-
Exploiting Richness of Learned Compressed Representation of Images for Semantic Segmentation
Authors:
Ravi Kakaiya,
Rakshith Sathish,
Ramanathan Sethuraman,
Debdoot Sheet
Abstract:
Autonomous vehicles and Advanced Driving Assistance Systems (ADAS) have the potential to radically change the way we travel. Many such vehicles currently rely on segmentation and object detection algorithms to detect and track objects around its surrounding. The data collected from the vehicles are often sent to cloud servers to facilitate continual/life-long learning of these algorithms. Consider…
▽ More
Autonomous vehicles and Advanced Driving Assistance Systems (ADAS) have the potential to radically change the way we travel. Many such vehicles currently rely on segmentation and object detection algorithms to detect and track objects around its surrounding. The data collected from the vehicles are often sent to cloud servers to facilitate continual/life-long learning of these algorithms. Considering the bandwidth constraints, the data is compressed before sending it to servers, where it is typically decompressed for training and analysis. In this work, we propose the use of a learning-based compression Codec to reduce the overhead in latency incurred for the decompression operation in the standard pipeline. We demonstrate that the learned compressed representation can also be used to perform tasks like semantic segmentation in addition to decompression to obtain the images. We experimentally validate the proposed pipeline on the Cityscapes dataset, where we achieve a compression factor up to $66 \times$ while preserving the information required to perform segmentation with a dice coefficient of $0.84$ as compared to $0.88$ achieved using decompressed images while reducing the overall compute by $11\%$.
△ Less
Submitted 27 July, 2023; v1 submitted 4 July, 2023;
originally announced July 2023.
-
Interaction of Shock Train with Cavity Shear Layer in a Scramjet Isolator
Authors:
Vignesh Ram Petha Sethuraman,
Yosheph Yang,
Jae Gang Kim
Abstract:
The interaction between the self-excited shock train flow and the cavity shear layer in a scramjet isolator is investigated numerically using detached-eddy simulations (DES). The effect of changing the position of the shock train by controlling the back pressure ratio and the effect of changing the cavity front wall angle are analyzed using unsteady statistics and modal analysis. The propagation m…
▽ More
The interaction between the self-excited shock train flow and the cavity shear layer in a scramjet isolator is investigated numerically using detached-eddy simulations (DES). The effect of changing the position of the shock train by controlling the back pressure ratio and the effect of changing the cavity front wall angle are analyzed using unsteady statistics and modal analysis. The propagation mechanism of the pressure disturbance was investigated by spatiotemporal cross-correlation coefficient analysis. In the present numerical study, a constant isolator section with a cavity front wall was considered, followed by a diffuser section simulated at Mach number 2.2 with three different back pressure ratios. The change in back pressure provides three different conditions. To understand the unsteady dynamics of the interaction of the shear layer with the shock train, the spatiotemporal trajectory of the wall pressure and the centerline pressure distribution, the spatiotemporal cross-correlation coefficient, and the modal analysis by dynamic mode decomposition are obtained. The results show that the low-frequency shock train oscillation dominates the cavity oscillation. The spatiotemporal cross-correlation between the wall surface and the cavity bottom wall indicates the propagation of local disturbances originating from the separated boundary layer caused by the shock and the recirculation zone in the corners of the cavity. Dynamic mode decomposition analysis shows the shear layer at the leading edge of the cavity and the downstream propagation of large eddies from the cavity. It also shows the pairing of coherent structures between the shock train and the recirculation zone of the cavity.
△ Less
Submitted 3 December, 2022;
originally announced December 2022.
-
Design of SCALES: A 2-5 Micron Coronagraphic Integral Field Spectrograph for Keck Observatory
Authors:
Andrew Skemer,
R. Deno Stelter,
Stephanie Sallum,
Nicholas MacDonald,
Renate Kupke,
Christopher Ratliffe,
Ravinder Banyal,
Amirul Hasan,
Hari Mohan Varshney,
Arun Surya,
A** Prakash,
Sivarani Thirupathi,
Ramya Sethuraman,
Govinda K. V.,
Michael P. Fitzgerald,
Eric Wang,
Marc Kassis,
Olivier Absil,
Carlos Alvarez,
Natasha Batalha,
Marc-Andre Boucher,
Cyril Bourgenot,
Timothy Brandt,
Zackery Briesemeister,
Katherine de Kleer
, et al. (27 additional authors not shown)
Abstract:
We present the design of SCALES (Slicer Combined with Array of Lenslets for Exoplanet Spectroscopy) a new 2-5 micron coronagraphic integral field spectrograph under construction for Keck Observatory. SCALES enables low-resolution (R~50) spectroscopy, as well as medium-resolution (R~4,000) spectroscopy with the goal of discovering and characterizing cold exoplanets that are brightest in the thermal…
▽ More
We present the design of SCALES (Slicer Combined with Array of Lenslets for Exoplanet Spectroscopy) a new 2-5 micron coronagraphic integral field spectrograph under construction for Keck Observatory. SCALES enables low-resolution (R~50) spectroscopy, as well as medium-resolution (R~4,000) spectroscopy with the goal of discovering and characterizing cold exoplanets that are brightest in the thermal infrared. Additionally, SCALES has a 12x12" field-of-view imager that will be used for general adaptive optics science at Keck. We present SCALES's specifications, its science case, its overall design, and simulations of its expected performance. Additionally, we present progress on procuring, fabricating and testing long lead-time components.
△ Less
Submitted 23 August, 2022;
originally announced August 2022.
-
Lung Segmentation and Nodule Detection in Computed Tomography Scan using a Convolutional Neural Network Trained Adversarially using Turing Test Loss
Authors:
Rakshith Sathish,
Rachana Sathish,
Ramanathan Sethuraman,
Debdoot Sheet
Abstract:
Lung cancer is the most common form of cancer found worldwide with a high mortality rate. Early detection of pulmonary nodules by screening with a low-dose computed tomography (CT) scan is crucial for its effective clinical management. Nodules which are symptomatic of malignancy occupy about 0.0125 - 0.025\% of volume in a CT scan of a patient. Manual screening of all slices is a tedious task and…
▽ More
Lung cancer is the most common form of cancer found worldwide with a high mortality rate. Early detection of pulmonary nodules by screening with a low-dose computed tomography (CT) scan is crucial for its effective clinical management. Nodules which are symptomatic of malignancy occupy about 0.0125 - 0.025\% of volume in a CT scan of a patient. Manual screening of all slices is a tedious task and presents a high risk of human errors. To tackle this problem we propose a computationally efficient two stage framework. In the first stage, a convolutional neural network (CNN) trained adversarially using Turing test loss segments the lung region. In the second stage, patches sampled from the segmented region are then classified to detect the presence of nodules. The proposed method is experimentally validated on the LUNA16 challenge dataset with a dice coefficient of $0.984\pm0.0007$ for 10-fold cross-validation.
△ Less
Submitted 16 June, 2020;
originally announced June 2020.
-
A Two-Stage Multiple Instance Learning Framework for the Detection of Breast Cancer in Mammograms
Authors:
Sarath Chandra K,
Arunava Chakravarty,
Nirmalya Ghosh,
Tandra Sarkar,
Ramanathan Sethuraman,
Debdoot Sheet
Abstract:
Mammograms are commonly employed in the large scale screening of breast cancer which is primarily characterized by the presence of malignant masses. However, automated image-level detection of malignancy is a challenging task given the small size of the mass regions and difficulty in discriminating between malignant, benign mass and healthy dense fibro-glandular tissue. To address these issues, we…
▽ More
Mammograms are commonly employed in the large scale screening of breast cancer which is primarily characterized by the presence of malignant masses. However, automated image-level detection of malignancy is a challenging task given the small size of the mass regions and difficulty in discriminating between malignant, benign mass and healthy dense fibro-glandular tissue. To address these issues, we explore a two-stage Multiple Instance Learning (MIL) framework. A Convolutional Neural Network (CNN) is trained in the first stage to extract local candidate patches in the mammograms that may contain either a benign or malignant mass. The second stage employs a MIL strategy for an image level benign vs. malignant classification. A global image-level feature is computed as a weighted average of patch-level features learned using a CNN. Our method performed well on the task of localization of masses with an average Precision/Recall of 0.76/0.80 and acheived an average AUC of 0.91 on the imagelevel classification task using a five-fold cross-validation on the INbreast dataset. Restricting the MIL only to the candidate patches extracted in Stage 1 led to a significant improvement in classification performance in comparison to a dense extraction of patches from the entire mammogram.
△ Less
Submitted 24 April, 2020;
originally announced April 2020.
-
Learning Decision Ensemble using a Graph Neural Network for Comorbidity Aware Chest Radiograph Screening
Authors:
Arunava Chakravarty,
Tandra Sarkar,
Nirmalya Ghosh,
Ramanathan Sethuraman,
Debdoot Sheet
Abstract:
Chest radiographs are primarily employed for the screening of cardio, thoracic and pulmonary conditions. Machine learning based automated solutions are being developed to reduce the burden of routine screening on Radiologists, allowing them to focus on critical cases. While recent efforts demonstrate the use of ensemble of deep convolutional neural networks(CNN), they do not take disease comorbidi…
▽ More
Chest radiographs are primarily employed for the screening of cardio, thoracic and pulmonary conditions. Machine learning based automated solutions are being developed to reduce the burden of routine screening on Radiologists, allowing them to focus on critical cases. While recent efforts demonstrate the use of ensemble of deep convolutional neural networks(CNN), they do not take disease comorbidity into consideration, thus lowering their screening performance. To address this issue, we propose a Graph Neural Network (GNN) based solution to obtain ensemble predictions which models the dependencies between different diseases. A comprehensive evaluation of the proposed method demonstrated its potential by improving the performance over standard ensembling technique across a wide range of ensemble constructions. The best performance was achieved using the GNN ensemble of DenseNet121 with an average AUC of 0.821 across thirteen disease comorbidities.
△ Less
Submitted 24 April, 2020;
originally announced April 2020.
-
A Systematic Search over Deep Convolutional Neural Network Architectures for Screening Chest Radiographs
Authors:
Arka Mitra,
Arunava Chakravarty,
Nirmalya Ghosh,
Tandra Sarkar,
Ramanathan Sethuraman,
Debdoot Sheet
Abstract:
Chest radiographs are primarily employed for the screening of pulmonary and cardio-/thoracic conditions. Being undertaken at primary healthcare centers, they require the presence of an on-premise reporting Radiologist, which is a challenge in low and middle income countries. This has inspired the development of machine learning based automation of the screening process. While recent efforts demons…
▽ More
Chest radiographs are primarily employed for the screening of pulmonary and cardio-/thoracic conditions. Being undertaken at primary healthcare centers, they require the presence of an on-premise reporting Radiologist, which is a challenge in low and middle income countries. This has inspired the development of machine learning based automation of the screening process. While recent efforts demonstrate a performance benchmark using an ensemble of deep convolutional neural networks (CNN), our systematic search over multiple standard CNN architectures identified single candidate CNN models whose classification performances were found to be at par with ensembles. Over 63 experiments spanning 400 hours, executed on a 11:3 FP32 TensorTFLOPS compute system, we found the Xception and ResNet-18 architectures to be consistent performers in identifying co-existing disease conditions with an average AUC of 0.87 across nine pathologies. We conclude on the reliability of the models by assessing their saliency maps generated using the randomized input sampling for explanation (RISE) method and qualitatively validating them against manual annotations locally sourced from an experienced Radiologist. We also draw a critical note on the limitations of the publicly available CheXpert dataset primarily on account of disparity in class distribution in training vs. testing sets, and unavailability of sufficient samples for few classes, which hampers quantitative reporting due to sample insufficiency.
△ Less
Submitted 24 April, 2020;
originally announced April 2020.
-
Fully Convolutional Model for Variable Bit Length and Lossy High Density Compression of Mammograms
Authors:
Aupendu Kar,
Sri Phani Krishna Karri,
Nirmalya Ghosh,
Ramanathan Sethuraman,
Debdoot Sheet
Abstract:
Early works on medical image compression date to the 1980's with the impetus on deployment of teleradiology systems for high-resolution digital X-ray detectors. Commercially deployed systems during the period could compress 4,096 x 4,096 sized images at 12 bpp to 2 bpp using lossless arithmetic coding, and over the years JPEG and JPEG2000 were imbibed reaching upto 0.1 bpp. Inspired by the reprise…
▽ More
Early works on medical image compression date to the 1980's with the impetus on deployment of teleradiology systems for high-resolution digital X-ray detectors. Commercially deployed systems during the period could compress 4,096 x 4,096 sized images at 12 bpp to 2 bpp using lossless arithmetic coding, and over the years JPEG and JPEG2000 were imbibed reaching upto 0.1 bpp. Inspired by the reprise of deep learning based compression for natural images over the last two years, we propose a fully convolutional autoencoder for diagnostically relevant feature preserving lossy compression. This is followed by leveraging arithmetic coding for encapsulating high redundancy of features for further high-density code packing leading to variable bit length. We demonstrate performance on two different publicly available digital mammography datasets using peak signal-to-noise ratio (pSNR), structural similarity (SSIM) index and domain adaptability tests between datasets. At high density compression factors of >300x (~0.04 bpp), our approach rivals JPEG and JPEG2000 as evaluated through a Radiologist's visual Turing test.
△ Less
Submitted 17 May, 2018;
originally announced May 2018.
