arXiv subjects

Tuan Truong

Publications and source records attributed to Tuan Truong.

27 records · Page 2Linked to original sources

arXiv2024

Benchmarking Pretrained Vision Embeddings for Near- and Duplicate Detection in Medical Images

Near- and duplicate image detection is a critical concern in the field of medical imaging. Medical datasets often contain similar or duplicate images from various sources, which can lead to significant performance issues and evaluation biases, especially in machine learning tasks due to data leakage between training and testing subsets. In this paper, we present an approach for identifying near- and duplicate 3D medical images leveraging publicly available 2D computer vision embeddings. We assessed our approach by comparing embeddings extracted from two state-of-the-art self-supervised pretrained models and two different vector index structures for similarity retrieval. We generate an experimental benchmark based on the publicly available Medical Segmentation Decathlon dataset. The proposed method yields promising results for near- and duplicate image detection achieving a mean sensitivity and specificity of 0.9645 and 0.8559, respectively.

cs.CV↗

arXiv2023

Fused Audio Instance and Representation for Respiratory Disease Detection

Audio-based classification techniques on body sounds have long been studied to aid in the diagnosis of respiratory diseases. While most research is centered on the use of cough as the main biomarker, other body sounds also have the potential to detect respiratory diseases. Recent studies on COVID-19 have shown that breath and speech sounds, in addition to cough, correlate with the disease. Our study proposes Fused Audio Instance and Representation (FAIR) as a method for respiratory disease detection. FAIR relies on constructing a joint feature vector from various body sounds represented in waveform and spectrogram form. We conducted experiments on the use case of COVID-19 detection by combining waveform and spectrogram representation of body sounds. Our findings show that the use of self-attention to combine extracted features from cough, breath, and speech sounds leads to the best performance with an Area Under the Receiver Operating Characteristic Curve (AUC) score of 0.8658, a sensitivity of 0.8057, and a specificity of 0.7958. Compared to models trained solely on spectrograms or waveforms, the use of both representations results in an improved AUC score, demonstrating that combining spectrogram and waveform representation helps to enrich the extracted features and outperforms the models that use only one representation.

cs.SD↗

arXiv2023

Medical Image Retrieval Using Pretrained Embeddings

A wide range of imaging techniques and data formats available for medical images make accurate retrieval from image databases challenging. Efficient retrieval systems are crucial in advancing medical research, enabling large-scale studies and innovative diagnostic tools. Thus, addressing the challenges of medical image retrieval is essential for the continued enhancement of healthcare and research. In this study, we evaluated the feasibility of employing four state-of-the-art pretrained models for medical image retrieval at modality, body region, and organ levels and compared the results of two similarity indexing approaches. Since the employed networks take 2D images, we analyzed the impacts of weighting and sampling strategies to incorporate 3D information during retrieval of 3D volumes. We showed that medical image retrieval is feasible using pretrained networks without any additional training or fine-tuning steps. Using pretrained embeddings, we achieved a recall of 1 for various tasks at modality, body region, and organ level.

cs.CV↗

arXiv2023

ACA-Net: Towards Lightweight Speaker Verification using Asymmetric Cross Attention

In this paper, we propose ACA-Net, a lightweight, global context-aware speaker embedding extractor for Speaker Verification (SV) that improves upon existing work by using Asymmetric Cross Attention (ACA) to replace temporal pooling. ACA is able to distill large, variable-length sequences into small, fixed-sized latents by attending a small query to large key and value matrices. In ACA-Net, we build a Multi-Layer Aggregation (MLA) block using ACA to generate fixed-sized identity vectors from variable-length inputs. Through global attention, ACA-Net acts as an efficient global feature extractor that adapts to temporal variability unlike existing SV models that apply a fixed function for pooling over the temporal dimension which may obscure information about the signal's non-stationary temporal variability. Our experiments on the WSJ0-1talker show ACA-Net outperforms a strong baseline by 5\% relative improvement in EER using only 1/5 of the parameters.

cs.SD↗

arXiv2021

How Transferable Are Self-supervised Features in Medical Image Classification Tasks?

Transfer learning has become a standard practice to mitigate the lack of labeled data in medical classification tasks. Whereas finetuning a downstream task using supervised ImageNet pretrained features is straightforward and extensively investigated in many works, there is little study on the usefulness of self-supervised pretraining. In this paper, we assess the transferability of ImageNet self-supervisedpretraining by evaluating the performance of models initialized with pretrained features from three self-supervised techniques (SimCLR, SwAV, and DINO) on selected medical classification tasks. The chosen tasks cover tumor detection in sentinel axillary lymph node images, diabetic retinopathy classification in fundus images, and multiple pathological condition classification in chest X-ray images. We demonstrate that self-supervised pretrained models yield richer embeddings than their supervised counterpart, which benefits downstream tasks in view of both linear evaluation and finetuning. For example, in view of linear evaluation at acritically small subset of the data, we see an improvement up to 14.79% in Kappa score in the diabetic retinopathy classification task, 5.4% in AUC in the tumor classification task, 7.03% AUC in the pneumonia detection, and 9.4% in AUC in the detection of pathological conditions in chest X-ray. In addition, we introduce Dynamic Visual Meta-Embedding (DVME) as an end-to-end transfer learning approach that fuses pretrained embeddings from multiple models. We show that the collective representation obtained by DVME leads to a significant improvement in the performance of selected tasks compared to using a single pretrained model approach and can be generalized to any combination of pretrained models.

cs.CV↗

arXiv2015

Know The Star, Know the Planet. IV. A Stellar Companion to the Host star of the Eccentric Exoplanet HD 8673b

HD 8673 hosts a massive exoplanet in a highly eccentric orbit (e=0.723). Based on two epochs of speckle interferometry a previous publication identified a candidate stellar companion. We observed HD 8673 multiple times with the 10 m Keck II telescope, the 5 m Hale telescope, the 3.63 m AEOS telescope and the 1.5m Palomar telescope in a variety of filters with the aim of confirming and characterizing the stellar companion. We did not detect the candidate companion, which we now conclude was a false detection, but we did detect a fainter companion. We collected astrometry and photometry of the companion on six epochs in a variety of filters. The measured differential photometry enabled us to determine that the companion is an early M dwarf with a mass estimate of 0.33-0.45 M?. The companion has a projected separation of 10 AU, which is one of the smallest projected separations of an exoplanet host binary system. Based on the limited astrometry collected, we are able to constrain the orbit of the stellar companion to a semi-major axis of 35{60 AU, an eccentricity ? 0.5 and an inclination of 75{85?. The stellar companion has likely strongly in uenced the orbit of the exoplanet and quite possibly explains its high eccentricity.

astro-ph.SR↗

arXiv2014

PULSE: The Palomar Ultraviolet Laser for the Study of Exoplanets

The Palomar Ultraviolet Laser for the Study of Exoplanets (PULSE) will dramatically expand the science reach of PALM-3000, the facility high-contrast extreme adaptive optics system on the 5-meter Hale Telescope. By using an ultraviolet laser to measure the dominant high spatial and temporal order turbulence near the telescope aperture, one can increase the limiting natural guide star magnitude for exquisite correction from mV < 10 to mV < 16. Providing the highest near-infrared Strehl ratios from any large telescope laser adaptive optics system, PULSE uniquely enables spectroscopy of low-mass and more distant young exoplanet systems, essential to formulating a complete picture of exoplanet populations.

astro-ph.IM↗

arXiv2013

Electric Field Conjugation with the Project 1640 coronagraph

The Project 1640 instrument on the 200-inch Hale telescope at Palomar Observatory is a coronagraphic instrument with an integral field spectrograph at the back end, designed to find young, self-luminous planets around nearby stars. To reach the necessary contrast for this, the PALM-3000 adaptive optics system corrects for fast atmospheric speckles, while CAL, a phase-shifting interferometer in a Mach-Zehnder configuration, measures the quasistatic components of the complex electric field in the pupil plane following the coronagraphic stop. Two additional sensors measure and control low-order modes. These field measurements may then be combined with a system model and data taken separately using a white-light source internal to the AO system to correct for both phase and amplitude aberrations. Here, we discuss and demonstrate the procedure to maintain a half-plane dark hole in the image plane while the spectrograph is taking data, including initial on-sky performance.

astro-ph.IM↗

arXiv2010

A Close Companion Search around L Dwarfs using Aperture Masking Interferometry and Palomar Laser Guide Star Adaptive Optics

We present a close companion search around sixteen known early-L dwarfs using aperture masking interferometry with Palomar laser guide star adaptive optics. The use of aperture masking allows the detection of close binaries, corresponding to projected physical separations of 0.6-10.0 AU for the targets of our survey. This survey achieved median contrast limits of Delta_K ~ 2.3 for separations between 1.2 - 4 lambda/D, and Delta_K ~ 1.4 at (2/3)lambda/D. We present four candidate binaries detected with moderate to high confidence (90-98%). Two have projected physical separations less than 1.5 AU. This may indicate that tight-separation binaries contribute more significantly to the binary fraction than currently assumed, consistent with spectroscopic and photometric overluminosity studies. Ten targets of this survey have previously been observed with the Hubble Space Telescope as part of companion searches. We use the increased resolution of aperture masking to search for close or dim companions that would be obscured by full aperture imaging, finding two candidate binaries. This survey is the first application of aperture masking with laser guide star adaptive optics at Palomar. Several new techniques for the analysis of aperture masking data in the low signal to noise regime are explored.

astro-ph.SR↗