Pinak Pujari

Projects

Reputation Systems (March 2006) (pdf)
Abstract: Transactions over the Internet with unknown users is a risky business. In today’s world of e-commerce relying just on the past historical transactional data or feedback does not entirely determine a player’s trustworthiness. Various other factors like – the trustworthiness of players who have given feedback; the kind of goods and amount involved in the transaction; the reputation of the player who gives feedback and one who receives feedback should ideally determine the extent to which the “feedback” given can damage or enhance the player’s reputation.
In this report I make an attempt to list down few incentives and heuristics which in theory should motivate a player to play honestly. Further, make dishonest players pay heavy penalty (in the form of decreasing their rating). The incentives are mainly economic – higher reputation favors better and more transactions.

Small Animal Radiotherapy Research Platform Cone Beam CT Reconstruction (May 2006) (pdf) (poster)
[ " Best Project 2006 " Award by the NSF Engineering Research Center for Computer-Integrated Surgical Systems and Technology ]

Goal of the Project:
To design and develop a software imaging system which does 3D volumetric reconstruction from 2D X-ray images of the tumor. This will provide the physician to plan the treatment. The problem is to develop a system which can help the physician to treat a cancer patient using radiotherapy with minimal damage to the healthy tissue around it.

2D images are acquired at small angle intervals, then reconstructed into a 3D volume using filtered back projection as described by Feldkamp algorithm. Any volumetric algorithm is inherently an O(n3) problem. In the case of Feldkamp and similar algorithms, the complexity is O(n3p), where p is the number of projection images used, and n is the dimension of the volume. The goal of this project was to provide a simple GUI for the imaging component, and to speed up image processing as much as possible to maximize the time available for treatment planning. In addition, viewing of the partial volume during reconstruction is desirable since it may allow the user to halt reconstruction early once the tumor region has been reconstructed with sufficient resolution.

Significance/features of this system:
The application developed will allow clinicians with limited or no computer expertise to operate the system in an intuitive and repeatable way. The system provides for calibration and error correction, including misalignment and simple distortion during image capture. The system provides dynamic intensity correction to eliminate image-to-image variations

The system, will allow automated acquisition and reconstruction of images, freeing researchers to develop treatment plans. The application will not only speed up data acquisition, but also allow for better characterization of any artifacts produced by the imaging system, correlating them with known calibration data. It also forms a critical link to other systems, including planning and PACS.

The system is intended to be intuitive to a non-technical user, providing the features expected in an modern image manipulation package as well as simple volume visualization and manipulation. Various aspects of the interface mimic those found in industry standard radiological packages, including the window and contrast manipulation controls, and the orthonormal controls for slice-by-slice viewing of the volume.

3D Reconstruction from Stereo Images (Dec. '05)

Design of privacy-preserving Data Warehouse, University of Hyderabad (April 2005)
Partial fulfillment towards the Bachelor's of Technology in Computer Science & Engineering.

Abstract: Organizations outsource data warehouse to a service provider. Since data is stored at the service provider (that may not be fully trusted) many new security and privacy challenges arise. The direct release of data invariably violates individual privacy. Thus, data must be processed before release in order to protect the privacy of the individuals they refer to. The objective of the proposed project is to design a data warehouse and implement different privacy-preserving algorithms. We are developing a system capable of disclosure-limited dissemination of tabular summaries of confidential micro data. The method forms part of a project developing a Web-based query system for statistical databases. A straightforward method for achieving individual privacy is to de-identify (anonymize) the data by replacing any explicit identifying information (e.g. name, date of birth) by some randomized values. This alone does not guarantee the full anonymity of data. We also propose to determine upper and lower bounds on the entries in multi-way tables of counts based on a set of released and possibly overlapping marginal tables which have practical importance for assessing disclosure risk. We attempt to do this by using the Frechet Bounds algorithm.

Research Paper

DNA Sequence alignment using AI techniques

Abstract: The sequence alignment problem is a very important problem in computational biology and bioinformatics. It has many applications in Human Genome project. The pair wise sequence alignment problem deals with alignment of two DNA sequences to find the best possible similarity. In this paper, we study in detail the techniques of pair wise sequence alignment problem. We give a brief account of the available algorithms such as dot-matrix method, dynamic programming method, divide and conquer algorithm, A* heuristics. We also discuss the time and space complexity of these algorithms. We, then propose a new algorithm based Artificial Intelligence techniques. We discuss the implementational issues of the algorithm. The algorithm is proposed based on a new idea and works very efficiently compared to other algorithms.

PINAK PUJARI

About Me

Courses

Fall 2005

Spring 2006

Fall 2006

Research Interest

Projects

Research Paper

DNA Sequence alignment using AI techniques

Contact Details