Search results

(1 - 5 of 5)
LOW-DOSE CARDIAC SPECT USING POST-FILTERING, DEEP LEARNING, AND MOTION CORRECTION
Title
LOW-DOSE CARDIAC SPECT USING POST-FILTERING, DEEP LEARNING, AND MOTION CORRECTION
Creator
Song, Chao
Date
2019
Description
Single photon emission computed tomography (SPECT) is an important technique in use today for the detection and evaluation of coronary artery...
Show moreSingle photon emission computed tomography (SPECT) is an important technique in use today for the detection and evaluation of coronary artery diseases. The image quality in cardiac SPECT can be adversely affected by cardiac motion and respiratory motion, both of which can lead to motion blur and non-uniform heart wall. In this thesis, we mainly investigate imaging de-noising algorithms and motion correction methods for improving the image quality in cardiac SPECT on both standard dose and reduced dose.First, we investigate a spatiotemporal post-processing approach based on a non-local means (NLM) filter for suppressing the noise in cardiac-gated SPECT images. Since in recent years low-dose studies have gained increased attention in cardiac SPECT owing to its potential radiation risk, to further improve the image quality on reduced dose, we investigate a novel de-noising method for low-dose cardiac-gated SPECT by using a three dimensional residual convolutional neural network (CNN). Furthermore, to reduce the negative effect of respiratory-binned acquisitions and assess the benefit of this approach in both standard dose and reduced dose using simulated acquisitions. Inspired by the success in respiratory correction, we investigate the potential benefit of cardiac motion correction for improving the detectability of perfusion defects. Finally, to combine the benefit of above two types of motion correction, dual-gated data acquisitions are implemented, wherein the acquired list-mode data are further binned into a number of intervals within cardiac and respiratory cycle according to the electrocardiography (ECG) signal and amplitude of the respiratory motion.
Show less
DEEP LEARNING IMAGE-DENOISING FOR IMPROVING DIAGNOSTIC ACCURACY IN CARDIAC SPECT
Title
DEEP LEARNING IMAGE-DENOISING FOR IMPROVING DIAGNOSTIC ACCURACY IN CARDIAC SPECT
Creator
Liu, Junchi
Date
2022
Description
Myocardial perfusion imaging (MPI) using single-photon emission computed tomography (SPECT) is a noninvasive imaging modality widely utilized...
Show moreMyocardial perfusion imaging (MPI) using single-photon emission computed tomography (SPECT) is a noninvasive imaging modality widely utilized for diagnosis of coronary artery diseases (CAD) in nuclear medicine. Because of the concern of potential radiation risks, the imaging dose administered to patients is limited in SPECT-MPI. Due to the low count statistics in acquired data, SPECT images can suffer from high levels of noise. In this study, we investigate the potential benefit of applying deep learning (DL) techniques for denoising in SPECT-MPI studies. Owing to the lack of ground truth in clinical studies, we adopt a noise-to-noise (N2N) training approach for denoising in full-dose studies. Afterwards, we investigate the benefit of applying N2N DL on reduced-dose studies to improve the detection accuracy of perfusion defects. To address the great variability in noise level among different subjects, we propose a scheme to account for the inter-subject variabilities in training a DL denoising network to improve its generalizability. In addition, we propose a dose-blind training approach for denoising at multiple reduced-dose levels. Moreover, we investigate several training schemes to address the issue that defect and non-defect image regions are highly unbalanced in a data set, where the overwhelming majority by non-defect regions tends to have a more pronounced contribution to the conventional loss function. We investigate whether these training schemes can effectively improve preservation of perfusion defects and yield better defect detection accuracy. In the experiments we demonstrated the proposed approaches with a set of 895 clinical acquisitions. The results show promising performance in denoising and improving the detectability of perfusion-defects with the proposed approaches.
Show less
Exploiting contextual information for deep learning based object detection
Title
Exploiting contextual information for deep learning based object detection
Creator
Zhang, Chen
Date
2020
Description
Object detection has long been an important research topic in computer vision area. It forms the basis of many applications. Despite the great...
Show moreObject detection has long been an important research topic in computer vision area. It forms the basis of many applications. Despite the great progress made in recent years, object detection is still a challenging task. One of the keys to improving the performance of object detection is to utilize the contextual information from the image itself or from a video sequence. Contextual information is defined as the interrelated condition in which something exists or occurs. In object detection, such interrelated condition can be related background/surroundings, support from image segmentation task, and the existence of the object in the temporal domain for video-based object detection. In this thesis, we propose multiple methods to exploit contextual information to improve the performance of object detection from images and videos.First, we focus on exploiting spatial contextual information in still-image based object detection, where each image is treated independently. Our research focuses on extracting contextual information using different approaches, which includes recurrent convolutional layer with feature concatenation (RCL-FC), 3-D recurrent neural networks (3-D RNN), and location-aware deformable convolution. Second, we focus on exploiting pixel-level contextual information from a related computer vision task, namely image segmentation. Our research focuses on applying a weakly-supervised auxiliary multi-label segmentation network to improve the performance of object detection without increasing the inference time. Finally, we focus on video object detection, where the temporal contextual information between video frames are exploited. Our first research involves modeling short-term temporal contextual information using optical flow and modeling long-term temporal contextual information using convLSTM. Another research focuses on creating a two-path convLSTM pyramid to handle multi-scale temporal contextual information for dealing with the change in object's scale. Our last work is the event-aware convLSTM that forces convLSTM to learn about the event that causes the performance to drop in a video sequence.
Show less
LOW-DOSE CARDIAC SPECT USING POST-FILTERING, DEEP LEARNING, AND MOTION CORRECTION
Title
LOW-DOSE CARDIAC SPECT USING POST-FILTERING, DEEP LEARNING, AND MOTION CORRECTION
Creator
Song, Chao
Date
2019
Description
Single photon emission computed tomography (SPECT) is an important technique in use today for the detection and evaluation of coronary artery...
Show moreSingle photon emission computed tomography (SPECT) is an important technique in use today for the detection and evaluation of coronary artery diseases. The image quality in cardiac SPECT can be adversely affected by cardiac motion and respiratory motion, both of which can lead to motion blur and non-uniform heart wall. In this thesis, we mainly investigate imaging de-noising algorithms and motion correction methods for improving the image quality in cardiac SPECT on both standard dose and reduced dose.First, we investigate a spatiotemporal post-processing approach based on a non-local means (NLM) filter for suppressing the noise in cardiac-gated SPECT images. Since in recent years low-dose studies have gained increased attention in cardiac SPECT owing to its potential radiation risk, to further improve the image quality on reduced dose, we investigate a novel de-noising method for low-dose cardiac-gated SPECT by using a three dimensional residual convolutional neural network (CNN). Furthermore, to reduce the negative effect of respiratory-binned acquisitions and assess the benefit of this approach in both standard dose and reduced dose using simulated acquisitions. Inspired by the success in respiratory correction, we investigate the potential benefit of cardiac motion correction for improving the detectability of perfusion defects. Finally, to combine the benefit of above two types of motion correction, dual-gated data acquisitions are implemented, wherein the acquired list-mode data are further binned into a number of intervals within cardiac and respiratory cycle according to the electrocardiography (ECG) signal and amplitude of the respiratory motion.
Show less
Voxel Transformer with Density-Aware Deformable Attention for 3D Object Detection
Title
Voxel Transformer with Density-Aware Deformable Attention for 3D Object Detection
Creator
Kim, Taeho
Date
2023
Description
The Voxel Transformer (VoTr) is a prominent model in the field of 3D object detection, employing a transformer-based architecture to...
Show moreThe Voxel Transformer (VoTr) is a prominent model in the field of 3D object detection, employing a transformer-based architecture to comprehend long-range voxel relationships through self-attention. However, despite its expanded receptive field, VoTr’s flexibility is constrained by its predefined receptive field. In this paper, we present a Voxel Transformer with Density-Aware Deformable Attention (VoTr-DADA), a novel approach to 3D object detection. VoTr-DADA leverages density-guided deformable attention for a more adaptable receptive field. It efficiently identifies key areas in the input using density features, combining the strengths of both VoTr and Deformable Attention. We introduce the Density-Aware Deformable Attention (DADA) module, which is specifically designed to focus on these crucial areas while adaptively extracting more informative features. Experimental results on the KITTI dataset and the Waymo Open dataset show that our proposed method outperforms the baseline VoTr model in 3D object detection while maintaining a fast inference speed.
Show less