Deep Learning-Based Bacterial Image Classification for Medical Diagnosis

Abstract—The rapid and accurate identification of bacteria is crucial to effective clinical diagnosis, infection control, andantibiotic stewardship, but conventional microscopic examination is time-consuming, reliant on experts, and subject to interobserver variation. Existing works mostly use either convolutional or transformer-based backbones for this task and few works include multi-scale convolutional features along with global attention, and few works provide explanations of their prediction. This paper presents a hybrid deep learning framework for the classification of microscopic images of bacteria that combines multi-scale residual/dense convolutional features, a channel-andspatial attention module, and integrates Gradient-weighted Class Activation Mapping (Grad-CAM) to provide visual explanations that are correlated with clinically relevant bacterial morphology. The framework is built on an open-source microbial image database of bacteria, and the preprocessing, augmentation, and train/validation/test procedure that is described is appropriate for class-imbalanced data, which is common in such databases. We describe the planned experimental setup, evaluation metrics (accuracy, precision, recall, F1-score, ROC-AUC, and confusion matrix), and compare with recent baselines based on CNN and Vision Transformer. Reported result values in this manuscript are simply placeholders for reporting structure and should be replaced with measured result values upon performing the experiments described. The analysis still suggests the intriguing potential for attention-augmented multi-scale fusion to enhance the discrimination of morphologically similar bacterial species, and the ability of Grad-CAM to provide clinically reliable and interpretable decision support. In this contribution, we will introduce a novel fusion-attention architecture, an interpretability-integrated evaluation protocol, and a critical comparison with the recent CNN/ViT literature.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-18
DOI
https://doi.org/10.5281/zenodo.24331981
Primary Topic
Cell Image Analysis Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Deep Learning-Based Bacterial Image Classification for Medical Diagnosis

Ekta, Mukesh Singla
1 links
Zenodo (CERN European Organization for Nuclear Research)
Cell Image Analysis Techniques
article

Deep Learning-Based Bacterial Image Classification for Medical Diagnosis

Ekta, Mukesh Singla
article en

Abstract

Abstract—The rapid and accurate identification of bacteria is crucial to effective clinical diagnosis, infection control, andantibiotic stewardship, but conventional microscopic examination is time-consuming, reliant on experts, and subject to interobserver variation. Existing works mostly use either convolutional or transformer-based backbones for this task and few works include multi-scale convolutional features along with global attention, and few works provide explanations of their prediction. This paper presents a hybrid deep learning framework for the classification of microscopic images of bacteria that combines multi-scale residual/dense convolutional features, a channel-andspatial attention module, and integrates Gradient-weighted Class Activation Mapping (Grad-CAM) to provide visual explanations that are correlated with clinically relevant bacterial morphology. The framework is built on an open-source microbial image database of bacteria, and the preprocessing, augmentation, and train/validation/test procedure that is described is appropriate for class-imbalanced data, which is common in such databases. We describe the planned experimental setup, evaluation metrics (accuracy, precision, recall, F1-score, ROC-AUC, and confusion matrix), and compare with recent baselines based on CNN and Vision Transformer. Reported result values in this manuscript are simply placeholders for reporting structure and should be replaced with measured result values upon performing the experiments described. The analysis still suggests the intriguing potential for attention-augmented multi-scale fusion to enhance the discrimination of morphologically similar bacterial species, and the ability of Grad-CAM to provide clinically reliable and interpretable decision support. In this contribution, we will introduce a novel fusion-attention architecture, an interpretability-integrated evaluation protocol, and a critical comparison with the recent CNN/ViT literature.

Zenodo (CERN European Organization for Nuclear Research)
Baba Mastnath University (IN)
Cell Image Analysis Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon