CsDETC: detection and counting of small target Cryptococcus spp

Area of Science:

• Medical Diagnostics
• Computer Vision
• Infectious Diseases

Background:

• Cryptococcus spp. infections cause serious conditions like cryptococcal meningitis (CM).
• Current diagnostic methods (CSF ink staining, culture) are manual, inefficient, and lack accuracy.
• Accurate and timely diagnosis of Cryptococcus spp. is crucial for effective clinical management.

Purpose of the Study:

• To develop an automated method for detecting and counting Cryptococcus spp. in cerebrospinal fluid (CSF).
• To improve the accuracy and efficiency of diagnosing Cryptococcus spp. infections.

Main Methods:

• Proposed CsDETC method integrating data augmentation, hypergraph computation (HGC-SCS), and attention-enhanced path aggregation network (AEPAN).
• Extensive data augmentation (cropping, flipping, rotation, perspective transformation) enhanced model generalization.
• Integrated CBAM into PAN for improved small target perception through channel and spatial attention.

Main Results:

• CsDETC demonstrated superior performance over advanced models like YOLOv10 and YOLO11 on a private dataset.
• Achieved significant improvements in mAP50 (93.6% vs. 91.3%), APs (51.0% vs. 49.5%), and MAE (1.865 vs. 2.370).
• Inference time increased minimally (0.7 ms).

Conclusions:

• CsDETC shows promise as a tool for accurate and timely identification and automated counting of Cryptococcus spp.
• Further validation with larger, diverse datasets is needed.
• Potential to aid clinicians in treatment planning and enhance diagnostic efficiency for cryptococcosis.