-
Interpretable Multimodal Learning for Cardiovascular Hemodynamics Assessment
Authors:
Prasun C Tripathi,
Sina Tabakhi,
Mohammod N I Suvon,
Lawrence Schöb,
Samer Alabed,
Andrew J Swift,
Shuo Zhou,
Hai** Lu
Abstract:
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 PA…
▽ More
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.
△ Less
Submitted 6 April, 2024;
originally announced April 2024.
-
Universal Feature Selection Tool (UniFeat): An Open-Source Tool for Dimensionality Reduction
Authors:
Sina Tabakhi,
Parham Moradi
Abstract:
The Universal Feature Selection Tool (UniFeat) is an open-source tool developed entirely in Java for performing feature selection processes in various research areas. It provides a set of well-known and advanced feature selection methods within its significant auxiliary tools. This allows users to compare the performance of feature selection methods. Moreover, due to the open-source nature of UniF…
▽ More
The Universal Feature Selection Tool (UniFeat) is an open-source tool developed entirely in Java for performing feature selection processes in various research areas. It provides a set of well-known and advanced feature selection methods within its significant auxiliary tools. This allows users to compare the performance of feature selection methods. Moreover, due to the open-source nature of UniFeat, researchers can use and modify it in their research, which facilitates the rapid development of new feature selection algorithms.
△ Less
Submitted 30 November, 2022;
originally announced November 2022.
-
Multimodal Learning for Multi-Omics: A Survey
Authors:
Sina Tabakhi,
Mohammod Naimul Islam Suvon,
Pegah Ahadian,
Hai** Lu
Abstract:
With advanced imaging, sequencing, and profiling technologies, multiple omics data become increasingly available and hold promises for many healthcare applications such as cancer diagnosis and treatment. Multimodal learning for integrative multi-omics analysis can help researchers and practitioners gain deep insights into human diseases and improve clinical decisions. However, several challenges a…
▽ More
With advanced imaging, sequencing, and profiling technologies, multiple omics data become increasingly available and hold promises for many healthcare applications such as cancer diagnosis and treatment. Multimodal learning for integrative multi-omics analysis can help researchers and practitioners gain deep insights into human diseases and improve clinical decisions. However, several challenges are hindering the development in this area, including the availability of easily accessible open-source tools. This survey aims to provide an up-to-date overview of the data challenges, fusion approaches, datasets, and software tools from several new perspectives. We identify and investigate various omics data challenges that can help us understand the field better. We categorize fusion approaches comprehensively to cover existing methods in this area. We collect existing open-source tools to facilitate their broader utilization and development. We explore a broad range of omics data modalities and a list of accessible datasets. Finally, we summarize future directions that can potentially address existing gaps and answer the pressing need to advance multimodal learning for multi-omics data analysis.
△ Less
Submitted 19 December, 2022; v1 submitted 29 November, 2022;
originally announced November 2022.
