66.7% (95% CI, 63.5-69.5) and 71.7% (95% CI, 68.8-74.4) (UC) and 65.4% (95% CI, 63.8-67.1) and 76% (95% CI, 74.1-77.1) (MSU). The second version improved results, demonstrating 96.2% (95% CI, 91.6-98.8) and 97% (95% CI, 92.6-99.2) (US), 87.5% (95% CI, 85.5-89.2) and 88.9% (95% CI, 87.0-90.5) (UC) and 91% (95% CI, 89.7-92.1) and 92% (95% CI, 91.1-93.4) (MSU) of agreement with LT results before and after revisions. The reliability of PHM results was confirmed by a routine study demonstrating 92% (95% CI, 90.0-94.2) overall agreement.

CONCLUSIONS

PHM showed high performance, with >90% of results in agreement with LT. PHM could help standardize and secure results, prioritize positive plates during analytical workflow and likely save LT time.

Performance evaluation of the Becton Dickinson Kiestra™ IdentifA/SusceptA

OBJECTIVES

New automated modules are required to provide fully automated solutions in diagnostic microbiology laboratories. We evaluated the performance of a Becton Dickinson Kiestra™ IdentifA/SusceptA prototype for MALDI-TOF identification (ID) and Phoenix™ antibiotic susceptibility testing (AST).

METHODS

The performance of the IdentifA/SusceptA coupled prototype was compared with manual processing for MALDI-TOF ID on 1302 clinical microbial isolates or ATCC strains and for Phoenix™ M50 AST on 484 strains, representing 61 species.

RESULTS

Overall, the IdentifA exhibited similar ID performances than manual spotting. Higher performances were observed for Gram-negative bacteria with an ID at the species level (score >2) of 96.5% (369/382) and 86.9% (334/384), respectively. A significantly better performance was observed with the IdentifA (95.2%, 81/85) compared with manual spotting (75.2%, 64/85) from colonies on MacConkey agar. Contrariwise, the IdentifA exhibited lower ID performances at the species level than manual processing for streptococci (76.1%, 96/126 compared with 92%, 115/125), coagulase-negative staphylococci (73.3%, 44/60 compared with 90%, 54/60) and yeasts (41.3%, 19/46 compared with 78.2%, 36/46). Staphylococcus aureus and enterococci were similarly identified by the two approaches, with ID rates of 92% (65/70) for the IdentifA and 92.7%, (64/69) for manual processing and 94.8%, (55/58) for the IdentifA and 98.2%, (57/58) for manual processing, respectively. The SusceptA exhibited an AST overall essential agreement of 98.82% (6863/6945), a category agreement of 98.86% (6866/6945), 1.05% (6/570) very major errors, 0.16% (10/6290) major errors, and 0.91% (63/6945) minor errors compared to the reference AST.

CONCLUSIONS

Overall, the automated IdentifA/SusceptA exhibited high ID and AST performances.

Automated preparation for identification and antimicrobial susceptibility testing: evaluation of a research use only prototype, the BD Kiestra IdentifA/SusceptA system

OBJECTIVE

The current BD Kiestra™ total laboratory automation (TLA) system automates specimen inoculation, incubation, and digital visualization of cultures prior to initiation of manual or semi-automated identification (ID) and antimicrobial susceptibility testing (AST). The current study aimed to compare the performance, in a clinical setting, of a fully automated research-use-only prototype, BD Kiestra™ IdentifA/SusceptA (automated system), to our current BD Kiestra™ TLA which utilizes manual or semi-automated IDs and ASTs (current system).

METHODS

Clinical samples yielding significant growth after processing by the BD Kiestra™ TLA were tested in parallel for ID and AST by both systems. IDs and ASTs were determined by Bruker matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry and BD Phoenix, respectively, with data stored and managed in the BD EpiCenter™. The automated system used a common inoculum preparation for both tests, whereas the current system used separate inocula. Results were compared to assess agreement between the systems.

RESULTS

On initial testing, 89% of IDs (466/523) and 92.4% of IDs (484/523) for the automated and current ID systems, respectively, yielded acceptable MALDI-TOF log scores of ≥1.7. On repeat testing, the respective acceptable scores were 97.1% (508/523) and 98.1% (513/523). For initial ASTs, the automated and current systems yielded 97.5% categorical agreement for 7325 drug-organism tests. After omitting discrepant MICs that differed by only one dilution and categorical discrepancies that were not reproducible, 0.2% unresolved discrepancies remained thus (99.8% categorical agreement).

CONCLUSIONS

The automated prototype is suitable for development into technology that will provide clinical microbiology laboratories with significant advantages such as improved efficiency, standardization, reproducibility, reduced technical error and greater safety.

Digital microbiology

BACKGROUND

Digitalization and artificial intelligence have an important impact on the way microbiology laboratories will work in the near future. Opportunities and challenges lie ahead to digitalize the microbiological workflows. Making efficient use of big data, machine learning, and artificial intelligence in clinical microbiology requires a profound understanding of data handling aspects.

OBJECTIVE

This review article summarizes the most important concepts of digital microbiology. The article gives microbiologists, clinicians and data scientists a viewpoint and practical examples along the diagnostic process.

SOURCES

We used peer-reviewed literature identified by a PubMed search for digitalization, machine learning, artificial intelligence and microbiology.

CONTENT

We describe the opportunities and challenges of digitalization in microbiological diagnostic processes with various examples. We also provide in this context key aspects of data structure and interoperability, as well as legal aspects. Finally, we outline the way for applications in a modern microbiology laboratory.

IMPLICATIONS

We predict that digitalization and the usage of machine learning will have a profound impact on the daily routine of laboratory staff. Along the analytical process, the most important steps should be identified, where digital technologies can be applied and provide a benefit. The education of all staff involved should be adapted to prepare for the advances in digital microbiology.

Consolidation of Clinical Microbiology Laboratories and Introduction of Transformative Technologies

Clinical microbiology is experiencing revolutionary advances in the deployment of molecular, genome sequencing-based, and mass spectrometry-driven detection, identification, and characterization assays. Laboratory automation and the linkage of information systems for big(ger) data management, including artificial intelligence (AI) approaches, also are being introduced. The initial optimism associated with these developments has now entered a more reality-driven phase of reflection on the significant challenges, complexities, and health care benefits posed by these innovations. With this in mind, the ongoing process of clinical laboratory consolidation, covering large geographical regions, represents an opportunity for the efficient and cost-effective introduction of new laboratory technologies and improvements in translational research and development. This will further define and generate the mandatory infrastructure used in validation and implementation of newer high-throughput diagnostic approaches. Effective, structured access to large numbers of well-documented biobanked biological materials from networked laboratories will release countless opportunities for clinical and scientific infectious disease research and will generate positive health care impacts. We describe why consolidation of clinical microbiology laboratories will generate quality benefits for many, if not most, aspects of the services separate institutions already provided individually. We also define the important role of innovative and large-scale diagnostic platforms. Such platforms lend themselves particularly well to computational (AI)-driven genomics and bioinformatics applications. These and other diagnostic innovations will allow for better infectious disease detection, surveillance, and prevention with novel translational research and optimized (diagnostic) product and service development opportunities as key results.

Machine learning in the clinical microbiology laboratory: has the time come for routine practice?

BACKGROUND

Machine learning (ML) allows the analysis of complex and large data sets and has the potential to improve health care. The clinical microbiology laboratory, at the interface of clinical practice and diagnostics, is of special interest for the development of ML systems.

AIMS

This narrative review aims to explore the current use of ML In clinical microbiology.

SOURCES

References for this review were identified through searches of MEDLINE/PubMed, EMBASE, Google Scholar, biorXiv, arXiV, ACM Digital Library and IEEE Xplore Digital Library up to November 2019.

CONTENT

We found 97 ML systems aiming to assist clinical microbiologists. Overall, 82 ML systems (85%) targeted bacterial infections, 11 (11%) parasitic infections, nine (9%) viral infections and three (3%) fungal infections. Forty ML systems (41%) focused on microorganism detection, identification and quantification, 36 (37%) evaluated antimicrobial susceptibility, and 21 (22%) targeted the diagnosis, disease classification and prediction of clinical outcomes. The ML systems used very diverse data sources: 21 (22%) used genomic data of microorganisms, 19 (20%) microbiota data obtained by metagenomic sequencing, 19 (20%) analysed microscopic images, 17 (18%) spectroscopy data, eight (8%) targeted gene sequencing, six (6%) volatile organic compounds, four (4%) photographs of bacterial colonies, four (4%) transcriptome data, three (3%) protein structure, and three (3%) clinical data. Most systems used data from high-income countries (n = 71, 73%) but a significant number used data from low- and middle-income countries (n = 36, 37%). Performance measures were reported for the 97 ML systems, but no article described their use in clinical practice or reported impact on processes or clinical outcomes.

IMPLICATIONS

In clinical microbiology, ML has been used with various data sources and diverse practical applications. The evaluation and implementation processes represent the main gap in existing ML systems, requiring a focus on their interpretability and potential integration into real-world settings.

Clinical microbiology

The clinical microbiology laboratory relies on traditional diagnostic methods such as culturing, Gram stains, and biochemical testing. Receipt of a high-quality specimen with an appropriate test order is integral to accurate testing. Recent technological advancements have led to decreased time to results and improved diagnostic accuracy. Examples of advancements discussed in this chapter include automation of bacterial culture processing and incubation, as well as introduction of mass spectrometry for the proteomic identification of microorganisms. In addition, molecular testing is increasingly common in the clinical laboratory. Commercially available multiplex molecular assays simultaneously test for a broad array of syndromic-related pathogens, providing rapid and sensitive diagnostic results. Molecular advancements have also transformed point-of-care (POC) microbiology testing, and molecular POC assays may largely supplant traditional rapid antigen testing in the future. Integration of new technologies with traditional testing methods has led to improved quality and value in the clinical microbiology laboratory.

After reviewing this chapter, the reader will be able to:•

List key considerations for specimen collection for microbiology testing.

•

Discuss the advantages and limitations of automation in the clinical microbiology laboratory.

•

Describe the evolution of microorganism identification methods.

•

Discuss the benefits and limitations of molecular microbiology point-of-care testing.

•

Summarize currently available multiplex molecular microbiology testing options.

Automation and artificial intelligence in the clinical laboratory

The daily operation of clinical laboratories will be drastically impacted by two disruptive technologies: automation and artificial intelligence (the development and use of computer systems able to perform tasks that normally require human intelligence). These technologies will also expand the scope of laboratory medicine. Automation will result in increased efficiency but will require changes to laboratory infrastructure and a shift in workforce training requirements. The application of artificial intelligence to large clinical datasets generated through increased automation will lead to the development of new diagnostic and prognostic models. Together, automation and artificial intelligence will support the move to personalized medicine. Changes in pathology and clinical doctoral scientist training will be necessary to fully participate in these changes. KEYWORDS: Automation; artificial intelligence; deep learning; laboratory medicine.

Clinical Microbiology Is Growing Up: The Total Laboratory Automation Revolution

BACKGROUND

Historically, culture-based microbiology laboratory testing has relied on manual methods, and automated methods (such as those that have revolutionized clinical chemistry and hematology over the past several decades) were largely absent from the clinical microbiology laboratory. However, an increased demand for microbiology testing and standardization of sample-collection devices for microbiology culture, as well as a dwindling supply of microbiology technologists, has driven the adoption of automated methods for culture-based laboratory testing in clinical microbiology.

CONTENT

We describe systems currently enabling total laboratory automation (TLA) for culture-based microbiology testing. We describe the general components of a microbiology automation system and the various functions of these instruments. We then introduce the 2 most widely used systems currently on the market: Becton Dickinson's Kiestra TLA and Copan's WASPLab. We discuss the impact of TLA on metrics such as turnaround time and recovery of microorganisms, providing a review of the current literature and perspectives from laboratory directors, managers, and technical staff. Finally, we provide an outlook for future advances in TLA for microbiology with a focus on artificial intelligence for automated culture interpretation.

SUMMARY

TLA is playing an increasingly important role in clinical microbiology. Although challenges remain, TLA has great potential to affect laboratory efficiency, turnaround time, and the overall quality of culture-based microbiology testing.

Laboratory Automation in Clinical Microbiology

Laboratory automation is currently the main organizational challenge for microbiologists. Automating classic workflows is a strenuous process for the laboratory personnel and a huge and long-lasting financial investment. The investments are rewarded through increases in quality and shortened time to report. However, the benefits for an individual laboratory can only be estimated after the implementation and depending on the classic workflows currently performed. The two main components of automation are hardware and workflow. This review focusses on the workflow aspects of automation and describes some of the main developments during recent years. Additionally, it tries to define some terms which are related to automation and specifies some developments which would further improve automated systems.

Significant increase in cultivation of Gardnerella vaginalis, Alloscardovia omnicolens, Actinotignum schaalii, and Actinomyces spp. in urine samples with total laboratory automation

While total laboratory automation (TLA) is well established in laboratory medicine, only a few microbiological laboratories are using TLA systems. Especially in terms of speed and accuracy, working with TLA is expected to be superior to conventional microbiology. We compared in total 35,564 microbiological urine cultures with and without incubation and processing with BD Kiestra TLA for a 6-month period each retrospectively. Sixteen thousand three hundred thirty-eight urine samples were analyzed in the pre-TLA period and 19,226 with TLA. Sixty-two percent (n = 10,101/16338) of the cultures processed without TLA and 68% (n = 13,102/19226) of the cultures processed with TLA showed growth. There were significantly more samples with two or more species per sample and with low numbers of colony forming units (CFU) after incubation with TLA. Regarding the type of bacteria, there were comparable amounts of Enterobacteriaceae in the samples, slightly less non-fermenting Gram-negative bacteria, but significantly more Gram-positive cocci, and Gram-positive rods. Especially Alloscardivia omnicolens, Gardnerella vaginalis, Actinomyces spp., and Actinotignum schaalii were significantly more abundant in the samples incubated and processed with TLA. The time to report was significantly lower in the TLA processed samples by 1.5 h. We provide the first report in Europe of a large number of urine samples processed with TLA. TLA showed enhanced growth of non-classical and rarely cultured bacteria from urine samples. Our findings suggest that previously underestimated bacteria may be relevant pathogens for urinary tract infections. Further studies are needed to confirm our findings.

Microbes are off the menu: Defective macrophage phagocytosis in COPD

In this issue of the Biomedical Journal, we learn about the pathophysiology of chronic obstructive pulmonary disease and how defective macrophage phagocytosis may lead to the build up of microbes and pollutants in inflamed lungs. We also focus on new findings that may take us a step closer to full automation in diagnostic bacteriology laboratories. Finally, we highlight the anti-tumor properties of microalgae and the application of algorithms to predict human emotion from electrocardiogram.