Tıbbi Mikrobiyolojide Yapay Zeka Uygulamaları

Giriş

The increasing volume and complexity of data generated in modern medical microbiology laboratories have driven the growing adoption of artificial intelligence (AI) and machine learning (ML) as advanced analytical tools capable of transforming diagnostic workflows and clinical decision- making (1,2). Traditional microbiological methods, including microscopy, culture-based identification, and antimicrobial susceptibility testing (AST), remain the cornerstone of clinical diagnostics; however, they present limitations such as prolonged turnaround times, workflow variability, and delayed detection of antimicrobial resistance (AMR), which may contribute to suboptimal antimicrobial use and increased resistance emergence (2).

Recent developments in AI offer complementary approaches that can rapidly analyze large- scale, heterogeneous microbiological datasets, encompassing imaging, genomic, spectral, and clinical data. Machine learning models learn patterns from labeled datasets by identifying relationships between input features (e.g., spectral peaks, genomic variants, image structures) and predefined outputs such as microbial species or resistance phenotypes. These models are trained on historical data and evaluated using validation datasets to estimate performance and generalizability, rather than replacing expert interpretation. AI functions as a decision-support framework that enhances diagnostic accuracy, reproducibility, laboratory efficiency, and antimicrobial stewardship when appropriately validated and integrated into clinical workflows (3).

As AI technologies continue to evolve, their applications in medical microbiology span multiple stages of the diagnostic pathway. In this section, these applications can be grouped into four domains: microbial identification and classification, antimicrobial resistance detection and prediction, virulence assessment and pathogen characterization, and antimicrobial stewardship and clinical decision support.

1. Microbial Identification and Classification

1.1 Image-Based Microbial Classification

Deep learning models, particularly convolutional neural networks (CNNs), have shown promising performance in image-based bacterial classification using microscopy images and culture plates. CNNs automatically learn hierarchical image features through convolutional layers, enabling classification without manual feature engineering. High-performance models such as DenseNet-121 have achieved accuracy levels exceeding 99% across datasets comprising multiple bacterial species. Notably, such performance was achieved using relatively modest datasets, demonstrating the effectiveness of transfer learning approaches in microbiology (4). Automated image interpretation workflows may significantly reduce manual review time, particularly in microscopy-based diagnostics such as mycobacterial or malaria detection, thereby improving laboratory efficiency and turnaround times.

1.2 MALDI-TOF Mass Spectrometry–Based Identification

The integration of ML with MALDI-TOF mass spectrometry represents one of the most advanced applications in microbial diagnostics. Beyond conventional species identification, Algorithms including support vector machines, random forests, and neural networks enable enhanced spectral pattern recognition beyond traditional database matching. A systematic review reported that 27 of 36 studies (75%) focusing on MALDI-TOF-based applications targeted species identification and antimicrobial susceptibility prediction, highlighting the central role of this technology in AI-driven microbiology workflows (5). These approaches may support earlier clinical decision-making by accelerating microbial characterization when appropriately validated.

1.3 Genomics-Driven Identification

Machine learning models applied to genomic sequencing data enable high-resolution pathogen identification and classification. Integration of ML with DNA sequencing has demonstrated strong diagnostic performance, with reported F1-scores up to 0.88 and AUC values up to 0.96. (1,6). This approach allows identification of complex microbial signatures and may enhance the detection of difficult-to-classify organisms.

2. Antimicrobial Resistance Detection and Prediction

Antimicrobial resistance prediction represents one of the most extensively studied applications of machine learning in medical microbiology, reflecting the urgent global need for rapid resistance detection strategies (7–9). ML models trained on genomic sequencing data, MALDI-TOF spectra, and integrated clinical datasets have demonstrated the ability to identify resistance determinants and predict antimicrobial susceptibility profiles with promising accuracy (1,5).

Compared with traditional risk-scoring approaches, machine learning models have shown improved specificity for resistance prediction while enabling earlier therapeutic optimization. These models may complement phenotypic AST by providing preliminary resistance predictions before laboratory confirmation, potentially reducing turnaround times and improving antimicrobial stewardship outcomes when adequately validated (4,7).

3. Virulence Assessment and Pathogen Characterization

Virulence assessment is an emerging application of machine learning in medical microbiology, focusing on identifying pathogen traits linked to disease severity and clinical outcomes. Integration with genomic technologies such as whole-genome sequencing and metagenomics enables detection of virulence-associated patterns and may support early risk stratification and targeted therapeutic strategies, especially in vulnerable populations. However, compared with other AI applications, virulence prediction remains less mature due to limited external validation, data heterogeneity, computational demands, and challenges in model interpretability, highlighting the need for further standardization and clinical validation (10).

4. Antimicrobial Stewardship and Clinical Decision Support

Antimicrobial stewardship emerged as a distinct application leveraging electronic health record data for clinical decision support. This application addresses the critical need to promote appropriate antibiotic use and reduce resistance through improved clinical workflows (11).

The clinical impact of ML-based stewardship tools has been quantitatively demonstrated. Implementation led to a 67% reduction in fluoroquinolone usage and an 18% reduction in inappropriate therapy. These tools support scaling up antimicrobial stewardship programs with personalized recommendations, representing a shift from traditional rule-based approaches to data- driven clinical decision support systems (9).

5. Clinical Impact and Implementation Challenges

Enhanced microbial identification using AI provides important clinical benefits, including improved diagnostic accuracy, increased workflow efficiency, reduced operational costs, and earlier targeted therapy selection (8). Improved pathogen identification also strengthens infection control and antimicrobial stewardship efforts. However, despite promising performance, clinical implementation remains limited, with external validation reported in only a small proportion of studies (≈11.11% in MALDI-TOF–based systematic reviews), raising concerns about generalizability across laboratories (5). Performance may decline due to dataset shift, class imbalance, or limited representation of rare pathogens, potentially affecting clinical reliability. Integration of machine learning with MALDI-TOF mass spectrometry extends beyond species identification by enabling advanced spectral pattern recognition and antimicrobial resistance prediction, supporting faster microbial characterization and earlier clinical decision-making when properly validated and incorporated into routine workflows (5).

Future research should prioritize standardized evaluation frameworks, multicenter external validation, and improved model interpretability to facilitate safe and scalable clinical implementation. As AI systems become increasingly integrated with laboratory and clinical workflows, they have the potential to transform medical microbiology into a more predictive, data- driven discipline supporting precision diagnostics and optimized antimicrobial stewardship. Human- in-the-loop designs, uncertainty reporting, and clinician override mechanisms are essential to ensure patient safety and maintain clinical trust.

Although most evidence on AI applications in medical microbiology derives from high- resource settings, implementation feasibility varies across national contexts depending on laboratory infrastructure, antimicrobial resistance (AMR) epidemiology, and regulatory frameworks. Countries with established AMR surveillance systems and centralized laboratory networks may facilitate integration of AI-driven diagnostics and stewardship tools. However, differences in digital maturity, data interoperability, and availability of annotated datasets remain key barriers, highlighting the need for context-specific validation studies and real-world implementation research (8,9).

6. Scope and Limitations

This chapter focuses on current diagnostic and clinical applications of AI in medical microbiology. Ethical considerations, regulatory frameworks, data privacy governance, and cost- effectiveness analyses are acknowledged as critical factors but are beyond the primary scope of this discussion and require dedicated evaluation in future work.

Kaynakça

1. Asnicar F, Thomas AM, Passerini A, Waldron L, Segata N. Machine learning for microbiologists. Nat Rev Microbiol. 2024 Apr;22(4):191–205.
2. Alsulimani A, Akhter N, Jameela F, Ashgar RI, Jawed A, Hassani MA, et al. The Impact of Artificial Intelligence on Microbial Diagnosis. Microorganisms. 2024 Jun;12(6):1051.
3. Wu Y, Gadsden SA. Machine learning algorithms in microbial classification: a comparative analysis. Front Artif Intell [Internet]. 2023 Oct 19 [cited 2026 Feb 16];6. Available from: https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2023.1200994/full
4. Burns BL, Rhoads DD, Misra A. The Use of Machine Learning for Image Analysis Artificial Intelligence in Clinical Microbiology. Journal of Clinical Microbiology. 2023 Jul 3;61(9):e02336-21.
5. Weis CV, Jutzeler CR, Borgwardt K. Machine learning for microbial identification and antimicrobial susceptibility testing on MALDI-TOF mass spectra: a systematic review. Clinical Microbiology and Infection. 2020 Oct 1;26(10):1310–7.
6. Xu C, Zhao LY, Ye CS, Xu KC, Xu KY. The application of machine learning in clinical microbiology and infectious diseases. Front Cell Infect Microbiol [Internet]. 2025 May 1 [cited 2026 Feb 16];15. Available from: https://www.frontiersin.org/journals/cellular-and-infection-microbiology/articles/10.3389/fcimb.2025.1545646/full
7. Elyan E, Hussain A, Sheikh A, Elmanama AA, Vuttipittayamongkol P, Hijazi K. Antimicrobial Resistance and Machine Learning: Challenges and Opportunities. IEEE Access. 2022;10:31561–77.
8. Tang R, Luo R, Tang S, Song H, Chen X. Machine learning in predicting antimicrobial resistance: a systematic review and meta-analysis. International Journal of Antimicrobial Agents. 2022 Nov 1;60(5):106684.
9. Anahtar MN, Yang JH, Kanjilal S. Applications of Machine Learning to the Problem of Antimicrobial Resistance: an Emerging Model for Translational Research. Journal of Clinical Microbiology. 2021 Jun 18;59(7):10.1128/jcm.01260-20.
10. Abhadionmhen AO, Asogwa CN, Ezema ME, Nzeh RC, Ezeora NJ, Abhadiomhen SE, et al. Machine Learning Approaches for Microorganism Identification, Virulence Assessment, and Antimicrobial Susceptibility Evaluation Using DNA Sequencing Methods: A Systematic Review. Mol Biotechnol. 2025 Nov 1;67(11):4038–66.
11. Pennisi F, Pinto A, Ricciardi GE, Signorelli C, Gianfredi V. The Role of Artificial Intelligence and Machine Learning Models in Antimicrobial Stewardship in Public Health: A Narrative Review. Antibiotics. 2025 Feb;14(2):134.