Papers by Orlando Aristizabal
2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI)
High-frequency ultrasound (HFU) is well suited for imaging embryonic mice due to its noninvasive ... more High-frequency ultrasound (HFU) is well suited for imaging embryonic mice due to its noninvasive and real-time characteristics. However, manual segmentation of the brain ventricles (BVs) and body requires substantial time and expertise. This work proposes a novel deep learning based endto-end auto-context refinement framework, consisting of two stages. The first stage produces a low resolution segmentation of the BV and body simultaneously. The resulting probability map for each object (BV or body) is then used to crop a region of interest (ROI) around the target object in both the original image and the probability map to provide context to the refinement segmentation network. Joint training of the two stages provides significant improvement in Dice Similarity Coefficient (DSC) over using only the first stage (0.818 to 0.906 for the BV, and 0.919 to 0.934 for the body). The proposed method significantly reduces the inference time (102.36 to 0.09 s/volume ≈1000x faster) while slightly improves the segmentation accuracy over the previous methods using slidewindow approaches.
2018 IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2018
Volumetric analysis of brain ventricle (BV) structure is a key tool in the study of central nervo... more Volumetric analysis of brain ventricle (BV) structure is a key tool in the study of central nervous system development in embryonic mice. High-frequency ultrasound (HFU) is the only non-invasive, real-time modality available for rapid volumetric imaging of embryos in utero. However, manual segmentation of the BV from HFU volumes is tedious, time-consuming, and requires specialized expertise. In this paper, we propose a novel deep learning based BV segmentation system for whole-body HFU images of mouse embryos. Our fully automated system consists of two modules: localization and segmentation. It first applies a volumetric convolutional neural network on a 3D sliding window over the entire volume to identify a 3D bounding box containing the entire BV. It then employs a fully convolutional network to segment the detected bounding box into BV and background. The system achieves a Dice Similarity Coefficient (DSC) of 0.8956 for BV segmentation on an unseen 111 HFU volume test set surpassing the previous state-of-the-art method (DSC of 0.7119) by a margin of 25%.
IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 2021
The rodent heart is frequently used to study human cardiovascular disease (CVD). Although advance... more The rodent heart is frequently used to study human cardiovascular disease (CVD). Although advanced cardiovascular ultrasound imaging methods are available for human clinical practice, application of these techniques to small animals remains limited due to the temporal and spatialresolution demands. Here, an ultrasound vector-flow workflow is demonstrated that enables visualization and quantification of the complex hemodynamics within the mouse heart. Wild type (WT) and fibroblast growth factor homologous factor 2 (FHF2)-deficient mice (Fhf2 KO/Y), which present with hyperthermia-induced ECG abnormalities highly reminiscent of Brugada syndrome, were used as a mouse model of human CVD. An 18-MHz linear array was used to acquire highspeed (30 kHz), plane-wave data of the left ventricle (LV) while increasing core body temperature up to 41.5°C. Hexplex (i.e., six output) processing of the raw data sets produced the output of vector-flow estimates (magnitude and phase); B-mode and color-Doppler images; Doppler spectrograms; and local time histories of vorticity and pericardium motion. Fhf2 WT/Y mice had repeatable beat-to-beat cardiac function, including vortex formation during diastole, at all temperatures. In contrast, Fhf2 KO/Y mice displayed dyssynchronous contractile motion that disrupted normal inflow vortex formation and impaired LV filling as temperature rose. The hexplex processing approach demonstrates the ability to visualize and quantify the interplay between hemodynamic and mechanical function in a mouse model of human CVD.
2020 IEEE International Ultrasonics Symposium (IUS), 2020
We have applied a deep learning framework, trained on mouse embryo images acquired with a 40 MHz ... more We have applied a deep learning framework, trained on mouse embryo images acquired with a 40 MHz annular array, to volumetric data acquired with a VisualSonics Vevo 3100 commercial scanner using a 40-MHz linear array. The deep learning framework was robust enough to accurately segment out the body and the brain ventricle from the 3D data generated by the commercial scanner. These results show that there is no need to retrain the algorithm with hundreds of new manually segmented datasets.
2019 IEEE International Ultrasonics Symposium (IUS), 2019
High-frequency ultrasound (HFU) is well suited for imaging embryonic mice in vivo because it is n... more High-frequency ultrasound (HFU) is well suited for imaging embryonic mice in vivo because it is non-invasive and real-time. Manual segmentation of the brain ventricles (BVs) and whole body from 3D HFU images is time-consuming and requires specialized training. This paper presents a deep-learning-based segmentation pipeline which automates several time-consuming, repetitive tasks currently performed to study genetic mutations in developing mouse embryos. Namely, the pipeline accurately segments the BV and body regions in 3D HFU images of mouse embryos, despite significant challenges due to position and shape variation of the embryos, as well as imaging artifacts. Based on the BV segmentation, a 3D convolutional neural network (CNN) is further trained to detect embryos with the Engrailed-1 (En1) mutation. The algorithms achieve 0.896 and 0.925 Dice Similarity Coefficient (DSC) for BV and body segmentation, respectively, and 95.8% accuracy on mutant classification. Through gradient based interrogation and visualization of the trained classifier, it is demonstrated that the model focuses on the morphological structures known to be affected by the En1 mutation.
IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 2021
Segmentation and mutant classification of high-frequency ultrasound (HFU) mouse embryo brain vent... more Segmentation and mutant classification of high-frequency ultrasound (HFU) mouse embryo brain ventricle (BV) and body images can provide valuable information for developmental biologists. However, manual segmentation and identification of BV and body requires substantial time and expertise. This paper proposes an accurate, efficient and explainable deep learning pipeline for automatic segmentation and classification of the BV and body. For segmentation, a two-stage framework is implemented. The first stage produces a low-resolution segmentation map, which is then used to crop a region of interest (ROI) around the target object and serve as the probability
The Journal of the Acoustical Society of America, 2018
High-frequency ultrasound Doppler modes have been used extensively for murine cardiovascular (CV)... more High-frequency ultrasound Doppler modes have been used extensively for murine cardiovascular (CV) studies, but traditional linear-array imaging modes are limited in terms of spatial and temporal resolution. Plane-wave imaging methods allow for high-speed vector-flow information to be obtained throughout a full image frame. Plane-wave imaging has been demonstrated in human CV studies, but its use in mouse models has received minimal attention. A Verasonics Vantage with an 18-MHz linear array was used to acquire plane-wave data at a frame rate of 30 kHz from the left ventricle of adult mice. Batches of 3 transmissions spanning ± 5 degrees were sent out. The mouse was placed supine on a heated imaging platform and then 2D + time data sequences. The data were beamformed using standard delay-and-sum methods and vector-flow estimates were obtained at each pixel location using a least-squares, multi-angle Doppler analysis approach. Vortex patterns in the left ventricle were visualized over several heart cycles s...
2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI)
High-frequency ultrasound (HFU) is well suited for imaging embryonic mice due to its noninvasive ... more High-frequency ultrasound (HFU) is well suited for imaging embryonic mice due to its noninvasive and real-time characteristics. However, manual segmentation of the brain ventricles (BVs) and body requires substantial time and expertise. This work proposes a novel deep learning based endto-end auto-context refinement framework, consisting of two stages. The first stage produces a low resolution segmentation of the BV and body simultaneously. The resulting probability map for each object (BV or body) is then used to crop a region of interest (ROI) around the target object in both the original image and the probability map to provide context to the refinement segmentation network. Joint training of the two stages provides significant improvement in Dice Similarity Coefficient (DSC) over using only the first stage (0.818 to 0.906 for the BV, and 0.919 to 0.934 for the body). The proposed method significantly reduces the inference time (102.36 to 0.09 s/volume ≈1000x faster) while slightly improves the segmentation accuracy over the previous methods using slidewindow approaches.
The Journal of the Acoustical Society of America, 2019
2018 IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2018
Volumetric analysis of brain ventricle (BV) structure is a key tool in the study of central nervo... more Volumetric analysis of brain ventricle (BV) structure is a key tool in the study of central nervous system development in embryonic mice. High-frequency ultrasound (HFU) is the only non-invasive, real-time modality available for rapid volumetric imaging of embryos in utero. However, manual segmentation of the BV from HFU volumes is tedious, time-consuming, and requires specialized expertise. In this paper, we propose a novel deep learning based BV segmentation system for whole-body HFU images of mouse embryos. Our fully automated system consists of two modules: localization and segmentation. It first applies a volumetric convolutional neural network on a 3D sliding window over the entire volume to identify a 3D bounding box containing the entire BV. It then employs a fully convolutional network to segment the detected bounding box into BV and background. The system achieves a Dice Similarity Coefficient (DSC) of 0.8956 for BV segmentation on an unseen 111 HFU volume test set surpassing the previous state-of-the-art method (DSC of 0.7119) by a margin of 25%.
Scientific reports, Jan 30, 2017
Real-time imaging of the embryonic murine cardiovascular system is challenging due to the small s... more Real-time imaging of the embryonic murine cardiovascular system is challenging due to the small size of the mouse embryo and rapid heart rate. High-frequency, linear-array ultrasound systems designed for small-animal imaging provide high-frame-rate and Doppler modes but are limited in regards to the field of view that can be imaged at fine-temporal and -spatial resolution. Here, a plane-wave imaging method was used to obtain high-speed image data from in utero mouse embryos and multi-angle, vector-flow algorithms were applied to the data to provide information on blood flow patterns in major organs. An 18-MHz linear array was used to acquire plane-wave data at absolute frame rates ≥10 kHz using a set of fixed transmission angles. After beamforming, vector-flow processing and image compounding, effective frame rates were on the order of 2 kHz. Data were acquired from the embryonic liver, heart and umbilical cord. Vector-flow results clearly revealed the complex nature of blood-flow p...
IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 2021
The rodent heart is frequently used to study human cardiovascular disease (CVD). Although advance... more The rodent heart is frequently used to study human cardiovascular disease (CVD). Although advanced cardiovascular ultrasound imaging methods are available for human clinical practice, application of these techniques to small animals remains limited due to the temporal and spatialresolution demands. Here, an ultrasound vector-flow workflow is demonstrated that enables visualization and quantification of the complex hemodynamics within the mouse heart. Wild type (WT) and fibroblast growth factor homologous factor 2 (FHF2)-deficient mice (Fhf2 KO/Y), which present with hyperthermia-induced ECG abnormalities highly reminiscent of Brugada syndrome, were used as a mouse model of human CVD. An 18-MHz linear array was used to acquire highspeed (30 kHz), plane-wave data of the left ventricle (LV) while increasing core body temperature up to 41.5°C. Hexplex (i.e., six output) processing of the raw data sets produced the output of vector-flow estimates (magnitude and phase); B-mode and color-Doppler images; Doppler spectrograms; and local time histories of vorticity and pericardium motion. Fhf2 WT/Y mice had repeatable beat-to-beat cardiac function, including vortex formation during diastole, at all temperatures. In contrast, Fhf2 KO/Y mice displayed dyssynchronous contractile motion that disrupted normal inflow vortex formation and impaired LV filling as temperature rose. The hexplex processing approach demonstrates the ability to visualize and quantify the interplay between hemodynamic and mechanical function in a mouse model of human CVD.
2020 IEEE International Ultrasonics Symposium (IUS), 2020
We have applied a deep learning framework, trained on mouse embryo images acquired with a 40 MHz ... more We have applied a deep learning framework, trained on mouse embryo images acquired with a 40 MHz annular array, to volumetric data acquired with a VisualSonics Vevo 3100 commercial scanner using a 40-MHz linear array. The deep learning framework was robust enough to accurately segment out the body and the brain ventricle from the 3D data generated by the commercial scanner. These results show that there is no need to retrain the algorithm with hundreds of new manually segmented datasets.
The Journal of the Acoustical Society of America, 2016
Plane-wave imaging methods allow for high-speed image capture at a time interval equal to round t... more Plane-wave imaging methods allow for high-speed image capture at a time interval equal to round trip acoustic propagation. Plane-wave imaging is ideally suited for cardiovascular imaging where fine-temporal resolution can reveal important information about cardiac mechanics and blood flow patterns. While plane-wave imaging has been demonstrated in humans for cardiovascular studies, its use in mouse models lags because instrumentation is not yet widely available at appropriate ultrasound frequencies. Thus, the amount of functional information that can be mined from mouse models of cardiovascular disease is limited. Here, an 18-MHz linear-array probe was used to acquire plane-wave data at a frame rate of 10 kHz from an in utero, E14.5 mouse embryo. The probe had 128 elements, 1.5 mm elevation aperture, and 8-mm elevation focus. The mother was placed supine on a heated mouse imaging platform, and then, a series of 2D + time data sequences were captured. The data were beamformed using standard delay-and-sum m...
2019 IEEE International Ultrasonics Symposium (IUS), 2019
High-frequency ultrasound (HFU) is well suited for imaging embryonic mice in vivo because it is n... more High-frequency ultrasound (HFU) is well suited for imaging embryonic mice in vivo because it is non-invasive and real-time. Manual segmentation of the brain ventricles (BVs) and whole body from 3D HFU images is time-consuming and requires specialized training. This paper presents a deep-learning-based segmentation pipeline which automates several time-consuming, repetitive tasks currently performed to study genetic mutations in developing mouse embryos. Namely, the pipeline accurately segments the BV and body regions in 3D HFU images of mouse embryos, despite significant challenges due to position and shape variation of the embryos, as well as imaging artifacts. Based on the BV segmentation, a 3D convolutional neural network (CNN) is further trained to detect embryos with the Engrailed-1 (En1) mutation. The algorithms achieve 0.896 and 0.925 Dice Similarity Coefficient (DSC) for BV and body segmentation, respectively, and 95.8% accuracy on mutant classification. Through gradient based interrogation and visualization of the trained classifier, it is demonstrated that the model focuses on the morphological structures known to be affected by the En1 mutation.
IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 2021
Segmentation and mutant classification of high-frequency ultrasound (HFU) mouse embryo brain vent... more Segmentation and mutant classification of high-frequency ultrasound (HFU) mouse embryo brain ventricle (BV) and body images can provide valuable information for developmental biologists. However, manual segmentation and identification of BV and body requires substantial time and expertise. This paper proposes an accurate, efficient and explainable deep learning pipeline for automatic segmentation and classification of the BV and body. For segmentation, a two-stage framework is implemented. The first stage produces a low-resolution segmentation map, which is then used to crop a region of interest (ROI) around the target object and serve as the probability
Journal of Clinical Oncology, 2015
324 Background: Pancreatic cancer is well known for its aggressive clinical course and resistance... more 324 Background: Pancreatic cancer is well known for its aggressive clinical course and resistance to chemotherapy. The advent of new mouse models of pancreatic cancer have accelerated our understanding of tumorigenesis and enabled preclinical testing of experimental therapeutics with a desire to translate these findings into meaningful clinical treatments. Methods: We have developed a model where pancreatic cells obtained from a KrasG12D;Trp53R172H genetically engineered mouse can be cultivated in two dimensional cell culture and implanted into the pancreas of a immunocompetent syngeneic mouse allowing for tumor formation in situ. In addition, we are using this model to study the effectiveness of new drug combination therapy such as gemcitabine, albumin-bound paclitaxel and CD40 agonist immunotherapy using overall survival as a primary endpoint. Results: These cells generate tumors of five millimeter diameter within two weeks of implantation with 100% efficiency. Because cancer cell...
The Journal of the Acoustical Society of America, 2018
High-frequency ultrasound Doppler modes have been used extensively for murine cardiovascular (CV)... more High-frequency ultrasound Doppler modes have been used extensively for murine cardiovascular (CV) studies, but traditional linear-array imaging modes are limited in terms of spatial and temporal resolution. Plane-wave imaging methods allow for high-speed vector-flow information to be obtained throughout a full image frame. Plane-wave imaging has been demonstrated in human CV studies, but its use in mouse models has received minimal attention. A Verasonics Vantage with an 18-MHz linear array was used to acquire plane-wave data at a frame rate of 30 kHz from the left ventricle of adult mice. Batches of 3 transmissions spanning ± 5 degrees were sent out. The mouse was placed supine on a heated imaging platform and then 2D + time data sequences. The data were beamformed using standard delay-and-sum methods and vector-flow estimates were obtained at each pixel location using a least-squares, multi-angle Doppler analysis approach. Vortex patterns in the left ventricle were visualized over several heart cycles s...
Uploads
Papers by Orlando Aristizabal