Manuel Tran

Manuel Tran, Sophia J. Wagner, Melanie Boxberg et al.

In computational pathology, we often face a scarcity of annotations and a large amount of unlabeled data. One method for dealing with this is semi-supervised learning which is commonly split into a self-supervised pretext task and a subsequent model fine-tuning. Here, we compress this two-stage training into one by introducing S5CL, a unified framework for fully-supervised, self-supervised, and semi-supervised learning. With three contrastive losses defined for labeled, unlabeled, and pseudo-labeled images, S5CL can learn feature representations that reflect the hierarchy of distance relationships: similar images and augmentations are embedded the closest, followed by different looking images of the same class, while images from separate classes have the largest distance. Moreover, S5CL allows us to flexibly combine these losses to adapt to different scenarios. Evaluations of our framework on two public histopathological datasets show strong improvements in the case of sparse labels: for a H&E-stained colorectal cancer dataset, the accuracy increases by up to 9% compared to supervised cross-entropy loss; for a highly imbalanced dataset of single white blood cells from leukemia patient blood smears, the F1-score increases by up to 6%.

Rushin H. Gindra, Giovanni Palla, Mathias Nguyen et al.

Spatial transcriptomics enables simultaneous measurement of gene expression and tissue morphology, offering unprecedented insights into cellular organization and disease mechanisms. However, the field lacks comprehensive benchmarks for evaluating multimodal learning methods that leverage both histology images and gene expression data. Here, we present HESCAPE, a large-scale benchmark for cross-modal contrastive pretraining in spatial transcriptomics, built on a curated pan-organ dataset spanning 6 different gene panels and 54 donors. We systematically evaluated state-of-the-art image and gene expression encoders across multiple pretraining strategies and assessed their effectiveness on two downstream tasks: gene mutation classification and gene expression prediction. Our benchmark demonstrates that gene expression encoders are the primary determinant of strong representational alignment, and that gene models pretrained on spatial transcriptomics data outperform both those trained without spatial data and simple baseline approaches. However, downstream task evaluation reveals a striking contradiction: while contrastive pretraining consistently improves gene mutation classification performance, it degrades direct gene expression prediction compared to baseline encoders trained without cross-modal objectives. We identify batch effects as a key factor that interferes with effective cross-modal alignment. Our findings highlight the critical need for batch-robust multimodal learning approaches in spatial transcriptomics. To accelerate progress in this direction, we release HESCAPE, providing standardized datasets, evaluation protocols, and benchmarking tools for the community

Manuel Tran, Amal Lahiani, Yashin Dicente Cid et al.

Vision Transformers (ViTs) and Swin Transformers (Swin) are currently state-of-the-art in computational pathology. However, domain experts are still reluctant to use these models due to their lack of interpretability. This is not surprising, as critical decisions need to be transparent and understandable. The most common approach to understanding transformers is to visualize their attention. However, attention maps of ViTs are often fragmented, leading to unsatisfactory explanations. Here, we introduce a novel architecture called the B-cos Vision Transformer (BvT) that is designed to be more interpretable. It replaces all linear transformations with the B-cos transform to promote weight-input alignment. In a blinded study, medical experts clearly ranked BvTs above ViTs, suggesting that our network is better at capturing biomedically relevant structures. This is also true for the B-cos Swin Transformer (Bwin). Compared to the Swin Transformer, it even improves the F1-score by up to 4.7% on two public datasets.

Manuel Tran, Yashin Dicente Cid, Amal Lahiani et al.

Training multimodal foundation models is challenging due to the limited availability of multimodal datasets. While many public datasets pair images with text, few combine images with audio or text with audio. Even rarer are datasets that align all three modalities at once. Critical domains such as healthcare, infrastructure, or transportation are particularly affected by missing modalities. This makes it difficult to integrate all modalities into a large pre-trained neural network that can be used out-of-the-box or fine-tuned for different downstream tasks. We introduce LoReTTa (Linking mOdalities with a tRansitive and commutativE pre-Training sTrAtegy) to address this understudied problem. Our self-supervised framework unifies causal modeling and masked modeling with the rules of commutativity and transitivity. This allows us to transition within and between modalities. As a result, our pre-trained models are better at exploring the true underlying joint probability distribution. Given a dataset containing only the disjoint combinations (A, B) and (B, C), LoReTTa can model the relation A <-> C with A <-> B <-> C. In particular, we show that a transformer pre-trained with LoReTTa can handle any mixture of modalities at inference time, including the never-seen pair (A, C) and the triplet (A, B, C). We extensively evaluate our approach on a synthetic, medical, and reinforcement learning dataset. Across different domains, our universal multimodal transformer consistently outperforms strong baselines such as GPT, BERT, and CLIP on tasks involving the missing modality tuple.