Transferable HMM probability matrices in multi‐orientation geometric medical volumes segmentation

Abstract

Acceptable error rate, low quality assessment, and time complexity are the major problems in image segmentation, which needed to be discovered. A variety of acceleration techniques have been applied and achieve real time results, but still limited in 3D. HMM is one of the best statistical techniques that played a significant rule recently. The problem associated with HMM is time complexity, which has been resolved using different accelerator. In this research, we propose a methodology for transferring HMM matrices from image to another skipping the training time for the rest of the 3D volume. One HMM train is generated and generalized to the whole volume. The concepts behind multi-orientation geometrical segmentation has been employed here to improve the quality of HMM segmentation. Axial, saggital, and coronal orientations have been considered individually and together to achieve accurate segmentation results in less processing time and superior quality in the detection accuracy. KEYWORDS hidden Markov model, image segmentation, medical imaging, transfer learning, 3D volumes 1 INTRODUCTION Nowadays, segmentation is one of the key problems in the field of computer vision. The understanding of a scene is considered to be the main problem of different computer-based applications that nourish from inferring knowledge from imagery, such as self-driving vehicles, human-computer interaction, virtual reality, medical applications, and many others. The growing interest of deep learning has added an enormous boost to the already rapidly developing field of computer vision where many semantic segmentation problems are being tackled using deep architectures. In the early stages of this century, civilization became more picture based. The employment of image processing and analysis is widely broadcast in many activities more than ever been. The expansion of social networks and due to globalization mechanisms helps in employing image processing analysis in this era. 1 Computer-aided diagnosis (CAD) systems are computer software, which have been mainly implemented for helping specialists doing their jobs; many applications have employed computer systems for better quality such as those in other works. 2-5 These software techniques improve diagnosing medical data. Big data and high-speed processing of this data helps the physician reach an impressive and high-quality diagnosis in real time with the mentioned software techniques or systems from possibly different sources. 6 Image segmentation plays an important role in both natural and medical analysis. 7 It is the process of segmenting or partitioning a digital image into a set of separated regions according to some connected parts. 8,9 Volume segmentation can be defined as declaring the voxels, which comprise the contour edges and the volume of objects of interest. Many techniques can be employed into segmentation. Those techniques could not be a portion of image processing, such as mathematical morphology and watershed 10,11 hard or fuzzy clustering, 12-14 fuzzy connectedness...

ANFIS and Deep Learning based missing sensor data prediction in IoT

“…Hidden Markov Model (HMM) is a stochastic finite state machine that is composed of four elements. 9 These elements are (i) states, (ii) possible observations, (iii) state transmission probability matrix, and (iv) emission probability matrix. 10 HMM has a number of advantages, but in the scope of this paper, we would like to focus on HMM's prediction capability.…”

“…With its strong statistical foundation, 11 HMM can detect patterns on time series efficiently. [9][10][11] In the work of Hassan and Nath, 11 HMM is used to predict stock market prices. The proposed model takes four inputs, which are opening, closing, high, and low prices of the current day.…”

“…HMM has a number of advantages, but in the scope of this paper, we would like to focus on HMM's prediction capability. With its strong statistical foundation, HMM can detect patterns on time series efficiently . In the work of Hassan and Nath, HMM is used to predict stock market prices.…”

Efficient 3D medical image segmentation algorithm over a secured multimedia network

“…Execution time complexity was resolved using hardware accelerators (GPU). In [7], we introduce Transfer Learning to accelerate the segmentation process. We propose a methodology for transferring HMM matrices from image to another skipping the training time for the rest of the 3D volume.…”

“…GPU-based parallel processing has been used as an efficient hardware solution, it is still inadequate for 3D volumes due to the huge amount of data in 3D volume and the large stack of Medical DICOM files that pose an overhead on the training process of HMMs. AlZu'bi et al accelerated the HMM segmentation process in [7] using the transfer learning. The Algorithm explained in Figure 2 explains the procedure for segmenting 2D images using HMMs [5] and the blocks which slow down the segmentation time are highlighted.…”