MorphoPath-VL: Morphology-Guided Vision–Language Learning for Waterborne Pathogen Detection in Microscopy Images
Contributors
Jenefa.A
Hemalatha S
Keywords
Proceeding
Track
General Track
License
Copyright (c) 2026 Sustainable Global Societies Initiative

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Abstract
In microscopy images, automated detection of waterborne pathogens can facilitate regular analysis of microscopic samples but image detectors or Vision-Language models that are only guided by class names do not take into account the combined shape and structural description explored in this study. MorphoPath-VL introduces morphology-guided vision-language learning to associate image content with pathogen names and complementary descriptions of shape and structure during detection. The method was tested on a dataset of microscopy images of waterborne pathogens, and compared to the conventional detectors, image-only matched backbone and vision-language class-name baseline. MorphoPath-VL achieved a precision of 92.38%, recall of 91.06%, and mean average precision (mAP) of 93.24% at an intersection over union (IoU) threshold of 0.5 and 68.45% for IoU thresholds between 0.5 and 0.95. These scores represent an improvement of 3.07, 3.41, 2.98 and 4.61 percentage points over Deformable Detection Transformer (Deformable DETR), respectively. Ablation analysis also showed that using morphological descriptions alone resulted in mAP@0.5:0.95 of 64.17%, while combining the two gave 68.45%. These quantitative gaps reflect the effectiveness of joint shape and structural guidance for pathogen detection and localization of bounding-boxes in the tested data set.