Related Experiment Video
Updated: Sep 13, 2025
09:37
Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition
Published on: August 18, 2022
2.5K
Deep learning-based classification of fungal and Acanthamoeba keratitis using confocal microscopy
Rohith Erukulla1, Kosar Esmaili2, Amir Rahdar3
1University of Illinois, College of Medicine, Chicago, IL, USA.
The Ocular Surface
|August 2, 2025
Summary
Deep learning models accurately classify fungal keratitis (FK) and Acanthamoeba keratitis (AK) from eye images. This AI approach aids in diagnosing infectious keratitis (IK) and subtyping FK for better treatment.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Fungal keratitis (FK) and Acanthamoeba keratitis (AK) present significant diagnostic and therapeutic challenges, often leading to poor visual outcomes.
- Infectious keratitis (IK) encompasses a range of microbial causes, necessitating accurate and timely identification for effective management.
Purpose of the Study:
- To evaluate the efficacy of deep learning (DL) models in classifying different types of infectious keratitis (IK).
- To assess the capability of DL in subtyping fungal keratitis (FK) into filamentous and non-filamentous forms.
- To determine the feasibility of using DL with in vivo confocal microscopy for IK diagnosis.
Main Methods:
- A dataset of 1975 images from Heidelberg Retinal Tomograph 3 (HRT 3) was utilized, including fungal keratitis (FK), Acanthamoeba keratitis (AK), and nonspecific keratitis (NSK).
- Transfer learning with a ResNet50 architecture was employed for classification and subtyping tasks.
- Data augmentation and class weighting were applied to address class imbalance, with models trained using Adam optimizer and 5-fold cross-validation.
Main Results:
- Model 1 achieved high weighted average accuracy (89%) for classifying IK types (FK, AK, NSK), with strong precision and recall for AK (93%, 96%) and FK (90%, 92%).
- Model 2 accurately subtyped FK with 85% accuracy, demonstrating high performance in distinguishing filamentous (F1-score 0.81) and non-filamentous (F1-score 0.85) forms.
- The models achieved high ROC AUC (0.94) and PR AUC (0.95) for FK subtyping, indicating robust diagnostic potential.
Conclusions:
- Deep learning models demonstrate significant potential for accurate classification of infectious keratitis (IK) subtypes.
- AI-driven analysis of confocal microscopy images can enhance diagnostic accuracy for FK and AK.
- These findings support the integration of DL into clinical practice for improved IK diagnosis and targeted treatment strategies.
Related Concept Videos
01:16Confocal Fluorescence Microscopy
14.3K
Confocal microscopy is an advanced microscopic technique. The prime advantage of the confocal microscope over other microscopy techniques is its ability to block the out-of-focus light from the illuminated samples using pinholes. It is widely used with fluorescence optics to obtain high-resolution, sharp contrast images. Unlike optical microscopes, confocal microscopes use a focused beam of light laser to scan the entire sample surface at different z-planes. These microscopes are, therefore,...
14.3K
01:28Three-Dimensional Microscopy in Microbiology
295
Three-dimensional imaging techniques are essential in cell biology, allowing researchers to visualize intricate cellular structures with high resolution. Two prominent methods, Differential Interference Contrast Microscopy (DIC) and Confocal Scanning Laser Microscopy (CSLM), provide distinct advantages for imaging live and thick specimens, respectively.Differential Interference Contrast MicroscopyDIC microscopy enhances contrast in transparent, unstained samples by converting phase...
295