الوحدة النصية و علاقتها بالنمط اللغوي الكبير: دراسة نظرية في ضوء اللسانيات النصية

The text is larger pattern of any language; it consists smaller units of a language. The Aristotle was among the earliest who discussed the unity of text, then letter on Arabian scholars especially philosophers started to relate it with Quranic text and analysis it according this concept; even some of them claim that the unity of Quranic text is the only reason in its “Ijaaz”. And more then that there were some others who discussed its different aspects in Arabic poetry and prose in the same time. On the other hand the western scholar studied the same “textual unity” in different ways in modern textual linguistics by defining the “unity” as a larger pattern of text. This work will discuss the all aspects of textual unit and its relations with larger pattern in the light of textual linguistics.

Registration of Medical Images Based on Common Subregions

Registration is an important and fundamental medical image analysis technique for the alignment of two or more images of the same organ into a single more informative and ideal image for receiving precise and complementary information. The high quality and more informative images help surgeons to accurately locate region of interest while the surgery is in progress. Reliable, accurate, robust and computationally efficient image registration is necessary and is always required in clinical practices. However, the development of more accurate and efficient registration techniques in clinically acceptable time-frames is always a challenge. Most of the registration approaches consider entire image content and global features for the alignment of two or more images. Such approaches are tend to be computationally intensive and inaccurate because it requires full image matching. In medical image registration, computational efficiency and high accuracy may be achieved by restricting the registration process to subregions within the image being registered. Registration based on subregions and local features consider salient regions (interested regions) in the whole medical image. These approaches are computationally efficient and accurate because the registration needs to be performed only for the specific region. Automatic detection and extraction of interested subregions in medical images is always required in IGS and radiotherapy. However, automatic detection and registration of interested subregions in medical images is difficult and prone to errors due to complex and non-linear nature, and the availability of limited features for registration. This work presents an automatic feature based approaches for the rigid and deformable registration of medical images with the aim of high accuracy and computational efficiency. Instead of globally registering one image (moving image or source image) to another image (fixed image or target image), interested common subregions in two images are first automatically detected. After the detection of interested common subregions in both images, the detected common subregions are registered with local transformation parameters. The obtained local transformation parameters are then applied on source image, which recovers it according to the coordinates of target image. Finally, the obtained recovered source image is aligned with fixed target image with global transformation and correct registration with high efficiency is therefore achieved. All the experiments are performed on real 2D brain MRI images of patients with tumor. To demonstrate the computational efficiency, accuracy, reliability and robustness of the proposed approaches, extensive experiments are performed and the results are compared with existing standard registration methods. The performance of the proposed methods is evaluated using popular statistical metrics i.e. mutual information (MI), mean square error (MSE), peak signal to noise ratio (PSNR), sum of square differences (SSD), cross correlation (CC) and computation time. The experimental results sows that the obtained values of MI, PSNR and CC for the proposed methods are high than existing methods. Similarly, the obtained values of MSE, SSD and computational time for the proposed methods are low compared to existing methods. Thus it is obvious from the experiments that the proposed registration approaches outperform than the existing registration approaches in terms of computational efficiency and registration accuracy. Moreover, the proposed approaches automatically detect the desired common subregions in rigid and deformable medical images and perform successful registration on it.