[en] A new approach in NDE (Non Destructive Evaluation) by Ultrasound is proposed. It is the implementation of the autoregressive spectral analysis (AR) of the Ultrasonic echographic signals applied to Nuclear components (L pipes for primary coolant circuit of PWR reactors). The purpose is to improve the signal to noise ratio by modeling the structural noise to cancel it by withness technic. The structural noise that disturb detection is of transient nature. It is parted in small sections supposed to be stationary. For each section, the spectral analysis obtained by correlogram and AR models (Burg and Marple) is compared. The spectral calibration problem of AR model induce us to use correlogram to examine the stationarity. A spectral representation versus depth and moving of the probe clearly shows the stationarity of the signal during inspection at constant depth and shows the non-stationarity when the various sections are observed at different depths

[en] We describe a localisation system for a robot moving in a known environment. Unlike the currently used methods for industrial robots, our approach does not require any beacons to be installed: the system uses odometry to estimate the vehicle position continuously, and corrects this estimation when necessary by identifying some objects of the environment through vision. These objects, used as landmarks, were previously recorded in a data base. The different parts of the system are presented, particularly the way the uncertainty on odometry is updated and how prior knowledge (position estimation and data base) is employed to facilitate landmark identification. 7 cm on xy and 1 deg on the heading is the typical precision obtained in term of localisation. (authors). 36 refs., 26 figs

[en] Steam generator tubes in nuclear power plants are periodically checked by means of eddy current probes. The output of a probe is composed of three types of signals: known events (rolling zone, support plates, U-bend part), noise (mainly metallurgical noise) and possible flaws. The latter are random transients, both in arrival time and in shape: they have to be detected and then estimated, before to be fed to the high level stages of a diagnostic system. The objective of the study presented is to develop a semi-automatic system, which could manage and process more than 1 M-bytes of data per tube and provide an operator with reliable diagnostics proposals within a few minutes. This can be achieved only by cooperation of several digital signal processing techniques: detection, segmentation, estimation, noise subtraction, adaptive filtering, modelization, pattern recognition. The paper describes some of these items

[en] We present the various researches we made on tomographic imaging by coding and reconstruction. These works are based on two chief kinds of methods: Coding Aperture Imaging and Transaxial Tomography, each of them including theoretical studies and realization of tomographic systems for definite applications. In this paper, we mainly present the more recent studies concerning reconstruction problems from missing data, with development of an original method by constrained natural pixels, and implementation of multislit coding in micro-imaging of laser plasmas

[en] This paper refers to the industrial application of image processing using Non Destructive Testing by radiography. The various problems involved by the conception of a numerical tool are described. This tool intends to help radiograph experts to quantify defects and to follow up their evolution, using numerical techniques. The sequences of processings that achieve defect segmentation and quantization are detailed. They are based on the thorough knowledge of radiographs formation techniques. The process uses various methods of image analysis, including textural analysis and morphological mathematics. The interface between the final product and users will occur in an explicit language, using the terms of radiographic expertise without showing any processing details. The problem is thoroughly described: image formation, digitization, processings fitted to flaw morphology and finally product structure in progress. 12 refs

[en] Pneumonia is a disease caused by inflammation of the lung tissue that is transmitted by various means, primarily bacteria. Early and accurate diagnosis is important in reducing the morbidity and mortality of the disease. The primary imaging method used for the diagnosis of pneumonia is lung x-ray. While typical imaging findings of pneumonia may be present on lung imaging, nonspecific images may be present. In addition, many health units may not have qualified personnel to perform this procedure or there may be errors in diagnoses made by traditional methods. For this reason, computer systems can be used to prevent error rates that may occur in traditional methods. Many methods have been developed to train data sets. In this article, a new model has been developed based on the layers of the ResNet50. The developed model was compared with the architectures InceptionV3, AlexNet, GoogleNet, ResNet50 and DenseNet201. In the developed model, the maximum accuracy rate was achieved as 97.22%. The model developed was followed by DenseNet201, ResNet50, InceptionV3, GoogleNet and AlexNet, respectively, according to their accuracy. With these developed models, the diagnosis of pneumonia can be made early and accurately, and the treatment management of the patient will be determined quickly. (authors)

[en] In response to challenges in liver occupancy such a variety of types and manifestations and difficulties in differentiating benign and malignant ones, this paper takes liver images of enhanced MRI scan as the research object, targets on the detection and identification of liver occupancy lesion areas and determining if it is benign or malignant. Accordingly, the paper proposed an auxiliary diagnosis method for liver image combining deep learning and MRI medical imaging. The first step is to establish a reusable standard dataset for MRI liver occupancy detection by pre-processing, image denoising, lesion annotation and data augmentation. Then it improves the classical region-based convolutional neural network (R-CNN) algorithm Faster R-CNN by incorporating CondenseNet feature extraction network, custom-designed anchor size and transfer learning pre-training. This is to further improve the detection accuracy and benign and malignant classification performance of liver occupancy. Experiments show that the improved model algorithm can effectively identify and localise liver occupancies in MRI images, and achieves a mean average precision (mAP) of 0.848 and an Area Under the Curve (AUC) of 0.926 on the MRI standard dataset. This study has important research significance and application value for reducing manual misses and misdiagnosis and improving the early clinical diagnosis rate of liver cancer. (authors)

[en] In dynamic tomography, the measured objects or organs are no-longer supposed to be static in the scanner during the acquisition but are supposed to move or to be deformed. Our approach is the analytic deformation compensation during the reconstruction. Our work concentrates on 3-dimensional cone beam tomography. We introduce a new large class of deformations preserving the 3-dimensional cone beam geometry. We show that deformations from this class can be analytically compensated. We present numerical experiments on phantoms showing the compensation of these deformations in 3-dimensional cone beam tomography. (authors)

[en] Identifying the histological phenotype of non-small cell lung cancer (NSCLC) is of crucial importance to its treatment and prognosis. Radiomics analysis is an emerging quantitative analysis method, which can automatically extract the mass, objective and indiscernible tumor features from medical images (such as CT, MRI and PET). The radiomics-based prediction model has the potential to non-invasively extract the tumor phenotype characteristics. However, the existing research ignores the stability of extracted features, which restricts the performance and robustness of the constructed model. White most of the methods in the literature use classification accuracy to solve the problem of radiomics features stability, in this paper we propose the use of SOM (Self-organizing Mapping) and K-means to evaluate the stability of different feature subsets. The subset with good clustering performance is selected as the optimal feature subset.When the optimal feature subset is used for modeling, compared with other feature subsets, the higher AUC(Area Under Curve) and lower SD (Standard Deviation) on the three classifiers show that the feature subset had excellent classification performance and good stability, and can distinguish NSCLC subtypes more accurately and robustly. (authors)

[en] Image denoising is an important concept in image processing for improving the image quality. It is difficult to remove noise from images because of the various causes of noise. Imaging noise is made up of many different types of noise, including Gaussian, impulse, salt, pepper, and speckle noise. Increasing emphasis has been paid to Convolution Neural Networks (CNNs) in image denoising. Image denoising has been researched using a variety of CNN approaches. For the evaluation of these methods, various datasets were utilized. Liver Tumor is the leading cause of cancer-related death worldwide. By using Computed Tomography (CT) to detect liver tumor early, millions of patients could be spared from death each year. Denoising a picture means cleaning up an image that has been corrupted by unwanted noise. Due to the fact that noise, edge, and texture are all high frequency components, denoising can be tricky, and the resulting images may be missing some finer features. Applications where recovering the original image content is vital for good performance benefit greatly from image denoising, including image reconstruction, activity recognition, image restoration, segmentation techniques, and image classification. Tumors of this type are difficult to detect and are almost always discovered at an advanced stage, posing a serious threat to the patient's life. As a result, finding a tumour at an early stage is critical. Tumors can be detected non-invasively using medical image processing. There is a pressing need for software that can automatically read, detect, and evaluate CT scans by removing noise from the images. As a result, any system must deal with a bottleneck in liver segmentation and extraction from CT scans. To segment and classify liver CT images after denoising images, a deep CNN technique is proposed in this research. An Image Quality Enhancement model with Image Denoising and Edge based Segmentation (IQE-ID-EbS) is proposed in this research that effectively reduces noise levels in the image and then performs edge based segmentation for feature extraction from the CT images. The proposed model is compared with the traditional models and the results represent that the proposed model performance is better. (authors)