Khawar Khurshid

Automated cardiac motion compensation in PET/CT for accurate reconstruction of PET myocardial perfusion images.

Error-free reconstruction of PET data with a registered CT attenuation map is essential for accurate quantification and interpretation of cardiac perfusion. Misalignment of the CT and PET data can produce an erroneous attenuation map that projects lung attenuation parameters onto the heart wall, thereby underestimating the attenuation and creating artifactual areas of hypoperfusion that can be misinterpreted as myocardial ischemia or infarction. The major causes of misregistration between CT and PET images are the respiratory motion, cardiac motion and gross physical motion of the patient. The misalignment artifact problem is overcome with automated cardiac registration software that minimizes the alignment error between the two modalities. Results show that the automated registration process works equally well for any respiratory phase in which the CT scan is acquired. Further evaluation of this procedure on 50 patients demonstrates that the automated registration software consistently aligns the two modalities, eliminating artifactual hypoperfusion in reconstructed PET images due to PET/CT misregistration. With this registration software, only one CT scan is required for PET/CT imaging, which reduces the radiation dose required for CT-based attenuation correction and improves the clinical workflow for PET/CT.

Speaker Verification Using Boosted Cepstral Features with Gaussian Distributions

An effective yet simple approach for text-dependent speaker verification is presented in this paper. The basic idea employs the fundamentals of Gaussian Mixture Model, which is a popular technique for speaker recognition in modern state-of-the- art systems. In this paper we introduce a novel technique for creating unique speaker models using spectral and prosodic features of speech signals which are further boosted to get the final robust discriminating speaker identity. Multi-class Adaptive Boosting (AdaBoost) algorithm is used for the classification of each speaker model. The Gaussian distributions of the Mel- Frequency Cepstral Coefficients (MFCC) of each speaker are pre- emphasized using the pitch information in the speech signals. The GMM combines the individual speaker models according to the set of mixing weights but our approach categorizes the speaker models in weak-learned sets which then linearly and optimally combine to form a strong classifier. The results of our algorithm show 98% correct speaker verification for the set of 16 speakers. On the average the percentage of false acceptance is 2% while the false rejection rate is 0%.

Automated PET/CT brain registration for accurate attenuation correction

Computed Tomography (CT) is used for the attenuation correction of Positron Emission Tomography (PET) to enhance the efficiency of data acquisition process and to improve the quality of the reconstructed PET data in the brain. Due to the use of two different modalities, chances of misalignment between PET and CT images are quite significant. The main cause of this misregistration is the motion of the patient during the PET scan and between the PET and CT scans. This misalignment produces an erroneous CT attenuation map that can project the bone and water attenuation parameters onto the brain, thereby under- or over-estimating the attenuation. To avoid the misregistration artifact and potential diagnostic misinterpretation, automated software for PET/CT brain registration has been developed. This software extracts the brain surface information from the CT and PET images and compensates for the translational and rotational misalignment between the two scans. This procedure has been applied to the dataset of a patient with visible perfusion defect in the brain, and the results show that the CTAC produced after the image registration eliminates that hypoperfusion artifact caused by the erroneous attenuation of the PET images.

Automated Software for PET/CT Image Registration to Avoid Unnecessary Invasive Cardiac Surgery

Computed tomography (CT) is used for the attenuation correction of positron emission tomography (PET) to enhance the efficiency of data acquisition process but due to the use of two different modalities chances of misalignment between PET and CT images are quite significant. There are multiple causes of this misalignment, the most important being the respiratory motion of the patient. The position of the heart in PET and CT can have a difference of up to 2 cm along the long axis of the body. If such a misaligned CT attenuation correction map is used for the PET reconstruction it will cause erroneous results by projecting the lungs attenuation parameters on the cardiac wall creating false areas of hypoperfusion which may be misinterpreted as myocardial ischemia or infarction and can lead to unnecessary invasive procedure like bypass or angioplasty. In this paper, an automated cardiac software alignment method is proposed to overcome this motion artifact. PET and CT heart geometries are aligned using efficient real time image processing techniques and results of this optimization procedure have been evaluated on 100 PET/CT cardiac data sets producing accurate cardiac alignment which eliminated PET/CT misregistration attenuation correction artifact.

A scalable architecture for geometric correction of multi-projector display systems

Abstract Multi-projector displays allow the realization of large and immersive projection environments by allowing the tiling of projections from multiple projectors. Such tiled displays require real time geometrical warping of the content that is being projected from each projector. This geometrical warping is a computationally intensive operation and is typically applied using high-end graphics processing units (GPUs) that are able to process a defined number of projector channels. Furthermore, this limits the applicability of such multi-projector display systems only to the content that is being generated using desktop based systems. In this paper we propose a platform independent FPGA based scalable hardware architecture for geometric correction of projected content that allows addition of each projector channel at a fractional increase in logic area. The proposed scheme provides real time correction of HD quality video streams and thus enables the use of this technology for embedded and standalone devices.

Automated PET/CT Cardiac Registration for Accurate Attenuation Correction

Alignment of PET and CT images is essential for accurate measurements of cardiac perfusion. Misalignment can produce an erroneous attenuation map that projects lung attenuation parameters onto the heart wall, thereby underestimating the attenuation, and creating artifactual areas of hypoperfusion which may be misinterpreted as myocardial ischemia or infarction. The main cause of misregistration between CT and PET images is the respiratory motion of the patient. In this paper, an automated cardiac software alignment method is proposed to overcome this motion artifact. In this approach, the heart is extracted from the PET data through windowing and c-mean clustering, and the CT scans are segmented to obtain the corresponding heart geometry. From this processed data, the heart geometries are registered, and a motion correction vector is calculated such that the alignment error of the two modalities is minimized. Results of this optimization procedure have been evaluated on 24 patient PET/CT cardiac data sets producing accurate cardiac alignment which eliminated PET/CT misregistration attenuation correction artifact

Efficient feature selection for Blind Image Quality Assessment based on natural scene statistics

Blind Image Quality Assessment (BIQA) has received considerable importance with the increase in the use of multimedia in our daily lives. The main objective of BIQA is to predict the quality of distorted images without any prior information about the original image. In this work, we propose an efficient feature selection method for blind image quality assessment based on natural scene statistics i.e., Distortion Identification-based Image Verity and Integrity Evaluation (DIIVINE). The proposed method produces better results for non-reference image quality assessment by selecting features, which produce the best Spearman Rank Order Correlation Constant (SROCC) scores averaged over 1000 random runs. The experimental results conducted on the LIVE database show that the proposed method strongly correlates to the subjective mean observer score and is competitive to the state-of-the-art image quality assessment techniques with a minimum number of features that reduces the computational expense.

A computationally low cost vision based tracking algorithm for human following robot

Recently, there has been an increasing interest in the field of human interactive robotics. Contrary to otherwise complex and resource hungry algorithms, we in this work have presented a computationally low cost algorithm for a human following robotic application. Instead of detecting the human, the algorithm makes use of a specific colour tag placed on the human subject which is detected by a camera mounted on the robot. Sensors including range sensor, magnetometer and optical encoders are utilized in tandem to assist the human following process. The method is tested on a custom built robotic platform running Raspberry pi minicomputer. We have performed and presented the results of several experiments for the evaluation of our method.

IM session identification by outlier detection in cross-correlation functions

The identification of encrypted Instant Messaging (IM) channels between users is made difficult by the presence of variable and high levels of uncorrelated background traffic. In this paper, we propose a novel Cross-correlation Outlier Detector (CCOD) to identify communicating end-users in a large group of users. Our technique uses traffic flow traces between individual users and IM service providers data center. We evaluate the CCOD on a data set of Yahoo! IM traffic traces with an average SNR of -6.11dB (data set includes ground truth). Results show that our technique provides 88% true positives (TP) rate, 3% false positives (FP) rate and 96% ROC area. Performance of the previous correlation-based schemes on the same data set was limited to 63% TP rate, 4% FP rate and 85% ROC area.

Breaching IM session privacy using causality

The breach of privacy in encrypted instant messenger (IM) service is a serious threat to user anonymity. Performance of previous de-anonymization strategies was limited to 65%. We perform network de-anonymization by taking advantage of the cause-effect relationship between sent and received packet streams and demonstrate this approach on a data set of Yahoo! IM service traffic traces. An investigation of various measures of causality shows that IM networks can be breached with a hit rate of 99%. A KCI Causality based approach alone can provide a true positive rate of about 97%. Individual performances of Granger, Zhang and IGCI causality are limited owing to the very low SNR of packet traces and variable network delays.