Table 3.
The characteristics of the studies included regarding the application of machine learning to image analysis in clinical microbiology laboratory.
| Authors (Year) |
n | Diagnosis Method | Input | Model/Analysis | Objective |
|---|---|---|---|---|---|
| Loh et al. [46] (2021) |
297 | Blood smear microscopy | Microscopic smear images | Mask R-CNN | Alternative method for automated rapid malaria screening |
| Holmström et al. [47] (2020) |
125 | Thin blood and Giemsa-stained thick smear microscopy | Microscopic smear images | Cloud-based machine-learning platform (Aiforia Cloud and Create), GoogLeNet network | Digitalization of blood smears, application of deep learning (DL) algorithms to detect Plasmodium falciparum |
| Oliveira et al. [48] (2022) |
676 | Thick blood smear films | Microscopy images | Multilayer perceptron (MLP) and decision tree (DT) | Automated malaria diagnosis |
| Sengar et al. [49] (2022) |
2329 | Thin blood smears | Microscopic images | Generative adversarial network (GAN), Vision Transformers (ViTs) | Automated, non-invasive multi-class Plasmodium vivax life cycle classification and malaria diagnosis |
| Park et al. [50] (2016) |
413 | Quantitative phase spectroscopy | Quantitative phase images of unstained cells | Linear discriminant classification (LDC), logistic regression (LR), and k-nearest neighbor classification (KNC), | Automated analysis for detection and staging of red blood cells infected with Plasmodium falciparum at trophozoite or schizont stage |
| Kassim et al. [51] (2021) |
5972 | Thick smear films | Annotated thick smear microscopy images | Mask regional–convolutional neural network (Mask R-CNN), ResNet50 classifier | Application of PlasmodiumVF-Net for automated malaria diagnosis on both image and patient level |
| Dey et al. [52] (2021) |
27,558 | Thick blood films | Blood smear cell microscopy images | ResNet 152 model integrated with the deep greedy network | Automating the detection of malaria parasites in thin blood smear images |
| Ufuktepe et al. [53] (2021) |
955 | Thin blood smears | Thin blood smear microscopy | Channel-wise feature pyramid network for medicine (CFPNet-M) | Red blood cell detection, counting infected cells or identifying parasite species |
| Hemachandran et al. [54] (2023) |
27,558 | Blood smears | Blood smear microscopy images | CNN, MobileNetV2, and ResNet50 | Automatic image identification system for parasite-infected RBC detection |
| Holmström et al. [55] (2017) |
7385 | Iodine-stained stool sample smears | Digital images from a mobile microscope and whole slide-scanner | Sequential algorithms | Automated detection of soil-transmitted helminths and Schistosoma haematobium |
| Kuok et al. [56] (2019) |
19,234 | Sputum smears stained by acid-fast staining | Smear microscopy images | Refined Faster region-based CNN, support vector machine (SVM) | Two-stage Mycobacterium tuberculosis identification system |
| Yang et al. [57] (2020) |
167 | Ziehl–Neelsen stained human tissue samples | Digitized images | CNNIN, CNNAL | Automated identification of mycobacteria in human tissues |
| Ibrahim et al. [58] (2021) |
1050 | Acid-fast staining of sputum | Microscopy images | AlexNet model | Automated detection of Mycobacterium tuberculosis using transfer learning |
| Xiong et al. [59] (2018) |
3,088,492 | Acid-fast stained tissue samples | Microscopy images | CIFAR-10 CNN | AI-assisted detection method for acid-fast stained TB bacillus |
| Horvath et al. [60] (2020) |
15,204 | Auramine-stained sputum smears | Slide microscopy images | DNN classifier, Keras, TensorFlow | Machine-assisted interpretation of auramine stains for microscopic tuberculosis diagnosis |
| Smith et al. [61] (2018) |
25,488 | Gram staining of blood cultures | Microscopy images | Inception v3 CNN, Python, TensorFlow | Automated interpretation of blood culture gram stains |
| Hoorali et al. [62] (2020) |
954 | Tissue slides of patients suffering from cutaneous anthrax | Microscopy images | UNet and UNet++, Keras, TensorFlow | Automatic and rapid diagnosis of anthrax via detection and segmentation of Bacillus anthracis |
| Kang et al. [63] (2020) |
84,000 | Hyperspectral microscope imaging (HMI) method | Hyperspectral microscope images | Linear discriminant analysis (LDA), support vector machine (SVM) and soft-max regression (SR) |
Identification of non-O157 Shiga toxin-producing Escherichia coli (STEC) using deep learning |
| Oyamada et al. [64] (2021) |
910 | Microscopic Agglutination Test (MAT) | MAT microscopic images | Support vector machine (SVM) | Determine agglutination within microscopic images for the diagnosis of leptospirosis |
| Zieliński et al. [65] (2017) |
660 | Stained clinical samples | DIBas dataset of digital bacterial images | CNN, support vector machine, random forest | Deep learning-based classification of bacterial genera and species |
| Ahmad et al. [66] (2023) |
480 | Stained clinical samples | High-resolution microscopic images from DIBas dataset | InceptionV3, MobileNetV2 | Deep ensemble approach-based pathogen classification in large-scale images |
| Van et al. [67] (2019] |
480 | Clinical throat specimens on CHROMagar confirmed by MALDI-TOF MS | Microscopic images | WASPLab PhenoMATRIX chromogenic detection module | AI-detection of Streptococcus pyogenes using CHROMagar |
| Gammel et al. [68] (2021) |
5913 | Patient samples collected from the nares plated onto BD BBL CHROMagar MRSA II and BD BBL CHROMagar Staph aureus | Digital images | Automated Plate Assessment System (APAS Independence) | Evaluation of an automated plate assessment system |
| Rattray et al. [69] (2023) |
335 | Culture specimens of clinical and environmental P. aeruginosa isolates | Digital colony images | ResNet-50, VGG-19, MobileNetV2 and Xception | Identification of from colony image data |
| Zhang et al. [70] (2022) |
960 | Escherichia coli cultures on agar medium | Digital colony images | Random cover targets algorithm (RCTA), YOLOv3 | Deep learning-based bacterial colony detection |
| Koo et al. [71] (2021) |
3707 | Slides with skin and nail specimens | Microscopy images | YOLO v4 | Automated detection of superficial fungal infections |
| Ma et al. [72] (2021) |
17,142 | Dissecting microscopy (DM)/stereomicroscopy platform | Original colony images | Xception | Validating a novel approach for the detection of Aspergillus fungi via stereomicroscopy |
| Liu et al. [73] (2015) |
1000 | Fecal specimens | Microscopic fecal images | ANN-1, ANN-2 | Automatic identification of fungi in fecal specimens |
| Meeda et al. [74] (2019) |
30 | Fungal cultures, confocal microscopy | Colony fingerprint digital images | Support vector machine (SVM) and random forest (RF) |
Rapid discrimination of fungal species by the colony fingerprinting |
| Khan et al. [75] (2018) |
119 | Raman spectroscopy | Spectral images | Support vector machine (SVM) | Analysis of hepatitis B virus infection in blood sera using ML |
| Rohaim et al. [76] (2020) |
199 | Reverse-transcribed loop-mediated isothermal amplification (LAMP) assay | Quantitative measurements using qRT-PCR | CNN model with binary cross-entropy and Adam | Rapid detection of SARS-CoV-2 using AI in loop-mediated isothermal amplification assays |
| Ito et al. [77] (2018) |
35 | Transmission electron microscopy (TEM) | Microscopy images | Cross-point method (CPM), RDP, spectral rings (SR), fully convolutional neural networks (FCN and FCN+) | Automated feline calicivirus particle detection in TEM images |
| Tong et al. [78] (2019) |
600 | Raman spectroscopy of serum samples | Raman spectra | Principal component analysis (PCA), support vector machine (SVM) | AI-aided detection of hepatitis B virus infection using Raman spectroscopy |
| Tabarov et al. [79] (2022) |
90 | Surface-enhanced Raman scattering spectroscopy (SERS) | SERS spectra | Support vector machine (SVM) | Detection of A and B influenza viruses by SERS coupled with ML |