https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/1694 2026-11-15T04:37:33Z https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/1772 Title: Ontology Based Approach for Diagnosis in Personalized Medicine Authors: Fernando, M.B.K.S. Abstract: Due to the complexity of the human body, overlapping phenotypes of patients and complex disease networks, disease diagnosis is a real challenge for the physicians. Medical diagnostic decision support systems like DXPLAIN, HELP, POSSUM, LDDB, Phenomizer etc have been developed as an approach to ease the process of disease diagnosis. Improved patient safety, improved quality of care and improved efficiency in health care delivery are potential benefits of MDDSS. This research proposes a novel and generic mathematical model into differen-tial diagnosis of genetic diseases instead of traditional method of analyzing gene mutations to expose genetic diseases. It is achieved by genotype-phenotype correlation through a common computer science concept called Ontology. Basically emerging genetic mutations of the patient are mapped to the standardized vocabulary called Human Phenotype Ontology (HPO) and subsequently differen-tial diagnosis is done using those terms. Differential diagnosis process is achieved by measuring ontology based semantic similarity by combining information theory and fuzzy relational theory. The system is capable of diagnosing the probability of occurrence of five complex diseases namely Lymphedema-Distichiasis Syndrome, Cornelia de Lange syndrome, Popliteal pterygium syndrome, Cohen Syndrome and Smith-Lemli-Opitz Syndrome. We evaluate our system by comparing the results obtained from our system with domain expert s diagnosis. Fourteen published pre-diagnosed set of real Cornelia de Lange syndrome patients data were used in this attempt. According to the results our system diagnoses Cornelia de Lange Syndrome with an average of 78.32% probability; Smith-Lemli-Opitz Syndrome has 61.35% probability and other three diseases with very low probability values. This thesaurus based appro-ach which considers a correlation of phenotypes and genotype of a patient can be used by a physician to make a better diagnosis of a disease. 0012-01-01T00:00:00Z https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/1773 Title: Digital Watermarking Based Video Device Authentication Authors: Fernando, W.K.B.A.K. Abstract: With the rapid growth of current information technologies, generating, editing and dis- tributing of digital multimedia data (audio, video, image, etc.) became increasingly trivial. Digital videos are in the front line when considering about this multimedia rev- olution. Replacement of classical analog videos with trendy digital videos, has made it easy to immediately capture visual feedbacks in to a video. Properly stored digital videos do not age and it is a proof of "this is what happened". But with the help of powerful video editing programs, it is very easy for an amateur to maliciously modify a digital video and create "perfect forgeries". Therefore it is dif- cult to get video records to stand up as 100% secure evidence in court. Even though videos are tending to get manipulated they are very important in many legal issues. Proving origin of a video is one of the important facts that should prove in a legal case where digital video considered as a silent witness such as in the case of a recording by a surveillance camcorder. It's a challenging task determine the origin of a video, as many malicious attackers can easily manipulate a video with the existing video editing tools. This project was conducted to nd a solution to this problem by incorporating video watermarking technology. A novel digital video device authentication approach based on secure self-embedding and real-time watermarking in the uncompressed domain is proposed to nd the origin of a piece of video footage. Real-time watermark embed- ding is achieved by combining all the processes related to watermark embedding, and by shortening the watermark embedding time by reusing the estimated watermark im- perceptibility among neighboring frames. Frequency domain watermarking is used for ensure the robustness. A prototype system is implemented to evaluate the proposed solution. 0012-01-01T00:00:00Z https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/1771 Title: Refining Vector Layers Using Satellite Images Authors: Fernando, W.V.A.H.R. Abstract: Vector layers represent features as line, polygons and points. Vector data is an essential compo- nent in a successful GIS as it involves in accurate and precise calculations of spatial measurements involving length, distance and area. Vector data is also capable of generating queries unable to solve using raster data. Such as which houses are within the 100 year ood level of a river, Which learners live in a particular suburb. Most common methods of creating vector data are digitizing of existing maps and satellite images, using data from eld surveys and using computer based algorithms. Accuracy of vector data will greatly depend on the scale of the map, resolution of the satellite image, e ciency of the algorithm, attention to details during digitizing and the ac- curacy of the instruments used in eld surveys. Hence, obtaining data from various data sources could lead to major compatibility issues over data accuracy and data integrity. This project is an e ort to develop a method to re ne existing vector layers using high resolution satellite images to create more accurate and meticulous vector layers. 0012-01-01T00:00:00Z https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/1764 Title: Augmenting the Real Composer on Automated Music Composition Authors: Wijesuriya, V.B. Abstract: Automated Music Composition (AMC) is an important aspect of the usage of com- puters in Music and it is extensively used in many commercial applications presently. Apart from instrumental music compositions, many of the other musical paradigms involve musical composition in nexus with a given set of lyrics. These compositions bear great resemblance to the lyrics provided and are sometimes composed solely based on the provided lyrics. Songs being one of the most famous examples of such types of music compositions, are greatly popular in the modern era. Many of the Automated Music Compositions techniques available today do not have the capability of correctly mapping structure and moods of lyrics to the com- position process. The author has built a computational model that would enable a potential composer to create musical compositions out of the lyrics using a set of controls, which is more robust compared to existing systems of AMC which do not take the features of the lyrics into account. The lyrics of a particular composition has a structure, contains di erent moods and encompasses rhythmic patterns or shapes which real composers take into account when setting a tune. The proposed model takes in these paradigms and automates the composition of a tune to a given set of lyrics thus augmenting the true composer by not only composing a sel ess tune but taking the real mood of an event into ac- count and tting up the composition to the event, thus giving the composition, a more natural avour. This work is delivered as a set of libraries. These have been developed on the Java platform. The outputs produced are in the form of MIDI les. The libraries can be used to design a host of applications. The generated compositions were subjected to validation by domain experts in the eld of music. The expert reposes shows that this work does generate good music compositions with a reasonable level of accuracy in theory of music, a good level of creativity and novelty. Objective of the work is to enable users to select their preferred composers and generate melodies for a given set of lyrics, thus making the user a potential composer though he might not possess all necessary skills needed by a professional composer. 0012-01-01T00:00:00Z