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

Abstract

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 ¼ 71, 73%) but a significant number used data from low-and middle-income countries (n ¼ 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.

Performance and impact of a multiplex PCR in ICU patients with ventilator-associated pneumonia or ventilated hospital-acquired pneumonia

“…The turnaround time of the Unyvero system is longer than the only other M-PCR system currently available for respiratory samples, the BioFire FilmArray (BioMerieux), that allows a diagnosis in 65 min instead of 4-5 h. Moreover, owing to the development of new techniques, M-PCR might become outdated even before they are widely used. Indeed, clinical metagenomics, the comprehensive sequencing of microbial and host genetic material in clinical samples [32], has the potential to improve microbiological diagnostics [33]. In a recent proof of concept study, Langelier et al combined microbiological and host transcriptome data in tracheal aspirates of 26 patients with a lower respiratory tract infection in the ICU and identified the causative pathogens with an AUC of 0.91 (95% CI, 0.83-0.97); however, this kind of techniques is expensive and not routinely available.…”

“…Special attention should be paid to the integration and implementation of systems into clinical practice and their adoption and utilization by clinicians. While waiting for the results of clinical trials, implementation outcomes such as appropriateness or fidelity may be key intermediate outcomes to study the success of strategies aiming to bring M-PCR systems to the clinical practice [33]. The question of whether or not M-PCR reduces antibiotic use, antibiotic resistance, direct costs, and indirect costs should be a priority area for future research.…”

Considerations for diagnostic COVID-19 tests

“…Some mature AI solutions are ready for application to support patient care 132 or clinical decision making, for example by reducing antibiotic use 133 (Box 3). More tools that improve the use of clinical and epidemiological big data will become increasingly available 134,135 and are currently being developed by laboratory scientists together with data scientists and software developers. New biomarkers are not only crucial for patient management by facilitating early diagnosis of severe COVID19, they are also important in the development of a COVID19 vaccine, as they can accelerate clinical trials, reduce costs, guide participant selection, reduce patient safety risks and enable easier verification of the mechanism of action.…”

Performance and impact of a multiplex PCR in ICU patients with ventilator-associated pneumonia or ventilated hospital-acquired pneumonia

“…Indeed, clinical metagenomics, the comprehensive sequencing of microbial and host genetic material in clinical samples, 33 has the potential to improve microbiological diagnostics. 34 In a recent proof of concept study, Langelier et al combined microbiological and host transcriptome data in tracheal aspirates of 26 patients with a lower respiratory tract infection in the ICU and identi ed the causative pathogens with an AUC of 0.91 (95% CI, 0.83-0.97) however this kind of techniques is expensive and not routinely available.…”

“…While waiting for the results of clinical trials, implementation outcomes such as appropriateness or delity may be key intermediate outcomes to study the success of strategies aiming to bring M-PCR systems to the clinical practice. 34 The question of whether or not M-PCR reduces antibiotic use, antibiotic resistance, direct costs, and indirect costs should be a priority area for future research. In our study, each episode of VAP or ventilated HAP was analyzed by a multidisciplinary expert panel of senior intensivists and clinical microbiologists.…”