Samuel Ofosu Mensah

Kerol Djoumessi, Samuel Ofosu Mensah, Philipp Berens

In many medical imaging tasks, convolutional neural networks (CNNs) efficiently extract local features hierarchically. More recently, vision transformers (ViTs) have gained popularity, using self-attention mechanisms to capture global dependencies, but lacking the inherent spatial localization of convolutions. Therefore, hybrid models combining CNNs and ViTs have been developed to combine the strengths of both architectures. However, such hybrid models are difficult to interpret, which hinders their application in medical imaging. In this work, we introduce an interpretable-by-design hybrid fully convolutional CNN-Transformer architecture for retinal disease detection. Unlike widely used post-hoc saliency methods for ViTs, our approach generates faithful and localized evidence maps that directly reflect the mode's decision process. We evaluated our method on two medical tasks focused on disease detection using color fundus images. Our model achieves state-of-the-art predictive performance compared to black-box and interpretable models and provides class-specific sparse evidence maps in a single forward pass. The code is available at: https://github.com/kdjoumessi/Self-Explainable-CNN-Transformer.

Samuel Ofosu Mensah, Maria Camila Roa Carvajal, Kerol Djoumessi et al.

Foundation models are used to extract transferable representations from large amounts of unlabeled data, typically via self-supervised learning (SSL). However, many of these models rely on architectures that offer limited interpretability, which is a critical issue in high-stakes domains such as medical imaging. We propose Dual-IFM, a foundation model that is interpretable-by-design in two ways: First, it provides local interpretability for individual images through class evidence maps that are faithful to the decision-making process. Second, it provides global interpretability for entire datasets through a 2D projection layer that allows for direct visualization of the model's representation space. We trained our model on over 800,000 color fundus photography from various sources to learn generalizable, interpretable representations for different downstream tasks. Our results show that our model reaches a performance range similar to that of state-of-the-art foundation models with up to $16\times$ the number of parameters, while providing interpretable predictions on out-of-distribution data. Our results suggest that large-scale SSL pretraining paired with inherent interpretability can lead to robust representations for retinal imaging.

Samuel Ofosu Mensah, Kerol Djoumessi, Philipp Berens

Federated learning (FL) provides a promising paradigm for collaboratively training machine learning models across distributed data sources while maintaining privacy. Nevertheless, real-world FL often faces major challenges including communication overhead during the transfer of large model parameters and statistical heterogeneity, arising from non-identical independent data distributions across clients. In this work, we propose an FL framework that 1) provides inherent interpretations using prototypes, and 2) tackles statistical heterogeneity by utilising lightweight adapter modules to act as compressed surrogates of local models and guide clients to achieve generalisation despite varying client distribution. Each client locally refines its model by aligning class embeddings toward prototype representations and simultaneously adjust the lightweight adapter. Our approach replaces the need to communicate entire model weights with prototypes and lightweight adapters. This design ensures that each client's model aligns with a globally shared structure while minimising communication load and providing inherent interpretations. Moreover, we conducted our experiments on a real-world retinal fundus image dataset, which provides clinical-site information. We demonstrate inherent interpretable capabilities and perform a classification task, which shows improvements in accuracy over baseline algorithms.

Samuel Ofosu Mensah, Bubacarr Bah, Willie Brink

Convolutional Neural Networks (CNNs) have successfully been used to classify diabetic retinopathy (DR) fundus images in recent times. However, deeper representations in CNNs may capture higher-level semantics at the expense of spatial resolution. To make predictions usable for ophthalmologists, we use a post-attention technique called Gradient-weighted Class Activation Mapping (Grad-CAM) on the penultimate layer of deep learning models to produce coarse localisation maps on DR fundus images. This is to help identify discriminative regions in the images, consequently providing evidence for ophthalmologists to make a diagnosis and potentially save lives by early diagnosis. Specifically, this study uses pre-trained weights from four state-of-the-art deep learning models to produce and compare localisation maps of DR fundus images. The models used include VGG16, ResNet50, InceptionV3, and InceptionResNetV2. We find that InceptionV3 achieves the best performance with a test classification accuracy of 96.07%, and localise lesions better and faster than the other models.