Amrita School of Engineering
Research Projects at CSE, Coimbatore Campus
A few projects have been initiated in the areas of multimedia indexing and searching, multimedia content adaptation and social computing in education. The first project is being started as part of the Amrita E-Learning Research center, and aims to provide a context based indexing and mining tool for video lectures.
The other project, started by a group of faculty in the CS department, explores the use of social computing tools in higher education, specifically for student projects. After studying the feasibility of using social computing tools in the university setup, a suitable system will be designed and deployed at different phases. An initial prototype has been developed and is being deployed for testing.
Faculty are also working on other research areas like Spatial Temporal Databases, Evolutionary Computing and Data Mining
AIMS Projects
The following Projects are developed in-house by our M.Tech.(CVIP) students under the guidance of faculty members to enable a user-friendly environment. It acts as a 2-tier platform where doctors and patients utilize the system effectively and faculty – students get the better utilization of knowledge sharing and resource media. It also helps in digitizing the hospitalization process easier.
Remote Clinical Monitoring
In this project, we were able to develop a sensor based system which can communicate at regular intervals via a normal mobile phone to another phone through the existing GSM network. At the client side, we have a temperature sensor which is connected to the PIC microcontroller which in turn is connected to the phone through an RS232 converter. This device continuously monitors the patient and sends the body temperature values to the server side mobile at the hospital. This mobile is connected to our SMS server. The data received periodically is stored in a DB and can be interpreted from time to time into graphs and other pictorial representation. Next phase is a network application which is installed in the hospital. Each doctor will be allotted separate login id and password. He/She can login to our system anywhere from the hospital. In case of any undesirable variations, the doctor will be informed through this messenger system. The doctor has to login into his account to make the messenger active. At any point, if the doctor is not online then an alert is sent to his mobile. This product also has a web application using which the doctor can monitor the patient even while he is travelling. For testing purpose, we used only Temperature sensor. It has provisions to interface seven more sensors to monitor various critical parametersof a patient.
3D Reconstruction of Impacted Canines in Human Jaw
3D reconstruction is the process of capturing the shape and appearance of real objects. Construction of 3D model of Canine Impacted Human jaw helps in finding the position and orientation of the impacted tooth. This model helps the dentists in doing Canine Impacted Human jaw treatments easily. Complicated concepts which are difficult to visualize can be simplified with the help of 3D reconstruction. Thus the doctors can to an extent identify the impacted tooth before starting the treatment. Patients are also in a comfortable state if the dentist is able to locate canines at first attempt instead of doing multiple cuts to make localization. With the help of this technique dental students can make measurement and analysis for learning purpose.
Structural and Functional Atlas for MRI Brain Images
This work aims at creating a structural and functional atlas for MRI Brain images, which would help to solve clinical problems, faced during the training periods and can also be referred as a data set for medical diagnosis. The existing Atlas of Brain, provides only the Structural details, and for the clinicians functional details are more crucial for decision making.
Magnetic Resonance (MR) images of brain from normal person were used as the data set for formation of the Atlas. All the views of MRI brain images, i.e., axial, sagittal and coronal were considered. These images are subjected to segmentation and labeling, in order to tag the required structural components of the human Brain. These segmented images are stored in a data base with each segment being labelled by the corresponding part name. An interactive and user friendly GUI has been developed which would help the doctors/ clinicians to use the atlas for teaching or for decision making/diagnosis.
The three different views of the brain MR images are displayed in the initial window. From these a desired view is chosen according to the slice number and gets loaded. Using popup menu option all organs are highlighted in the present slice with corresponding name, if it is present in the loaded slice. The selected region is highlighted in different color. A synchronisation facility is provided that allows viewing the same part (selected in one of the view) in all the three views. The Functional circuit diagram and the functional details are displayed in a window when a particular organ is selected. The functional details are loaded from the text file associated with organ name from the data base and also the doctor can edit it to make changes if required.
For segmentation of the images various algorithms like Thresholding, Region Growing, Level Sets have been used. We have also tried MLS registration based segmentation approach. Also an interactive segmentation tool kit is developed, so as to enable the doctors to segment and label the images manually.
The data set for the atlas was acquired by taking the MRI Scan of a person, who was certified by the doctors at AIMS, Kochi to be a normal person without any abnormalities/ anomalies in the brain.
3D Reconstruction of Facial Structures From 2D Images
3D-operation planning which makes the preoperative visualization of patient’s postoperative appearance is one of the main applications in reconstructive and cranio-facial surgeries. The current preoperative surgical planning with respect to tissue changes is based mostly on frontal analysis of 2D information sources such as photographs and X-rays. It is necessary to develop an automatic guided software for performing cosmetic surgery by the process of simulating three dimensional visualization techniques suitable for the display of complex structures of the facial skeleton and of skull base.
This work proposes an approach to develop a suitable computer-based three dimensional visualization technique which can construct a three dimensional image from the given set of two-dimensional images.Three dimensional surface reconstruction of cranial anatomy is obtained from CT (Computer Tomographic ) or MRI(Magnetic Resonance Images) images.
The 2D slices or images which is the result of a horizontal or vertical scan of human head and face can be used to create the 3D image of the particular head. The 3D image can be then used as a model for surgery simulations. The cosmetic surgery can be simulated or performed first on the 3D model which can then be executed in real surgical environments.
The 2D slices which are in dicom format will be taken as the input for the system. The images will be then subjected to pre-processing of filtering to remove noises and segmentation based on a threshold to select the relevant data in case of necessities. The surgeon can browse and select a particular section or whole of dicom images from the given dataset using a dicom viewer by which he can check the whole information and data in the images. An iso-surface reconstruction algorithm based on marching cubes is used to develop a 3D model from a collection of 2D dataset of images. The iso-surfaces will be extracted and reconstructed to form a three dimensional model of the human head which will be a 3D volume which can be rotated and viewed from any angle by the surgeon .
2D Atlas of Human Heart for Educational Purposes
Human heart anatomy is the study of morphological structures and relationships between morphological structures. Hence such a study when visualized in the form of a two dimensional atlas can hold strong interactive sessions between doctors and medical students. Atlas is a visualization technique developed for understanding the morphological structures of a human body. Interacting directly with morphological structures contributes to a more comprehensive understanding of them in spatial relation to their surroundings.
The purpose of this project is to develop new interactive system, where the system provides a 2D atlas of the human heart in detail. Image segmentation is done to segment different parts of the morphological structures while performing structure analysis in the system. The proposed model was taken on the basis of the segmentation algorithm which gave the best resultant image. The segmentation algorithms used were K-Means, Watershed, Region Growing and Thresholding. Out of these four algorithms, region growing was considered to be the best because the algorithm was able to extract multiple regions of the heart. These extracted regions were then color masked for unique identification. Indexing and analyzing these 2D data in graphical user interface environment, enables an effective usage for designers and even final users. The idea of developing an atlas for the human anatomy is to provide a virtual environment that can facilitate surgical planning, virtual surgery, virtual endoscopy, and training simulations for doctors and medical students.
The final phase of this project is to detect the affected coronary arteries by using various automatic segmentation and edge detection method over the given set of CT images. From the results produced by various algorithms, the doctor will decide which algorithm has given a better output based on visual perception.
Localization of Lesions in Brachial Plexus
The brachial plexus is formed by the union of the fifth, sixth, seventh, and eighth cervical ventral rami and the first thoracic ventral ramus. The brachial plexus is a very important structure and can be injured through various types of trauma. The size and complexity of brachial plexus shows that neither single Nerve Conduction Study nor muscle assessable by Needle Electrode Examination is capable of evaluating it in its entirety. This projects aims at identifying the location of lesion in the brachial plexus.