A Step towards Interpretable Multimodal AI Models with MultiFIX

2025-05-16

AI Summary

To address the lack of trustworthiness and interpretability of multimodal AI models in high-stakes domains such as healthcare, this paper proposes MultiFIX—a novel framework that introduces modality-specific interpretability modules *early* in the deep feature extraction pipeline: symbolic regression for tabular data and Grad-CAM integration for imaging data, enabling concurrent feature-level and fusion-level interpretability. Crucially, MultiFIX pioneers an “interpretability-forward” design, embedding transparency directly into the modeling workflow without compromising predictive accuracy. Evaluated on multiple synthetic multimodal datasets, MultiFIX achieves accuracy comparable to black-box baselines while providing verifiable, quantitative attribution of modality contributions and fully traceable, stepwise fusion decision paths. These advances significantly enhance clinical deployability and stakeholder trust.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Real-world problems are often dependent on multiple data modalities, making multimodal fusion essential for leveraging diverse information sources. In high-stakes domains, such as in healthcare, understanding how each modality contributes to the prediction is critical to ensure trustworthy and interpretable AI models. We present MultiFIX, an interpretability-driven multimodal data fusion pipeline that explicitly engineers distinct features from different modalities and combines them to make the final prediction. Initially, only deep learning components are used to train a model from data. The black-box (deep learning) components are subsequently either explained using post-hoc methods such as Grad-CAM for images or fully replaced by interpretable blocks, namely symbolic expressions for tabular data, resulting in an explainable model. We study the use of MultiFIX using several training strategies for feature extraction and predictive modeling. Besides highlighting strengths and weaknesses of MultiFIX, experiments on a variety of synthetic datasets with varying degrees of interaction between modalities demonstrate that MultiFIX can generate multimodal models that can be used to accurately explain both the extracted features and their integration without compromising predictive performance.

Problem

• Enhancing interpretability in multimodal AI models for diverse data sources
• Ensuring trustworthy predictions in high-stakes domains like healthcare
• Balancing model accuracy with explainable feature integration across modalities

Innovation

• Multimodal fusion with interpretability-driven pipeline
• Replaces black-box components with explainable blocks
• Maintains predictive performance while ensuring explainability