LGH-Transformer: A Hybrid CNN–Transformer Model for Multi-Class Bladder Tissue Classification from Cystoscopic Images
4 affiliations
- 1Department of Urology, Faculty of Medicine, Çanakkale Onsekiz Mart University, 17100 Çanakkale, Turkey
- 2Department of Medical System Biology, Graduate School of Sciences, Çanakkale Onsekiz Mart University, 17100 Çanakkale, Turkey
- 3Department of Medical Biology, Faculty of Medicine, Çanakkale Onsekiz Mart University, 17100 Çanakkale, Turkey
- 4Department of Software Engineering, Faculty of Engineering, Bandirma Onyedi Eylül University, 10200 Balıkesir, Turkey
Abstract
Automated analysis of cystoscopic images may support consistent characterization of bladder tissue patterns, but reported performance is highly sensitive to patient composition, data partitioning, and evaluation methodology. Objective: This study provides an exploratory comparison of convolutional neural networks, Transformer-based models, and a proposed Local–Global Hybrid Transformer (LGH-Transformer) for four-class bladder tissue image classification. The dataset comprised 1754 frames and used the publicly released 1257/308/189 training–validation–held-out test split. The annotation-defined held-out test composition was HGC = 74, LGC = 53, NST = 37, and NTL = 25 (n = 189). Images were resized to 224 × 224 pixels, normalized, and augmented only in the training subset. Adam and RMSprop were evaluated. Accuracy and macro-averaged precision, recall, and F1-score were calculated from the integer-count evaluation matrices. The highest held-out accuracy, 187/189 (0.9894), was reached by seven model–optimizer configurations: DenseNet201–RMSprop, XCiT Medium 24 P16–Adam, ConvNeXt Base–Adam, CoAtNet–Adam, ConvNeXt Base–RMSprop, CoAtNet–RMSprop, and the proposed LGH-Transformer–RMSprop; MobileNet–Adam and Xcep-tion–RMSprop each yielded 185/189 (0.9788), and LGH-Transformer–Adam yielded 186/189 (0.9841). On a 189-frame test set, differences in one to two frames do not establish statistical superiority, so these configurations are reported as a top-accuracy tier rather than as a ranked superiority claim. Within this tier, the LGH-Transformer had the smallest computational footprint (≈18.2 M parameters and ≈4.1 GFLOPs, versus 20.0–88.6 M parameters and 4.3–16.2 GFLOPs for the other top-tier models). Component-wise ablation experiments showed accuracy decreases of 0.0106, 0.0053, and 0.0053 after removing the CNN tokenizer, local attention, and global attention, respectively. The LGH-Transformer reaches the top accuracy tier of this 18-architecture benchmark at the lowest parameter and FLOP budget within that tier, supporting an internally consistent, efficiency-oriented comparison within the available cohort. The absence of task-aligned external validation and the small number of held-out patients require the findings to be interpreted as exploratory internal validation rather than evidence of clinical generalizability.
Related
- Interpretable Deep Learning of Endogenous Fluorophore Concentration and Entropy for Lung Tissue Classification
- Piezo1 Activation by Yoda1 Inhibits Mechanosensitive ATP Release in Human Urothelial Cells and Bladder Overactivity in Rats with Bladder Outlet Obstruction
- From epidemiology to molecular pathology: current research status and future prospects on the role of toxic metal exposure in the development and progression of bladder cancer
- GlycoStamp-seq: Single-cell nascent sialoglycometabolomics captures sialylation dynamics in immune cells of an oxaliplatin-treated bladder cancer mouse model
- CBAT-Net: a unified CNN–BiLSTM–MultiHead transformer architecture for joint Q3/Q8 protein secondary structure prediction
BibTeX
@article{kurt2026lghtransformer,
title = {{LGH-Transformer: A Hybrid CNN–Transformer Model for Multi-Class Bladder Tissue Classification from Cystoscopic Images}},
author = {Hasan Anıl Kurt and Sabire Kılıçarslan and Meliha Merve Çiçekliyurt and Serhat Kiliçarslan},
journal = {International journal of molecular sciences},
year = {2026},
month = oct,
doi = {10.3390/ijms27198929},
url = {https://europepmc.org/article/PMC/PMC13666353},
note = {CC BY},
}
RIS
TY - JOUR
TI - LGH-Transformer: A Hybrid CNN–Transformer Model for Multi-Class Bladder Tissue Classification from Cystoscopic Images
AU - Hasan Anıl Kurt
AU - Sabire Kılıçarslan
AU - Meliha Merve Çiçekliyurt
AU - Serhat Kiliçarslan
PY - 2026
DA - 2026/10/01
JO - International journal of molecular sciences
DO - 10.3390/ijms27198929
UR - https://europepmc.org/article/PMC/PMC13666353
AB - Automated analysis of cystoscopic images may support consistent characterization of bladder tissue patterns, but reported performance is highly sensitive to patient composition, data partitioning, and evaluation methodology. Objective: This study provides an exploratory comparison of convolutional neural networks, Transformer-based models, and a proposed Local–Global Hybrid Transformer (LGH-Transformer) for four-class bladder tissue image classification. The dataset comprised 1754 frames and used the publicly released 1257/308/189 training–validation–held-out test split. The annotation-defined held-out test composition was HGC = 74, LGC = 53, NST = 37, and NTL = 25 (n = 189). Images were resized to 224 × 224 pixels, normalized, and augmented only in the training subset. Adam and RMSprop were evaluated. Accuracy and macro-averaged precision, recall, and F1-score were calculated from the integer-count evaluation matrices. The highest held-out accuracy, 187/189 (0.9894), was reached by seven model–optimizer configurations: DenseNet201–RMSprop, XCiT Medium 24 P16–Adam, ConvNeXt Base–Adam, CoAtNet–Adam, ConvNeXt Base–RMSprop, CoAtNet–RMSprop, and the proposed LGH-Transformer–RMSprop; MobileNet–Adam and Xcep-tion–RMSprop each yielded 185/189 (0.9788), and LGH-Transformer–Adam yielded 186/189 (0.9841). On a 189-frame test set, differences in one to two frames do not establish statistical superiority, so these configurations are reported as a top-accuracy tier rather than as a ranked superiority claim. Within this tier, the LGH-Transformer had the smallest computational footprint (≈18.2 M parameters and ≈4.1 GFLOPs, versus 20.0–88.6 M parameters and 4.3–16.2 GFLOPs for the other top-tier models). Component-wise ablation experiments showed accuracy decreases of 0.0106, 0.0053, and 0.0053 after removing the CNN tokenizer, local attention, and global attention, respectively. The LGH-Transformer reaches the top accuracy tier of this 18-architecture benchmark at the lowest parameter and FLOP budget within that tier, supporting an internally consistent, efficiency-oriented comparison within the available cohort. The absence of task-aligned external validation and the small number of held-out patients require the findings to be interpreted as exploratory internal validation rather than evidence of clinical generalizability.
SN - 1422-0067
ER -