Benchmarking Convolutional, Transformer, Hybrid, and Vision Language Models for Multi Disease Retinal Screening

Early detection of vision-threatening conditions such as diabetic retinopathy, glaucoma, and age-related macular degeneration depends on retinal fundus image analysis, but manual assessment is slow and expert-dependent. Automated convolutional neural networks classify fundus images accurately yet act as black boxes, and existing retinal vessel segmentation methods lose discriminative power under pathology and seldom exploit attention or transformer backbones. Using the FIVES and DRIVE fundus datasets, we develop a two-pipeline framework that pairs four-class disease classification with attention- and transformer-based vessel segmentation, organised in three stages: (1) FIVES images are augmented by rotation and horizontal and vertical flips and used to fine-tune eight ImageNet-pretrained CNNs: ResNet101, DenseNet169, Xception, InceptionV3, DenseNet121, InceptionResNetV2, ResNet50, and EfficientNetB0. (2) Five gradient-based explanation methods, Grad-CAM, Grad-CAM++, Score-CAM, Faster Score-CAM, and Layer-CAM, are computed on the final convolutional block of each classifier and compared qualitatively across architectures. (3) Ten U-Net variants are benchmarked for vessel segmentation: TransUNet (hybrid CNN--Transformer encoder) and Attention U-Net (gated skip connections), evaluated with ResNet50V2, ResNet101V2, and ResNet152V2 backbones, along with additional Attention U-Net configurations using DenseNet backbones, and the fully transformer-based Swin-UNet. ResNet101 gives the highest classification accuracy: 94.17% (F1 0.942) $>$ 88.33% for EfficientNetB0. For segmentation, the architecture ranking is consistent on both datasets: Attention U-Net $>$ TransUNet $>$ Swin-UNet. The strongest configuration is Attention U-Net with a ResNet101V2 backbone: FIVES IoU 0.722, Dice 0.838; DRIVE IoU 0.648, Dice 0.787, lifting DRIVE IoU 60.80 $\rightarrow$ 64.83 over a prior custom U-Net.

Skin cancer is a common and fast rising malignancy worldwide. Early detection is critical for improving outcomes. Deep learning models trained on dermoscopic and clinical images can support automated and fast triage. However, many studies evaluate only a limited set of architectures. Experimental setups also vary across studies. In this paper, we present a unified evaluation of twelve deep learning models for binary skin cancer detection on the PAD-UFES-20 dataset. The models span four families: convolutional neural networks (CNN), vision transformers (ViT), hybrid convolution transformer backbones, and vision language models (VLM). Performance is assessed using AUC, the maximum F1 score with its precision and recall, and sensitivity at 80% specificity, reflecting screening oriented requirements. Our results show that well tuned CNNs already provide strong baselines, but transformer based families consistently improve discrimination. Hybrid models (MaxViT Tiny, CoAtNet0) and a SigLIP based VLM achieve the best overall trade off between ranking performance and clinically relevant operating points, while CLIP based model offers high precision. The full codebase for all experiments is publicly released. Together, these findings offer practical guidance on which model families are most suitable for real world deployment in skin cancer screening and establish a reproducible reference point for future work on PAD-UFES-20.

The quality of diabetic retinopathy (DR) screening relies on the ability to correctly grade severity; however, many deep-learning (DL) classifiers cannot be easily interpreted in the clinical context. This study presents a methodology that combines strong discriminative models with multimodal explanations, converting retinal pixels into clinically interpretable outputs. Using the APTOS 2019 benchmark, we evaluated six representative CNN- and transformer-based backbones under a controlled protocol with stratified five-fold cross-validation. We then compared ensembling strategies (hard voting, weighted soft voting, stacking) and investigated a hybrid class-level fusion variant to exploit grade-specific advantages. For interpretability, we produced Grad-CAM++ visual attribution maps and short textual rationales using vision-language models (VLMs) conditioned on the fundus image and classifier outputs under conservative prompting constraints. Modern CNN backbones (ResNet-50 and ConvNeXt-Tiny) provided the strongest single-model baselines, with cross-validated QWK up to 0.919 and 0.914, respectively. Ensembling improved ordinal agreement, and weighted soft voting was the most consistent across folds (QWK 0.934 +/- 0.017). Hybrid class-level fusion was competitive but did not yield a statistically reliable improvement over standard fusion in paired fold comparisons (Holm-adjusted p >= 1.000). For explanation quality, Grad-CAM++ offered plausible but coarse localization, and VLM rationales were generally grade-consistent. Quantitatively, VLM variants showed a trade-off between clinical completeness and template-level semantic similarity (coverage 0.700 vs. BERTScore 0.072), while image-text alignment was comparable (CLIPScore approximately 0.34).