Project Abstract Submission : Entry # 456. Part 1 - Team. Part 2 - Project. Team Leader Name. Maroua Filali. Team Leader .

Transcription

1 Part 1 - Team Team Leader Name Maroua Filali Team Leader mf @qu.edu.qa 2nd Team Member Name Ealaf Hussein 2nd Team Member eh @qu.edu.qa 3rd Team Member Name Salma Shalaby 3rd Team Member ss @qu.edu.qa Faculty Advisor Name Abbes Amira Faculty Advisor abbes.amira@qu.edu.qa Faculty Advisor University Job Title Professor of Computer Engineering How many male team members? 0 How many female team members? 3 Your University Qatar University Country of Residence Qatar Part 2 - Project 1/5

2 Project Title Automatic aneurysm segmentation on Reconfigurable SoC General Topic / Problem statement General Topic / Problem statement: - Aneurysms are caused by dilation in the walls of a blood vessel, becoming so thin, such that blood flow to it can lead to rupture. Cerebral aneurysms are found in the brain. Ruptured aneurysm causes blood leakage into the area surrounding it, which is a lifethreatening condition. Hence, early and accurate diagnosis of aneurysm is highly required. Knowing some parameters about the aneurysm, such as size and location, helps in deciding the right treatment. - Ruptured brain aneurysms are fatal in about 40% of cases. Of those who survive, about 66% suffer some permanent neurological deficit. Accurate early diagnosis is critical, as the initial haemorrhage may be fatal. Despite widespread neuroimaging availability, misdiagnosis or delays in diagnosis occurs in up to 25% of patients with subarachnoid haemorrhage (SAH) when initially presenting for medical treatment. Our project aims to develop a new algorithm for image segmentation to help the diagnosis of cerebral aneurysm. - This project aims to implement a real-time automated segmentation technique for cerebral aneurysm on the ZYNQ system on chip (SoC), and realizing the results in a 3D plane, utilizing a Virtual Reality Lab to create an interactive environment for training purposes. Image segmentation is the process of clustering pixels based on shared characteristics. It is used in the medical field for many purposes such as diagnosis, surgery planning, and locating tumours. Medical image segmentation can be accelerated on reconfigurable hardware as it requires huge computation power for 3D DICOM data. - The significance of the project is presented in the need for an automated system to segment cerebral aneurysms. Since medical images have high resolution, they require a high computational power. This can be provided through a Field Programmable Gate Array (FPGA), which provides an accelerated computation that is needed for this system. PRODUCTS/STUDIES: A KALMAN filter-pca based approach for brain aneurysm segmentation Novel VLSI Architecture for Real Time Medical Image Segmentation Real Time Implementation of Medical Images Segmentation Based on PSO Segmentation of Bowel Images and its implementation using Virtex FPGA kit Segmentation of MR Images by Using Grow and Learn Network on FPGAs Segmentation and separation of Cerebral Aneurysm: A multi-phase Approach Geodesic Active Contour with adaptive Configuration for cerebral vessel and aneurysm segmentation. Automatic Detection, Segmentation and classification of Abdominal Aortic Aneurysm using Deep Learning. Intracranial aneurysm segmentation in 3D CT angiography: Method and quantitative validation with and without prior noise filtering. Project Description 2/5

3 - The design of the project will be carried through several phases. The data will be provided to the FPGA. In this stage, the 2D slices of aneurysm will be segmented using thresholding and wavelet transform. Once the results of the segmentation are available, they can be sent to a variety of devices that can be interfaced through the FPGA. For instance, a Bluetooth Dongle can be connected to the FPGA, to send the result to a smartphone platform. Also, the results can be displayed in a monitor through an interface with an HDMI connection. Moreover, the slices will be converted to a 3D volume image, and will then be shown in a Virtual Reality environment. The main objectives of this project are: To develop an efficient algorithm to detect and segment cerebral aneurysm on FPGA and its simulation on MATLAB. To implement the proposed algorithms on ZYNQ SoC Zedboard. To evaluate the FPGA implementation of the proposed algorithm based on subjective and objective evaluation using different performance metrics. To show the results in an interactive virtual reality environment. To demonstrate and visualize the segmented aneurysm in 3D volume in VR environment - The limitations of the project happen in case there are no enough data to be provided to the system, hence, the algorithm would be data-biased. To solve this, we have managed to get a set of data for developing the algorithm, then more data will be used for testing and evaluating the algorithm. Proposed Approach/Methodology 3/5

4 Proposed Approach/Methodology: - To achieve the objectives of the project, a set of software and hardware are needed. For the software, MATLAB is used to develop and simulate the segmentation algorithm. It is also used to implement the graphical user interface of the system. To configure this system on the Zedboard, several software will be used. These software packages are: Vivado IDE: It is used to create the components of the hardware system of the System on Chip design, such as the memory, interfaces, peripherals, etc. Also, it integrates and packages intellectual property (IP). SDK: It contains driver support for all the Xilinx IPs. In addition, it has library support for ARM and NEON extensions. MATLAB / Simulink: provide suitable functions for image processing, and the algorithm can be converted into a code that can be implemented on the Zynq platform (C/C++). System Generator: employs the Simulink design for the FPGA system design, providing a high-level model-based design for the hardware platform. ZedBoard : is a low-cost board that features the Zynq device (XC7Z020). This platform is used because it has interfaces for all the peripherals needed for this system. Plus, it allows different implementation approaches, hence, offering more flexibility for hardware/software design. Team structure and responsibilities: Ealaf Hussein Hardware/Software interfacing System Integration with VR Maroua Filali Hardware implementation of the segmentation algorithm using different HW/SW co-design approaches Salma Shalaby Developing the algorithm in MATLAB. Evaluation using different Evaluation metrics. And objective and Subjective Evaluation Risks and Contingency plans: Risk 1: -A robotic arm was to be integrated with the project. Contingency Plan 1: -The robotic arm will be emulated in Virtual Reality VR. Risk 2: -3D construction of the segmentation is not accurate using MATLAB functions. Contingency Plan 2: -Utilize the Virtual Reality for accurate Visualisation of the segmentation. Project plan/schedule: Senior Design 1 (SEP 18th DEC 15th 2016) : - Literature review and related work including: o clinical problem o Image segmentation o SoC Zedboard o Robotic Surgery o Virtual Reality - Develop and simulate the algorithm Senior Design 2 (FEB 12th MAY 25th 2017): - Implement algorithm on hardware (Zedboard). - Testing and Verification. - Evaluation. 4/5

5 Expected Deliverables The expected deliverables for this project is automated medical image segmentation algorithm for cerebral aneurysm with the help of the ZYNQ system on chip (SoC). In addition, the constructed 3D volume of the segmentation will be demonstrated in an interactive environment using Virtual Reality lab to be used for training purposes. Submission in PDF DELL_abstract_AneurysmSegmentation-segmentationFINALSTAMP.pdf How did you hear about the competition? University Professor or Staff 5/5