Identification of Aspergillus at section and species levels by artificial intelligence-based microscopic morphology image recognition
Clinical Snapshot
PICO Framework
| P — Population | Clinical isolates of Aspergillus species from 12 species across 8 sections, represented by 11,689 high-resolution microscopic images |
| I — Intervention | FungalNet deep learning model (ResNet-50 with Focal Loss algorithm) for microscopic image analysis |
| C — Comparator | Three established deep learning models (GoogLeNet, ResNet-50, and Xception) and traditional morphological identification methods |
| O — Outcomes | Classification accuracy at section level (98.45%) and species level (97.85%) for Aspergillus identification |
Bottom Line
This diagnostic accuracy study evaluated FungalNet, a deep learning model for Aspergillus species identification from microscopic images. Using 11,689 high-quality images from clinical isolates, the model achieved 98.45% accuracy at section level and 97.85% at species level, outperforming established models. However, the study has significant limitations including single-center design, unclear reference standards, and lack of detailed diagnostic performance metrics beyond overall accuracy. While the technology shows promise for enhancing fungal identification in clinical laboratories, the results require validation across multiple centers with diverse patient populations and standardized reference methods. The clinical utility is potentially high given the importance of accurate Aspergillus identification for antifungal therapy selection, but implementation would require significant infrastructure investment and validation studies before routine clinical adoption.
Key Findings
P Value: Not reported
Effect Size: 98.45% accuracy at section level, 97.85% at species level
Primary Outcome: Classification accuracy for Aspergillus identification
Nnt Or Sensitivity: Sensitivity and specificity not reported - only overall accuracy provided
Confidence Interval: Not reported
Clinical Application
Requires high-resolution microscopy (100× oil immersion), standardized staining protocols, and computational infrastructure for deep learning model implementation Could enhance diagnostic capabilities in Australian pathology laboratories, potentially supporting TGA-approved antifungal stewardship programs and improving patient outcomes in tertiary hospitals with invasive aspergillosis cases Clinical microbiology laboratories performing Aspergillus identification from clinical specimens
Abstract
Rapid and accurate identification of Aspergillus species in clinical microbiology laboratories is crucial for aspergillosis diagnosis and antifungal therapy. However, traditional methods still face challenges in distinguishing phylogenetically related species due to their morphological similarities. This study presents FungalNet, a deep learning model integrating ResNet-50 architecture with Focal Loss algorithm, specifically designed to enhance feature extraction for Aspergillus identification. A total of 12,000 high-resolution images were obtained from lactophenol cotton blue-stained slide preparations under a 100× oil immersion objective, among which 311 images were excluded through a novel quality control approach combining fivefold cross-validation and expert manual review. The performance of four deep learning models (FungalNet and three established models) for identifying Aspergillus species and sections was evaluated using the remaining 11,689 qualified images. FungalNet demonstrated superior classification performance, achieving overall accuracies of 98.45% and 97.85% at the section and species levels, respectively. These results indicate that FungalNet shows significant promise for rapid and accurate identification of Aspergillus species. With further optimization and multicenter validation, this tool could potentially be integrated into routine diagnostic workflows to enhance the efficiency and reliability of fungal identification in clinical settings.IMPORTANCEThis study integrates microscopic morphology identification with deep learning to address the challenge of accurate Aspergillus species identification. Twelve clinically isolated Aspergillus species belonging to eight different sections were included. From touch-tape slide preparations with lactophenol cotton blue staining under a 100× oil immersion objective, 11,689 qualified images were collected and analyzed using FungalNet (our proposed model) along with three established models (GoogLeNet, ResNet-50, and Xception). The results showed that FungalNet demonstrated superior performance in Aspergillus identification, achieving the highest classification accuracy at both section (98.45%) and species (97.85%) levels. Given its rapid turnaround time and cost-effectiveness, this AI-based image analysis approach shows promising potential for the rapid and accurate identification of Aspergillus species in clinical microbiology laboratories.
References
- 1.Tan, M., Guo, Z., Wang, Y., Xu, X., Cao, W., Liu, Z., & Jiang, C. (2026). Identification of Aspergillus at section and species levels by artificial intelligence-based microscopic morphology image recognition. Journal of Clinical Microbiology. https://doi.org/10.1128/jcm.00012-26
Related Research
Clinical infectious diseases : an official publication of the Infectious Diseases Society of America
Utility of Repeating Blood Cultures in Candidemia: Insights From 2 Large Clinical Data Repositories
24 July 2026
mSystems
MeLSI: Metric Learning for Statistical Inference in microbiome community composition analysis
22 July 2026
Archives of microbiology
SigMine and OPathDb: a literature-mining pipeline and database of potential opportunistic pathogens
13 July 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service