Patient-specific cardiovascular simulation has become a paradigm in cardiovascular research and is emerging as a powerful tool in basic, translational and clinical research. In this paper we discuss the recent development of a fully open-source SimVascular software package, which provides a complete pipeline from medical image data segmentation to patient-specific blood flow simulation and analysis. This package serves as a research tool for cardiovascular modeling and simulation, and has contributed to numerous advances in personalized medicine, surgical planning and medical device design. The SimVascular software has recently been refactored and expanded to enhance functionality, usability, efficiency and accuracy of image-based patient-specific modeling tools. Moreover, SimVascular previously required several licensed components that hindered new user adoption and code management and our recent developments have replaced these commercial components to create a fully open source pipeline. These developments foster advances in cardiovascular modeling research, increased collaboration, standardization of methods, and a growing developer community.

Abstract. In this work, we present a novel 3D-Convolutional Neural Network (CNN) architecture called I2I-3D that predicts boundary location in volumetric data. Our fine-to-fine, deeply supervised framework addresses three critical issues to 3D boundary detection: (1) e cient, holistic, end-to-end volumetric label training and prediction (2) precise voxel-level prediction to capture fine scale structures prevalent in medical data and (3) directed multi-scale, multi-level feature learning. We evaluate our approach on a dataset consisting of 93 medical image volumes with a wide variety of anatomical regions and vascular structures. In the process, we also introduce HED-3D, a 3D extension of the state-of-theart 2D edge detector (HED). We show that our deep learning approach out-performs, the current state-of-the-art in 3D vascular boundary detection (structured forests 3D), by a large margin, as well as HED applied to slices, and HED-3D while successfully localizing fine structures. With our approach, boundary detection takes about one minute on a typical 512x512x512 volume.

The periocular region, the region of the face surrounding the eyes, has gained increasing attention in biometrics in recent years. This region of the face is of particular interest when trying to identify a person whose face is partially occluded. We propose the novel idea of applying the information obtained from the periocular region to identify the gender of a person, which is a type of soft biometric recognition. We gradually narrow the region of interest of the face to explore the feasibility of using smaller, eye-centered regions for building a robust gender classifier around the periocular region alone. Our experimental results show that at least an 85% classification rate is still obtainable using only the periocular region with a database of 936 low resolution images collected from the web.

Abstract. Computational simulations provide detailed hemodynamics and physiological data that can assist in clinical decision-making. However, accurate cardiovascular simulations require complete 3D models constructed from image data. Though edge localization is a key aspect in pinpointing vessel walls in many segmentation tools, the edge detection algorithms widely utilized by the medical imaging community have remained static. In this paper, we propose a novel approach to medical image edge detection by adopting the powerful structured forest detector and extending its application to the medical imaging domain. First, we specify an effective set of medical imaging driven features. Second, we directly incorporate an adaptive prior to create a robust three-dimensional edge classifier. Last, we boost our accuracy through an intelligent sampling scheme that only samples areas of importance to edge fidelity.Through experimentation, we demonstrate that the proposed method outperforms widely used edge detectors and probabilistic boosting tree edge classifiers and is robust to error in a prori information.

No abstract