Examensarbeten för masterexamen // Master Theses
Länka till denna samling:
Browse
Browsar Examensarbeten för masterexamen // Master Theses efter Program "Applied physics (MPAPP), MSc"
Visar 1 - 5 av 5
Sökresultat per sida
Sortera efter
- PostA Model for an Augmented Reality Tool in Tumour Removal Laparoscopic Surgery(2020) Månsson, Lisa; Chalmers tekniska högskola / Institutionen för matematiska vetenskaper; Lundh, Torbjörn; Modin, Klas; Lundh, TorbjörnOne of the most lethal types of cancer is hepatocellular carcinoma, cancer in the liver. Because of the many risks entailed with open surgery, the use of laparoscopic surgery has increased, with no exception in liver resections. Instead of a big cut, minimally invasive techniques are used, placing small ports on the abdomen where surgical tools as well as a laparoscope can be inserted. The surgeons orient themselves from the outside, creating a perception of the inside through the 2D images from the camera and a preoperative 3D image on the side as a map. Since the liver is an essentially homogeneous organ, it can be hard to orient from this information, why surgeons in Gothenburg have developed markers to place on the liver’s surface. With help of such markers, the goal is to develop an augmented reality tool for intraoperative guidance, mapping the laparoscopic 2D image to the corresponding 3D position, and be able to project a tumour area in the laparoscopic view. In this work, an inventory of laparoscopic liver resection was performed and a camera model as well as a simulated liver environment was developed. To map a 2D image to the 3D environment, an algorithm for estimation of the camera pose, named POSIT, was examined. It was concluded that the limitations of POSIT were not compatible with the problem, but the platform developed, consisting of a simulated liver environment and the camera model, can be used as a framework for future work, in which testing of other algorithms to estimate the camera pose can be performed.
- PostApplication Classification - Feature Selection and Classification of Vehicle Applications from Usage data(2016) Herder, Björn; Chalmers tekniska högskola / Institutionen för matematiska vetenskaper; Chalmers University of Technology / Department of Mathematical Sciences
- PostClustering Cancer Tumours using Unsupervised Deep Learning Techniques(2016) Lilja, Oskar; Chalmers tekniska högskola / Institutionen för matematiska vetenskaper; Chalmers University of Technology / Department of Mathematical SciencesThe modern technology of DNA microarrays has made high-dimensional genomic data available for large-scale analysis. This thesis investigates how unsupervised deep learning techniques may be used as a class discovery method analysing cancer tumour data. Furthermore, the possibility of inferring which genes most strongly contribute in the differentiation of cancer types is discussed. Gene expression data from The Cancer Genome Atlas of 10 different cancer tumour types are analysed. A deep autoencoder network clearly separates cancer tumours as well as known subtypes of tumours already in 2-dimensions. The results are compared with other dimensionality reduction methods like principal component analysis.
- PostEstimation of Ego Vehicle Motion from Lidar Point Cloud Data(2015) Richard, Lindén; Chalmers tekniska högskola / Institutionen för matematiska vetenskaper; Chalmers University of Technology / Department of Mathematical Sciences
- PostSwedish Dialect Classification using Artificial Neural Networks and Guassian Mixture Models(2017) Blomqvist, Viktor; Lidberg, David; Chalmers tekniska högskola / Institutionen för matematiska vetenskaper; Chalmers University of Technology / Department of Mathematical SciencesVariations due to speaker dialects are one of the main problems in automatic speech recognition. A possible solution to this issue is to have a separate classifier identify the dialect of a speaker and then load an appropriate speech recognition system. This thesis investigates classification of seven Swedish dialects based on the SweDia2000 database. Classification was done using Gaussian mixture models, which are a widely used technique in speech processing. Inspired by recent progress in deep learning techniques for speech recognition, convolutional neural networks and multi-layered perceptrons were also implemented. Data was preprocessed using both mel-frequency coefficients, and a novel feature extraction technique using path signatures. Results showed high variance in classification accuracy during cross validations even for simple models, suggesting a limitation in the amount of available data for the classification problems formulated in this project. The Gaussian mixture models reached the highest accuracy of 61.3% on test set, based on singe-word classification. Performance is greatly improved by including multiple words, achieving around 80% classification accuracy using 12 words.