Ehsan Golkar

Feature extraction algorithm for fill level and cap inspection in bottling machine

Automated Visual Inspection Systems (AVIS) have a strong ability to improve bottle manufacturing quality control by means of inspecting products automatically instead of through manual inspections. AVIS automatically tends to make a suitable decision in process and results classification according to the images of the products via image processing and Artificial Intelligence techniques. Since bottling is one of the most common packaging styles in the food and medical industries, in this paper we will concentrate on the visual inspection of bottles. Checking the quality of the cap closure and over-filling/under-filling checks for the level of the liquid in the bottle have been investigated to reach an optimized bottle product. Therefore, in this research general hardware and modules for these systems are investigated. Besides, new techniques of bottle inspection are reviewed along with presenting previous work of other researchers. Subsequently we will propose a feature extraction algorithm to inspect cap closure and level of the liquid in the bottle together, in the same system. According to the new proposed method, our system classifies three situations for cap condition and three situations for the condition of the level of the liquid. As a result the system has investigated 9 situations. The algorithm of the system will accept its system when the liquid level is in the correct position and the cap is in the normal condition. Other situations will be rejected. The proper algorithm which is proposed here using bottle visual inspection techniques leads our system to reach an optimized liquid level with a high quality of the cap closure.

Respiratory motion tracking using the kinect camera

Respiratory motion presents a challenge in developing techniques to increase the accuracy of image acquisition or guided interventions in abdominal and thoracic organs. The use of distance cameras, such as that from Vision RT Ltd has been considered as a surrogate for respiratory motion tracking in external beam radiotherapy that does not require the use of markers. However, it is very costly and thus this project suggests using the Microsoft Xbox Kinect™ for tracking respiratory motion. The aim of the project is to utilize the Kinect camera which for instance, can be put inside the CT scanner room in the hospital. This system records the 3D motion data and further analyses the consistency of the respiratory motion. A Graphical User Interface (GUI) was developed to display the breathing signal and statistics of the data. This respiratory motion tracking system will facilitate clinicians in choosing a suitable treatment option for lung cancer patients with less cost.

Development of a respiratory motion tracking system using a distance camera for diagnostic imaging and external beam radiotherapy delivery

This paper describes the development of a system that uses off the shelf ranging technology (distance cameras) to acquire a real-time multidimensional respiratory signal from a 3D surface reconstruction of the patients chest and abdomen without the use of markers. The preliminary measurements are discussed within the context of the use of such a system with an existing CT scanner.

A composite registration framework for respiratory motion modelling from 4D MRI

4D MRI is current gaining attention as an imaging modality which is able to capture inter-cycle variability of respiratory motion. Such information is beneficial for example in radiotherapy planning and delivery. However the motion extracted from 4D MRI is dependant on the method of processing the images. Typically a deformable registration algorithm is used so that the 3D motion of every point in the image can be found. We have used several popular registration algorithms which have been recently published with publicly available data. Our findings suggest that due to the low contrast of 4D MRI, the registration is only accurate near clear boundaries, and less accurate elsewhere, to different degrees depending on the algorithm used. We thus suggest modelling organ deformation as an affine transformation, which is then embedded into the deformation field for more accurate modelling of respiratory motion.

Real-Time Curvature Defect Detection on Outer Surfaces Using Best-Fit Polynomial Interpolation

This paper presents a novel, real-time defect detection system, based on a best-fit polynomial interpolation, that inspects the conditions of outer surfaces. The defect detection system is an enhanced feature extraction method that employs this technique to inspect the flatness, waviness, blob, and curvature faults of these surfaces. The proposed method has been performed, tested, and validated on numerous pipes and ceramic tiles. The results illustrate that the physical defects such as abnormal, popped-up blobs are recognized completely, and that flames, waviness, and curvature faults are detected simultaneously.

Comparison between the Kinect™ V1 and Kinect™ V2 for respiratory motion tracking

In this paper, we aim to assess the accuracy as well as compare between the Microsoft Kinect™ version 1 and Microsoft Kinect™ version 2 with regards to the purpose of respiratory motion tracking. We find that both correlate well to an alternative method of respiratory motion measurement i.e. a respiratory belt, up to a distance of around 2 m. However, we find that the Kinect™ version 2 has a slightly higher correlation, which can be explained by it being a newer device, as well as having a slightly higher cost.

Comparison of image registration similarity measures for an abdominal organ segmentation framework

Automated segmentation is a primary step in medical diagnosis applications. This paper presents an image segmentation propagation method via image registration. Therefore, segmentation propagation accuracy depends on the accuracy of image registration. Three similarity measures were considered, namely Sum of Squared (Intensity) Differences (SSD), Mutual Information (MI) and Cross Correlation (CC) and their results were compared to each other. The results shows that there are slight differences between the use of these different similarity measures, however, the results are similar in the evaluation using MR images.

An automated liver tumour segmentation from abdominal CT scans for hepatic surgical planning

PurposeSegmentation of liver tumours is an important part of the 3D visualisation of the liver anatomy for surgical planning. The spatial relationship between tumours and other structures inside the liver forms the basis of preoperative surgical risk assessment. However, the automatic segmentation of liver tumours from abdominal CT scans is riddled with challenges. Tumours located at the border of the liver impose a big challenge as the surrounding tissues could have similar intensities.MethodsIn this work, we introduce a fully automated liver tumour segmentation approach in contrast-enhanced CT datasets. The method is a multi-stage technique which starts with contrast enhancement of the tumours using anisotropic filtering, followed by adaptive thresholding to extract the initial mask of the tumours from an identified liver region of interest. Localised level set-based active contours are used to extend the mask to the tumour boundaries.ResultsThe proposed method is validated on the IRCAD database with pathologies that offer highly variable and complex liver tumours. The results are compared quantitatively to the ground truth, which is delineated by experts. We achieved an average dice similarity coefficient of 75% over all patients with liver tumours in the database with overall absolute relative volume difference of 11%. This is comparable to other recent works, which include semiautomated methods, although they were validated on different datasets.ConclusionsThe proposed approach aims to segment tumours inside the liver envelope automatically with a level of accuracy adequate for its use as a tool for surgical planning using abdominal CT images. The approach will be validated on larger datasets in the future.

Assessing the 3D accuracy of consumer grade distance camera measurement of respiratory motion

Recently range imagers or distance camera systems have garnered interest for measuring respiratory motion without using markers, which can then be used as a surrogate in diagnosis and treatment for example in diagnostic imaging or radiotherapy. However, their use may have limitations, especially among lower cost systems, whereby their accuracy decrease greatly with the distance of the patient from the camera. This is considering the fact that the motion amplitude of the anterior surface of the body in normal breathing is typically around 1 cm or less, which is at the limit of accuracy of these systems. This accuracy limitation is even more pertinent when the fact that the 1 cm accuracy is desired over the whole anterior surface that is image and not just an average measurement of distance. We study this limitation in a low cost system i.e. the Microsoft KinectTM, using both version 1 and version 2 of the sensor. The 3D accuracy of both versions is compared with an alternative method of respiratory motion measurement i.e. a respiratory belt, at a distance of around 1.35 m. This study can be a guide for the design and application of range imaging systems in the clinical setting.

Comparison of intensity based deformable registration methods for respiratory motion modelling from 4D MRI

Deformable image registration is a key part of modern medical image processing and analysis. The aim of image registration is to align one image to another image. In this paper, three deformable image registration methods (NiftyReg, MRF-based and lreg) are compared based on their estimated motion field from 4D MRI data for respiratory motion modelling. The result shows that all of these methods are able to extract respiratory motion with different degrees of certainty. In terms of overall displacement for each organ, lreg with piecewise affine transformation produces more realistic motion than NiftyReg and MRF-Based registration. Finally, we can conclude that deformable image registration can be used to extract respiratory motion for applications such as integration into external beam radiotherapy treatment planning and delivery.