Multimodal deep learning (MDL) has emerged as a transformative approach in computational pathology. By integrating complementary information from multiple data sources, MDL models have demonstrated superior predictive performance across diverse clinical tasks compared to unimodal models. However, the assumption that combining modalities inherently improves performance remains largely unexamined. We hypothesise that multimodal gains depend critically on the predictive quality of individual modalities, and that integrating weak modalities may introduce noise rather than complementary information. We test this hypothesis on a prostate cancer dataset with histopathology, radiology, and clinical data to predict time-to-biochemical recurrence. Our results confirm that combining high-performing modalities yield superior performance compared to unimodal approaches. However, integrating a poor-performing modality with other higher-performing modalities degrades predictive accuracy. These findings demonstrate that multimodal benefit requires selective, performance-guided integration rather than indiscriminate modality combination, with implications for MDL design across computational pathology and medical imaging.

F ondazione Policlinico Universitario Campus Bio-Medico, Rome, Italy Research Unit of Radiation Oncology, Department of Medicine and Surgery, Universit ` a Campus Bio-Medico di Roma, Rome, Italy Research Unit of Anatomical Pathology, Department of Medicine and Surgery, Universit ` a Campus Bio-Medico di Roma, Rome, Italy Department of Diagnostics and Intervention, Radiation Physics, Biomedical Engineering, Ume a University, Ume a, Sweden Abstract --This study proposes a novel approach combining Multimodal Deep Learning with intrinsic eXplainable Artificial Intelligence techniques to predict pathological response in non-small cell lung cancer patients undergoing neoadjuvant therapy. Due to the limitations of existing radiomics and unimodal deep learning approaches, we introduce an intermediate fusion strategy that integrates imaging and clinical data, enabling efficient interaction between data modalities. The proposed Multimodal Doctor-in-the-Loop method further enhances clinical relevance by embedding clinicians' domain knowledge directly into the training process, guiding the model's focus gradually from broader lung regions to specific lesions. Results demonstrate improved predictive accuracy and explainability, providing insights into optimal data integration strategies for clinical applications. I NTRODUCTION Non-small cell lung cancer (NSCLC) is the most common subtype of lung cancer, constituting approximately 85% of lung cancer cases [1]. Currently, surgery remains the main treatment for early-stage and resectable locally advanced NSCLC, although a notable number of patients experience post-surgery recurrence. Neoadjuvant therapy (NA T) has shown potential in improving overall survival rates and reducing the risk of distant disease recurrence [2]. Achieving a complete pathological response after NA T, indicating the absence of tumor cells in all specimens, may have a potential prognostic role and serve as a surrogate survival endpoint [3]. Evaluating pathological response before surgical resection provides valuable insights into tumor sensitivity to the administered therapy, enabling clinicians to tailor the type of treatment to the needs of patients and reserve surgical interventions only for patients who are most likely to benefit from them.

Cooperative perception enabled by Vehicle-to-Everything (V2X) communication holds significant promise for enhancing the perception capabilities of autonomous vehicles, allowing them to overcome occlusions and extend their field of view. However, existing research predominantly relies on simulated environments or static datasets, leaving the feasibility and effectiveness of V2X cooperative perception especially for intermediate fusion in real-world scenarios largely unexplored. In this work, we introduce V2X-ReaLO, an open online cooperative perception framework deployed on real vehicles and smart infrastructure that integrates early, late, and intermediate fusion methods within a unified pipeline and provides the first practical demonstration of online intermediate fusion's feasibility and performance under genuine real-world conditions. Additionally, we present an open benchmark dataset specifically designed to assess the performance of online cooperative perception systems. This new dataset extends V2X-Real dataset to dynamic, synchronized ROS bags and provides 25,028 test frames with 6,850 annotated key frames in challenging urban scenarios. By enabling real-time assessments of perception accuracy and communication lantency under dynamic conditions, V2X-ReaLO sets a new benchmark for advancing and optimizing cooperative perception systems in real-world applications. The codes and datasets will be released to further advance the field.

Accurate classification of histological subtypes of non-small cell lung cancer (NSCLC) is essential in the era of precision medicine, yet current invasive techniques are not always feasible and may lead to clinical complications. This study presents a multi-stage intermediate fusion approach to classify NSCLC subtypes from CT and PET images. Our method integrates the two modalities at different stages of feature extraction, using voxel-wise fusion to exploit complementary information across varying abstraction levels while preserving spatial correlations. We compare our method against unimodal approaches using only CT or PET images to demonstrate the benefits of modality fusion, and further benchmark it against early and late fusion techniques to highlight the advantages of intermediate fusion during feature extraction. Additionally, we compare our model with the only existing intermediate fusion method for histological subtype classification using PET/CT images. Our results demonstrate that the proposed method outperforms all alternatives across key metrics, with an accuracy and AUC equal to 0.724 and 0.681, respectively. This non-invasive approach has the potential to significantly improve diagnostic accuracy, facilitate more informed treatment decisions, and advance personalized care in lung cancer management.

In healthcare, the integration of multimodal data is pivotal for developing comprehensive diagnostic and predictive models. However, managing missing data remains a significant challenge in real-world applications. We introduce MARIA (Multimodal Attention Resilient to Incomplete datA), a novel transformer-based deep learning model designed to address these challenges through an intermediate fusion strategy. Unlike conventional approaches that depend on imputation, MARIA utilizes a masked self-attention mechanism, which processes only the available data without generating synthetic values. This approach enables it to effectively handle incomplete datasets, enhancing robustness and minimizing biases introduced by imputation methods. We evaluated MARIA against 10 state-of-the-art machine learning and deep learning models across 8 diagnostic and prognostic tasks. The results demonstrate that MARIA outperforms existing methods in terms of performance and resilience to varying levels of data incompleteness, underscoring its potential for critical healthcare applications.

Deep learning has revolutionized biomedical research by providing sophisticated methods to handle complex, high-dimensional data. Multimodal deep learning (MDL) further enhances this capability by integrating diverse data types such as imaging, textual data, and genetic information, leading to more robust and accurate predictive models. In MDL, differently from early and late fusion methods, intermediate fusion stands out for its ability to effectively combine modality-specific features during the learning process. This systematic review aims to comprehensively analyze and formalize current intermediate fusion methods in biomedical applications. We investigate the techniques employed, the challenges faced, and potential future directions for advancing intermediate fusion methods. Additionally, we introduce a structured notation to enhance the understanding and application of these methods beyond the biomedical domain. Our findings are intended to support researchers, healthcare professionals, and the broader deep learning community in developing more sophisticated and insightful multimodal models. Through this review, we aim to provide a foundational framework for future research and practical applications in the dynamic field of MDL.

Pulmonary Arterial Wedge Pressure (PAWP) is an essential cardiovascular hemodynamics marker to detect heart failure. In clinical practice, Right Heart Catheterization is considered a gold standard for assessing cardiac hemodynamics while non-invasive methods are often needed to screen high-risk patients from a large population. In this paper, we propose a multimodal learning pipeline to predict PAWP marker. We utilize complementary information from Cardiac Magnetic Resonance Imaging (CMR) scans (short-axis and four-chamber) and Electronic Health Records (EHRs). We extract spatio-temporal features from CMR scans using tensor-based learning. We propose a graph attention network to select important EHR features for prediction, where we model subjects as graph nodes and feature relationships as graph edges using the attention mechanism. We design four feature fusion strategies: early, intermediate, late, and hybrid fusion. With a linear classifier and linear fusion strategies, our pipeline is interpretable. We validate our pipeline on a large dataset of $2,641$ subjects from our ASPIRE registry. The comparative study against state-of-the-art methods confirms the superiority of our pipeline. The decision curve analysis further validates that our pipeline can be applied to screen a large population. The code is available at https://github.com/prasunc/hemodynamics.

Diffusion models have been widely used for conditional data cross-modal generation tasks such as text-to-image and text-to-video. However, state-of-the-art models still fail to align the generated visual concepts with high-level semantics in a language such as object count, spatial relationship, etc. We approach this problem from a multimodal data fusion perspective and investigate how different fusion strategies can affect vision-language alignment. We discover that compared to the widely used early fusion of conditioning text in a pretrained image feature space, a specially designed intermediate fusion can: (i) boost textto-image alignment with improved generation quality and (ii) improve training and inference efficiency by reducing low-rank text-to-image attention calculations. We perform experiments using a text-to-image generation task on the MS-COCO dataset. We compare our intermediate fusion mechanism with the classic early fusion mechanism on two common conditioning methods on a U-shaped ViT backbone. Our intermediate fusion model achieves higher CLIP Score and lower FID, with 20% reduced FLOPs, and 50% increased training speed compared against a strong U-ViT baseline with an early fusion.

An object handover between a robot and a human is a coordinated action which is prone to failure for reasons such as miscommunication, incorrect actions and unexpected object properties. Existing works on handover failure detection and prevention focus on preventing failures due to object slip or external disturbances. However, there is a lack of datasets and evaluation methods that consider unpreventable failures caused by the human participant. To address this deficit, we present the multimodal Handover Failure Detection dataset, which consists of failures induced by the human participant, such as ignoring the robot or not releasing the object. We also present two baseline methods for handover failure detection: (i) a video classification method using 3D CNNs and (ii) a temporal action segmentation approach which jointly classifies the human action, robot action and overall outcome of the action. The results show that video is an important modality, but using force-torque data and gripper position help improve failure detection and action segmentation accuracy.

Hateful memes are a growing menace on social media. While the image and its corresponding text in a meme are related, they do not necessarily convey the same meaning when viewed individually. Hence, detecting hateful memes requires careful consideration of both visual and textual information. Multimodal pre-training can be beneficial for this task because it effectively captures the relationship between the image and the text by representing them in a similar feature space. Furthermore, it is essential to model the interactions between the image and text features through intermediate fusion. Most existing methods either employ multimodal pre-training or intermediate fusion, but not both. In this work, we propose the Hate-CLIPper architecture, which explicitly models the cross-modal interactions between the image and text representations obtained using Contrastive Language-Image Pre-training (CLIP) encoders via a feature interaction matrix (FIM). A simple classifier based on the FIM representation is able to achieve state-of-the-art performance on the Hateful Memes Challenge (HMC) dataset with an AUROC of 85.8, which even surpasses the human performance of 82.65. Experiments on other meme datasets such as Propaganda Memes and TamilMemes also demonstrate the generalizability of the proposed approach. Finally, we analyze the interpretability of the FIM representation and show that cross-modal interactions can indeed facilitate the learning of meaningful concepts. The code for this work is available at https://github.com/gokulkarthik/hateclipper.