scholarly journals Predictive and discriminative localization of pathology using high resolution class activation maps with CNNs

2021 ◽  
Vol 7 ◽  
pp. e622
Author(s):  
Sumeet Shinde ◽  
Priyanka Tupe-Waghmare ◽  
Tanay Chougule ◽  
Jitender Saini ◽  
Madhura Ingalhalikar
Keyword(s):  
High Resolution ◽  
Bone Fractures ◽  
Single Layer ◽  
Image Resolution ◽  
Neural Net ◽  
Original Image ◽  
Feature Maps ◽  
Final Layer ◽  
Small Cohort ◽  
Activation Mapping

Purpose Existing class activation mapping (CAM) techniques extract the feature maps only from a single layer of the convolutional neural net (CNN), generally from the final layer and then interpolate to upsample to the original image resolution to locate the discriminative regions. Consequently these provide a coarse localization that may not be able to capture subtle abnormalities in medical images. To alleviate this, our work proposes a technique called high resolution class activation mapping (HR-CAMs) that can provide enhanced visual explainability to the CNN models. Methods HR-CAMs fuse feature maps by training a network using the input from multiple layers of a trained CNN, thus gaining information from every layer that can localize abnormalities with greater details in original image resolution. The technique is validated qualitatively and quantitatively on a simulated dataset of 8,000 images followed by applications on multiple image analysis tasks that include (1) skin lesion classification (ISIC open dataset—25,331 cases) and (2) predicting bone fractures (MURA open dataset—40,561 images) (3) predicting Parkinson’s disease (PD) from neuromelanin sensitive MRI (small cohort-80 subjects). Results We demonstrate that our model creates clinically interpretable subject specific high resolution discriminative localizations when compared to widely used CAMs and Gradient-CAMs. Conclusion HR-CAMs provide finer delineation of abnormalities thus facilitating superior explainability to CNNs as has been demonstrated from its rigorous validation.

scholarly journals Instance Segmentation for Governmental Inspection of Small Touristic Infrastructure in Beach Zones Using Multispectral High-Resolution WorldView-3 Imagery

2021 ◽  
Vol 10 (12) ◽  
pp. 813
Author(s):  
Osmar Luiz Ferreira de Carvalho ◽  
Rebeca dos Santos de Moura ◽  
Anesmar Olino de Albuquerque ◽  
Pablo Pozzobon de Bem ◽  
Rubens de Castro Pereira ◽  
...  
Keyword(s):  
High Resolution ◽  
Object Detection ◽  
Sliding Window ◽  
Coastal Areas ◽  
Image Resolution ◽  
Small Object ◽  
Original Image ◽  
Individual Object ◽  
High Resolution Images ◽  
Instance Segmentation

Misappropriation of public lands is an ongoing government concern. In Brazil, the beach zone is public property, but many private establishments use it for economic purposes, requiring constant inspection. Among the undue targets, the individual mapping of straw beach umbrellas (SBUs) attached to the sand is a great challenge due to their small size, high presence, and agglutinated appearance. This study aims to automatically detect and count SBUs on public beaches using high-resolution images and instance segmentation, obtaining pixel-wise semantic information and individual object detection. This study is the first instance segmentation application on coastal areas and the first using WorldView-3 (WV-3) images. We used the Mask-RCNN with some modifications: (a) multispectral input for the WorldView3 imagery (eight channels), (b) improved the sliding window algorithm for large image classification, and (c) comparison of different image resizing ratios to improve small object detection since the SBUs are small objects (<322 pixels) even using high-resolution images (31 cm). The accuracy analysis used standard COCO metrics considering the original image and three scale ratios (2×, 4×, and 8× resolution increase). The average precision (AP) results increased proportionally to the image resolution: 30.49% (original image), 48.24% (2×), 53.45% (4×), and 58.11% (8×). The 8× model presented 94% AP50, classifying nearly all SBUs correctly. Moreover, the improved sliding window approach enables the classification of large areas providing automatic counting and estimating the size of the objects, proving to be effective for inspecting large coastal areas and providing insightful information for public managers. This remote sensing application impacts the inspection cost, tribute, and environmental conditions.

Molecular Scale Imaging with a Multilayer Superlens

2006 ◽  
Vol 919 ◽  
Author(s):  
Pratik Chaturvedi ◽  
Nicholas X. Fang
Keyword(s):  
High Resolution ◽  
Single Layer ◽  
Diffraction Limit ◽  
Image Resolution ◽  
Optical Metrology ◽  
Optimized Design ◽  
Low Loss ◽  
Molecular Scale ◽  
Illumination Wavelength ◽  
Better Than

AbstractIt has been experimentally demonstrated that a single layer of silver functions as a “superlens” [Fang et al, Science 308, 534 (2005)], providing image resolution much better than the diffraction limit. Resolution as high as 60 nanometer (λ/6) half-pitch was achieved. In this paper, we explore the possibility of further refining the image resolution using a “multilayer superlens” design. With optimized design of silver-alumina multilayer superlens, our numerical simulations show a feasibility of resolving 15nm features, about 1/26th of the illumination wavelength. We present preliminary experimental results targeted towards achieving the molecular scale imaging resolution. The development of potential low-loss and high resolution superlens opens the door to exciting applications in nanoscale optical metrology and nanomanufacturing.

Towards 1-Ångstrom-resolution STEM

1993 ◽  
Vol 51 ◽  
pp. 996-997
Author(s):  
H.S. von Harrach ◽  
D.E. Jesson ◽  
S.J. Pennycook
Keyword(s):  
High Resolution ◽  
Field Emission ◽  
Phase Contrast ◽  
Spherical Aberration ◽  
A Priori ◽  
Dark Field ◽  
Image Resolution ◽  
Objective Lens ◽  
Annular Dark Field ◽  
First Results

Phase contrast TEM has been the leading technique for high resolution imaging of materials for many years, whilst STEM has been the principal method for high-resolution microanalysis. However, it was demonstrated many years ago that low angle dark-field STEM imaging is a priori capable of almost 50% higher point resolution than coherent bright-field imaging (i.e. phase contrast TEM or STEM). This advantage was not exploited until Pennycook developed the high-angle annular dark-field (ADF) technique which can provide an incoherent image showing both high image resolution and atomic number contrast.This paper describes the design and first results of a 300kV field-emission STEM (VG Microscopes HB603U) which has improved ADF STEM image resolution towards the 1 angstrom target. The instrument uses a cold field-emission gun, generating a 300 kV beam of up to 1 μA from an 11-stage accelerator. The beam is focussed on to the specimen by two condensers and a condenser-objective lens with a spherical aberration coefficient of 1.0 mm.

scholarly journals RE-APPRAISAL OF THE CAVOTRICUSPID REGION AS A FUNCTIONAL ISTHMUS DURING ATRIAL FLUTTER: A HIGH RESOLUTION ACTIVATION MAPPING STUDY

2021 ◽  
Vol 77 (18) ◽  
pp. 429
Author(s):  
Swati Rao ◽  
Agatha Kwasnik ◽  
Hemal Nayak ◽  
Zaid Aziz ◽  
Gaurav Upadhyay ◽  
...  
Keyword(s):  
High Resolution ◽  
Atrial Flutter ◽  
Mapping Study ◽  
Activation Mapping

scholarly journals High-Resolution SAR Image Classification Using Multi-Scale Deep Feature Fusion and Covariance Pooling Manifold Network

2021 ◽  
Vol 13 (2) ◽  
pp. 328
Author(s):  
Wenkai Liang ◽  
Yan Wu ◽  
Ming Li ◽  
Yice Cao ◽  
Xin Hu
Keyword(s):  
High Resolution ◽  
Image Classification ◽  
Feature Fusion ◽  
Representation Learning ◽  
Sar Image ◽  
Gabor Filtering ◽  
Feature Maps ◽  
Sar Images ◽  
Multi Scale ◽  
Deep Feature

The classification of high-resolution (HR) synthetic aperture radar (SAR) images is of great importance for SAR scene interpretation and application. However, the presence of intricate spatial structural patterns and complex statistical nature makes SAR image classification a challenging task, especially in the case of limited labeled SAR data. This paper proposes a novel HR SAR image classification method, using a multi-scale deep feature fusion network and covariance pooling manifold network (MFFN-CPMN). MFFN-CPMN combines the advantages of local spatial features and global statistical properties and considers the multi-feature information fusion of SAR images in representation learning. First, we propose a Gabor-filtering-based multi-scale feature fusion network (MFFN) to capture the spatial pattern and get the discriminative features of SAR images. The MFFN belongs to a deep convolutional neural network (CNN). To make full use of a large amount of unlabeled data, the weights of each layer of MFFN are optimized by unsupervised denoising dual-sparse encoder. Moreover, the feature fusion strategy in MFFN can effectively exploit the complementary information between different levels and different scales. Second, we utilize a covariance pooling manifold network to extract further the global second-order statistics of SAR images over the fusional feature maps. Finally, the obtained covariance descriptor is more distinct for various land covers. Experimental results on four HR SAR images demonstrate the effectiveness of the proposed method and achieve promising results over other related algorithms.

New LWD Image Improvement Method to Mitigate Interpretation Risks

2021 ◽  
Author(s):  
Tianhua Zhang ◽  
Shiduo Yang ◽  
Chandramani Shrivastava ◽  
Adrian A ◽  
Nadege Bize-Forest
Keyword(s):  
High Resolution ◽  
Real Time ◽  
Best Practice ◽  
Synthetic Data ◽  
Image Resolution ◽  
Drilling Process ◽  
Drilling Tool ◽  
Stick Slip ◽  
Geological Interpretation ◽  
Wellbore Temperature

Abstract With the advancement of LWD (Logging While Drilling) hardware and acquisition, the imaging technology becomes not only an indispensable part of the drilling tool string, but also the image resolution increases to map layers and heterogeneity features down to less than 5mm scale. This shortens the geological interpretation turn-around time from wireline logging time (hours to days after drilling) to semi-real time (drilling time or hours after drilling). At the same time, drilling motion is complex. The depth tracking is on the surface referenced to the surface block movement. The imaging sensor located downhole can be thousands of feet away from the surface. Mechanical torque and drag, wellbore friction, wellbore temperature and weight on bit can make the downhole sensor movement motion not synchronized with surface pipe depth. This will cause time- depth conversion step generate image artifacts that either stop real-time interpretation of geological features or mis-interpret features on high resolution images. In this paper, we present several LWD images featuring distortion mechanism during the drilling process using synthetic data. We investigated how heave, depth reset and downhole sensor stick/slip caused image distortions. We provide solutions based on downhole sensor pseudo velocity computation to minimize the image distortion. The best practice in using Savitsky-Golay filter are presented in the discussion sections. Finally, some high-resolution LWD images distorted with drilling-related artifacts and processed ones are shown to demonstrate the importance of image post-processing. With the proper processed images, we can minimize interpretation risks and make drilling decisions with more confidence.

scholarly journals Identifying Atoms in High Resolution Transmission Electron Micrographs Using a Deep Convolutional Neural Net

2018 ◽  
Vol 24 (S1) ◽  
pp. 512-513 ◽  
Author(s):  
Jakob Schiøtz ◽  
Jacob Madsen ◽  
Pei Liu ◽  
Ole Winther ◽  
Jens Kling ◽  
...  
Keyword(s):  
High Resolution ◽  
Electron Micrographs ◽  
Transmission Electron Micrographs ◽  
Neural Net ◽  
Transmission Electron

scholarly journals Building Extraction in Very High Resolution Imagery by Dense-Attention Networks

2018 ◽  
Vol 10 (11) ◽  
pp. 1768 ◽  
Author(s):  
Hui Yang ◽  
Penghai Wu ◽  
Xuedong Yao ◽  
Yanlan Wu ◽  
Biao Wang ◽  
...  
Keyword(s):  
Deep Learning ◽  
High Resolution ◽  
Building Extraction ◽  
Learning Networks ◽  
Feature Maps ◽  
Low Level ◽  
High Resolution Imagery ◽  
Very High Resolution Imagery ◽  
High Level ◽  
Very High

Building extraction from very high resolution (VHR) imagery plays an important role in urban planning, disaster management, navigation, updating geographic databases, and several other geospatial applications. Compared with the traditional building extraction approaches, deep learning networks have recently shown outstanding performance in this task by using both high-level and low-level feature maps. However, it is difficult to utilize different level features rationally with the present deep learning networks. To tackle this problem, a novel network based on DenseNets and the attention mechanism was proposed, called the dense-attention network (DAN). The DAN contains an encoder part and a decoder part which are separately composed of lightweight DenseNets and a spatial attention fusion module. The proposed encoder–decoder architecture can strengthen feature propagation and effectively bring higher-level feature information to suppress the low-level feature and noises. Experimental results based on public international society for photogrammetry and remote sensing (ISPRS) datasets with only red–green–blue (RGB) images demonstrated that the proposed DAN achieved a higher score (96.16% overall accuracy (OA), 92.56% F1 score, 90.56% mean intersection over union (MIOU), less training and response time and higher-quality value) when compared with other deep learning methods.

scholarly journals The Effect of Image Resolution on Automated Classification of Chest X-rays

2021 ◽  
Author(s):  
Md Inzamam Ul Haque ◽  
Abhishek K Dubey ◽  
Jacob D Hinkle
Keyword(s):  
High Resolution ◽  
Automated Analysis ◽  
Image Resolution ◽  
Free Text ◽  
X Rays ◽  
Training Time ◽  
X Ray ◽  
Radiology Reports ◽  
Chest X Ray ◽  
High Resolution Images

Deep learning models have received much attention lately for their ability to achieve expert-level performance on the accurate automated analysis of chest X-rays. Although publicly available chest X-ray datasets include high resolution images, most models are trained on reduced size images due to limitations on GPU memory and training time. As compute capability continues to advance, it will become feasible to train large convolutional neural networks on high-resolution images. This study is based on the publicly available MIMIC-CXR-JPG dataset, comprising 377,110 high resolution chest X-ray images, and provided with 14 labels to the corresponding free-text radiology reports. We find, interestingly, that tasks that require a large receptive field are better suited to downscaled input images, and we verify this qualitatively by inspecting effective receptive fields and class activation maps of trained models. Finally, we show that stacking an ensemble across resolutions outperforms each individual learner at all input resolutions while providing interpretable scale weights, suggesting that multi-scale features are crucially important to information extraction from high-resolution chest X-rays.

scholarly journals Mastcam Image Resolution Enhancement with Application to Disparity Map Generation for Stereo Images with Different Resolutions

Sensors ◽  
2019 ◽  
Vol 19 (16) ◽  
pp. 3526 ◽  
Author(s):  
Ayhan ◽  
Kwan
Keyword(s):  
Deep Learning ◽  
High Resolution ◽  
Image Enhancement ◽  
Image Resolution ◽  
Map Estimation ◽  
Stereo Images ◽  
Disparity Map ◽  
Camera Image ◽  
The Right ◽  
The Impact

In this paper, we introduce an in-depth application of high-resolution disparity map estimation using stereo images from Mars Curiosity rover’s Mastcams, which have two imagers with different resolutions. The left Mastcam has three times lower resolution as that of the right. The left Mastcam image’s resolution is first enhanced with three methods: Bicubic interpolation, pansharpening-based method, and a deep learning super resolution method. The enhanced left camera image and the right camera image are then used to estimate the disparity map. The impact of the left camera image enhancement is examined. The comparative performance analyses showed that the left camera enhancement results in getting more accurate disparity maps in comparison to using the original left Mastcam images for disparity map estimation. The deep learning-based method provided the best performance among the three for both image enhancement and disparity map estimation accuracy. A high-resolution disparity map, which is the result of the left camera image enhancement, is anticipated to improve the conducted science products in the Mastcam imagery such as 3D scene reconstructions, depth maps, and anaglyph images.

Sign in / Sign up

Export Citation Format

Share Document