There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality.

Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

I and i

[신간의 별자리x] 우리/미술, 그리고 ‘슬픔의 박물관’

From the Publisher: 
The accessible presentation of this book gives both a general view of the entire computer vision enterprise and also offers sufficient detail to be able to build useful applications. Users learn techniques that have proven to be useful by first-hand experience and a wide range of mathematical methods. A CD-ROM with every copy of the text contains source code for programming practice, color images, and illustrative movies. Comprehensive and up-to-date, this book includes essential topics that either reflect practical significance or are of theoretical importance. Topics are discussed in substantial and increasing depth. Application surveys describe numerous important application areas such as image based rendering and digital libraries. Many important algorithms broken down and illustrated in pseudo code. Appropriate for use by engineers as a comprehensive reference to the computer vision enterprise.

Computer vision : a modern approach = 计算机视觉 : 一种现代的方法

Biomedical imaging is a driver of scientific discovery and a core component of medical care and is being stimulated by the field of deep learning. While semantic segmentation algorithms enable image analysis and quantification in many applications, the design of respective specialized solutions is non-trivial and highly dependent on dataset properties and hardware conditions. We developed nnU-Net, a deep learning-based segmentation method that automatically configures itself, including preprocessing, network architecture, training and post-processing for any new task. The key design choices in this process are modeled as a set of fixed parameters, interdependent rules and empirical decisions. Without manual intervention, nnU-Net surpasses most existing approaches, including highly specialized solutions on 23 public datasets used in international biomedical segmentation competitions. We make nnU-Net publicly available as an out-of-the-box tool, rendering state-of-the-art segmentation accessible to a broad audience by requiring neither expert knowledge nor computing resources beyond standard network training.

nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation

The volume and complexity of diagnostic imaging is increasing at a pace faster than the availability of human expertise to interpret it. Artificial intelligence has shown great promise in classifying two-dimensional photographs of some common diseases and typically relies on databases of millions of annotated images. Until now, the challenge of reaching the performance of expert clinicians in a real-world clinical pathway with three-dimensional diagnostic scans has remained unsolved. Here, we apply a novel deep learning architecture to a clinically heterogeneous set of three-dimensional optical coherence tomography scans from patients referred to a major eye hospital. We demonstrate performance in making a referral recommendation that reaches or exceeds that of experts on a range of sight-threatening retinal diseases after training on only 14,884 scans. Moreover, we demonstrate that the tissue segmentations produced by our architecture act as a device-independent representation; referral accuracy is maintained when using tissue segmentations from a different type of device. Our work removes previous barriers to wider clinical use without prohibitive training data requirements across multiple pathologies in a real-world setting.

/pdf/clinically-applicable-deep-learning-for-diagnosis-and-56r182cw0z.pdf

Clinically applicable deep learning for diagnosis and referral in retinal disease

Radio frequency identification and detection (RFID) technology is a wireless communication technology that enables users to uniquely identify tagged objects or people. RFID is rapidly becoming a cost-effective technology. RFID can be supplied as read only or read/write, does not require contact or line of sight to operate, can function under a variety of environmental conditions, and provides a high level of data integrity. Hence in the present work we seek to do comprehensive studies on the performance of the simulated hardware RFID system (operated in the range of frequency 120-130 MHz).

Realization and simulation of the hardware for RFID system and its performance study

Over the last couple of decades, the quality of surgical interventions has improved owing to the use of computer vision and robotic assistance. One such application of computer vision, namely, detection of surgical tools in videos is gaining attention of the medical image processing community. The main motivation for detection, localization, and annotation of surgical tools is to develop applications for surgical wsorkflow analysis. Such an analysis can aid in report generation, real-time decision support, etc. Cataract surgery is one of the common surgical procedure where surgeons do have direct visual access to the surgical site. Extremely small tools are used for this procedure and the surgeons observe the surgical site through a surgical microscope. In such cases, detecting the presence of tools can act an additional aid to the surgeon as well as other surgical staffs. We propose a framework consisting of a Convolutional Neural Network (CNN) which learns to distinguish and detect the presence of various surgical tools by learning robust features from the frames of a surgical video. Various deep neural architectures are hence evaluated for the task of detecting tools. The baseline models used for the purpose are pretrained on Imagenet dataset and they render upto 50% prediction accuracy. All the experiments have been validated on the dataset released as part of the Cataracts Grand Challenge. A framework for localization and detection of tools has also been proposed, which is capable of extracting visual features from glimpses of an image, by adaptively selecting and processing only the selected regions at high resolution.

Deep Neural Architecture for Localization and Tracking of Surgical Tools in Cataract Surgery

Breast cancer is the most common form of cancer in women. Early diagnosis can significantly improve lifeexpectancy and allow different treatment options. Clinicians favor 2D ultrasonography for breast tissue abnormality screening due to high sensitivity and specificity compared to competing technologies. However, inter- and intra-observer variability in visual assessment and reporting of lesions often handicaps its performance. Existing Computer Assisted Diagnosis (CAD) systems though being able to detect solid lesions are often restricted in performance. These restrictions are inability to (1) detect lesion of multiple sizes and shapes, and (2) differentiate between hypo-echoic lesions from their posterior acoustic shadowing. In this work we present a completely automatic system for detection and segmentation of breast lesions in 2D ultrasound images. We employ random forests for learning of tissue specific primal to discriminate breast lesions from surrounding normal tissues. This enables it to detect lesions of multiple shapes and sizes, as well as discriminate between hypo-echoic lesion from associated posterior acoustic shadowing. The primal comprises of (i) multiscale estimated ultrasonic statistical physics and (ii) scale-space characteristics. The random forest learns lesion vs. background primal from a database of 2D ultrasound images with labeled lesions. For segmentation, the posterior probabilities of lesion pixels estimated by the learnt random forest are hard thresholded to provide a random walks segmentation stage with starting seeds. Our method achieves detection with 99.19% accuracy and segmentation with mean contour-to-contour error < 3 pixels on a set of 40 images with 49 lesions.

Random forest learning of ultrasonic statistical physics and object spaces for lesion detection in 2D sonomammography

Correction for ‘Global spectral and local molecular connects for optical coherence tomography features to classify oral lesions towards unravelling quantitative imaging biomarkers’ by Satarupa Banerjee et al., RSC Adv., 2016, 6, 7511–7520.

/pdf/correction-global-spectral-and-local-molecular-connects-for-1h91qfflge.pdf

Correction: Global spectral and local molecular connects for optical coherence tomography features to classify oral lesions towards unravelling quantitative imaging biomarkers

Overlapping of cervical cells and poor contrast of cell cytoplasm are the major issues in accurate detection and segmentation of cervical cells An unsupervised cell segmentation approach is presented here Cell clump segmentation was carried out using the extended depth of field (EDF) image created from the images of different focal planes A modified Otsu method with prior class weights is proposed for accurate segmentation of nuclei from the cell clumps The cell cytoplasm was further segmented from cell clump depending upon the number of nucleus detected in that cell clump Level set model was used for cytoplasm segmentation

Debdoot Sheet

Papers

Realization and simulation of the hardware for RFID system and its performance study

Deep Neural Architecture for Localization and Tracking of Surgical Tools in Cataract Surgery

Random forest learning of ultrasonic statistical physics and object spaces for lesion detection in 2D sonomammography

Correction: Global spectral and local molecular connects for optical coherence tomography features to classify oral lesions towards unravelling quantitative imaging biomarkers

Unsupervised Segmentation of Overlapping Cervical Cell Cytoplasm.