Search

This work describes a multiplication and accumulation (MAC) accelerator integrated with a memory interface. The link is designed to take advantage of naturally existing sparsity in a neural network. The link operating at 16 Gb/s achieves 0.1875-pJ/bit signaling efficiency for random data but, for sparse data, signaling efficiency can improve to 0.0375 pJ/bit. Similarly, the MAC unit accelerates the computation utilizing the phase domain accumulation process and provides a 40% improvement in energy efficiency for sparse data and at the same achieves inference accuracy of 94% for the MNIST data set.

Abstract Introduction: Here we describe our technique and results of beating heart pulmonary thromboendarterectomy (PTE) with cardiopulmonary bypass (CPB) in four patients for treatment of chronic thromboembolic pulmonary hypertension (CTEPH). Methods: Retrospective analysis of data from patients who underwent PTE for CTEPH between January 2019 and September 2020. Patients were followed up with clinical assessment, 2D echocardiography, and computed tomography pulmonary angiogram. Results: Four patients were operated for CTEPH using our technique. Moderate tricuspid regurgitation (TR) and severe TR were found in two patients each. Severe right ventricular (RV) dysfunction was found in all cases. Thrombi were classified as Jamieson type II in three cases and type I in one case. Postoperative median direct manometric pulmonary artery (PA) pressures decreased (from 46.5 mmHg to 23.5 mmHg), median CPB time was 126 minutes, and median temperature was 33.35 °C. Mechanical ventilation was for a median of 19.5 hours. There was one re-exploration. Median intensive care unit stay was 7.5 days. There was no mortality. Postoperative 2D echocardiography revealed decrease in median PA systolic pressures (from 85 mmHg to 33 mmHg), improvement in RV function by tricuspid annular plane systolic excursion (median 14 mm vs. 16 mm), and improved postoperative oxygen saturations (88.5% vs. 99%). In follow-up (ranging between 2-15 months), all patients reported improvement in quality of life and were in New York Heart Association class I. Conclusion: With our described simple modifications, advances in perfusion, and blood conservation technologies, one can avoid the need for deep hypothermic circulatory arrest during PTE.

Corpus Callosum (CC) is an important white-matter structure in the human brain. Magnetic resonance imaging (MRI) is a non-invasive medical imaging technique that provides high resolution images for the structures. Segmentation is an important step in medical image analysis. This paper proposes a fully automated technique for segmentation of CC on the midsagittal slice of T1-weighted brain MR images. The proposed technique consists of three modules. First it clusters all homogenous regions in the image with an adaptive mean shift (AMS) technique. The automatic CC contour initialization (ACI) is achieved using the region analysis, template matching and location analysis, thus identify the CC region. Finally, the boundary of recognized CC region is used as the initial contour in the Geometric Active Contour (GAC) model, and is evolved to obtain the final segmentation result of CC. Experimental results demonstrate that the proposed AMS-ACI technique is able to provide accurate initial CC contour, and the proposed AMS-ACI-GAC technique overcomes the problem of user-guided initialization in existing GAC techniques, and provides a reliable and accurate performance in CC segmentation.

Skin lesion segmentation is a primary step for skin lesion analysis, which can benefit the subsequent classification task. It is a challenging task since the boundaries of pigment regions may be fuzzy and the entire lesion may share a similar color. Prevalent deep learning methods for skin lesion segmentation make predictions by ensembling different convolutional neural networks (CNN), aggregating multi-scale information, or by multi-task learning framework. The main purpose of doing so is trying to make use of as much information as possible so as to make robust predictions. A multi-task learning framework has been proved to be beneficial for the skin lesion segmentation task, which is usually incorporated with the skin lesion classification task. However, multi-task learning requires extra labeling information which may not be available for the skin lesion images. In this paper, a novel CNN architecture using auxiliary information is proposed. Edge prediction, as an auxiliary task, is performed simultaneously with the segmentation task. A cross-connection layer module is proposed, where the intermediate feature maps of each task are fed into the subblocks of the other task which can implicitly guide the neural network to focus on the boundary region of the segmentation task. In addition, a multi-scale feature aggregation module is proposed, which makes use of features of different scales and enhances the performance of the proposed method. Experimental results show that the proposed method obtains a better performance compared with the state-of-the-art methods with a Jaccard Index (JA) of 79.46, Accuracy (ACC) of 94.32, SEN of 88.76 with only one integrated model, which can be learned in an end-to-end manner.

Introduction: Cardiac surgery-associated kidney injury (CSA-AKI) is common but relatively less is known about its progression. The present study is aimed at evaluating the incidence and course of CSA-AKI and its relationship with the different durations of cardiopulmonary bypass (CPB) and cross clamp times. Materials and Methods : Occurrences of CSA-AKI are evaluated as per the Akin Kidney Injury Network (AKIN) criteria over the course of 5 postoperative day (POD) in 100 patients. The relationship of different durations of CPB and aortic cross clamp time with CSA-AKI is analyzed by Chi-squared test for trend and other appropriate tests using INSTAT software. Results: One hundred (43 male, 57 female; mean age of 37.01 ± 12.28 years, and baseline mean serum creatinine 0.99 ± 0.20 mg %) patients undergone mostly valve replacement, and congenital heart disease correction was evaluated. Nearly 49% suffered CSA-AKI (81.63% AKIN Class I) with maximum numbers on 2 nd POD. Serum creatinine followed a falling trend 3 rd POD onward except in 8.16% cases of CSA-AKI. Oliguria was absent even in AKIN Class II. The CPB time >70 min and cross clamp time >60 min increase CSA-AKI risk by an OR of 4.76 and 2.84, respectively (P < 0.05). Conclusion: CSA-AKI is very prevalent; mostly of AKIN Class I and increases with increasing CPB and cross clamp time. Urine output is not a reliable indicator of CSA-AKI. The AKIN Class II on the very 1 st POD or increasing trend of serum creatinine beyond 3 rd POD should alert for early intervention.

Despite the technological advances in medical diagnosis, accurate detection of infectious tuberculosis (TB) still poses challenges due to complex image features and thus infectious TB continues to be a public health problem of global proportions. Currently, the detection of TB is mainly conducted visually by radiologists examining chest radiographs (CXRs). To reduce the backlog of CXR examination and provide more precise quantitative assessment, computer-aided detection (CAD) systems for potential lung lesions have been increasingly adopted and commercialized for clinical practice. CADs work as supporting tools to alert radiologists on suspected features that could have easily been neglected. In this paper, an effective CAD system aimed for acinar shadow regions detection in CXRs is proposed. This system exploits textural and photometric features analysis techniques which include local binary pattern (LBP), grey level co-occurrence matrix (GLCM) and histogram of oriented gradients (HOG) to analyze target regions in CXRs. Classification of acinar shadows using Adaboost is then deployed to verify the performance of a combination of these techniques. Comparative study in different image databases shows that the proposed CAD system delivers consistent high accuracy in detecting acinar shadows.

Accurate and efficient detection of cell nuclei is an important step towards the development of a pathology-based Computer Aided Diagnosis. Generally, high-resolution histopathology images are very large, in the order of billion pixels, therefore nuclei detection is a highly compute intensive task, and software implementation requires a significant amount of processing time. To assist the doctors in real time, special hardware accelerators, which can reduce the processing time, are required. In this paper, we propose a Field Programmable Gate Array (FPGA) implementation of automated nuclei detection algorithm using generalized Laplacian of Gaussian filters. The experimental results show that the implemented architecture has the potential to provide a significant improvement in processing time without losing detection accuracy.

Adaptive search is one of the fastest fractal compression algorithms and has gained great success in many industrial applications. By substituting the luminance offset by the range block mean, the authors create a completely new version for both the encoding and decoding algorithms. In this paper, theoretically, they prove that the proposed decoding algorithm converges at least as fast as the existing decoding algorithms using the luminance offset. In addition, they prove that the attractor of the decoding algorithm can be represented by a linear combination of range-averaged images. These theorems are very important contributions to the theory and applications of fractal image compression. As a result, the decoding image can be represented as the sum of the DC and AC component images, which is similar with discrete cosine transform or wavelet transform. To further speed up this algorithm and reduce the complexity of range and domain blocks matching, they propose two improvements in this paper, that is, employing the post-quantisation and geometric neighbouring local search to replace the currently used pre-quantisation and the global search, respectively. The corresponding experimental results show the proposed encoding and decoding algorithms can provide a better performance compared with the existing algorithms.