33 research outputs found
Synthesizing Skin Lesion Images Using Generative Adversarial Networks
No full textAccording to the World Health Organization cancer is the second leading cause of death globally [1], and the most common cancer in the world is skin cancer [2]. To get a skin cancer diagnosis, a patient will usually mention their concern to a doctor. The doctor will then refer the patient to a dermatologist, who takes a closer look to determine whether the lesion is abnormal and in need of further inspection. If it’s abnormal, the next step is usually a biopsy of the lesion for further testing in a laboratory. The test results determine the final diagnosis. The dermatologist uses an array of different techniques and tools to examine the lesion, one of which is often a dermoscope. A dermoscope is a handheld device that makes it possible to see structures in the skin that normally are invisible to the naked eye. Nowadays dermoscopes are usually equipped with a camera, allowing the physician to digitally store images of the lesion. This makes it easier to get a second opinion. Either from another physician, or in the interest of this thesis, from some state-of-the-art automated skin cancer detection system, i.e. computer-aided diagnosis (CAD). The idea of having CAD tools act as second opinions has become more plausible than ever, with the recent improvements in machine learning (ML) seen in the last decade. Most importantly, the last years developments within the sub-fields of neural network (NN), deep learning (DL), and computer vision. This have caused plenty of new startups, companies, software, and applications for medical image analysis. In the subject of skin cancer, there has been a great deal of activity around the field of machine learning, ranging from simple mobile applications to comprehensive research into clinical utility. In 2017, Andre Esteva et al. published a paper which showed that machine learning models can perform at the level of dermatologists [3]. In 2018, H. A. Haenssel et al. constructed a convolutional neural network that in most cases outperformed 58 dermatologist. They claimed that any physicians, regardless of experience, may benefit from including a machine learning model in their evaluation [4]. A main reason for the success of neural networks is their ability to automatically extract useful features from data (representation learning). Data can also be said to be the methods greatest weakness: the performance of the model becomes extremely dependent on the quantity and quality of the data its being fed. The problem of acquiring high-quality data is particularly challenging when working with medical data. Because of the highly restrictive privacy protection of sensitive patient data, the resources needed for correctly labeling it, and the highly imbalanced nature of medical data, the process of collecting medical data is both difficult and expensive. In this thesis, we will try to reduce the problem of low amounts of data in machine learning by artificially creating more data using generative models, more specifically Generative Adversarial Network (GAN). We will create synthetic images of skin lesions, expanding the original data set. In addition to generating visually accurate synthetic data, our goal is to improve machine learning models by adding synthetic training data
Synthesizing skin lesion images using CycleGANs – a case study
No full textGenerative adversarial networks (GANs) have seen some success as a way to synthesize training data for supervised machine learning models. In this work, we design two novel approaches for synthetic image generation based on CycleGANs, aimed at generating realistic-looking, class-specific dermoscopic skin lesion images. We evaluate the images’ usefulness as additional training data for a convolutional neural network trained to perform a difficult lesion classification task. We are able to generate visually striking images, but their value for augmenting the classifier’s training data set is low. This is in-line with other researcher’s investigations into similar GAN models, indicating the need for further research into forcing GAN models to produce samples further from the training data distribution, and to find ways of guiding the image generation using feedback from the ultimate classification objective.publishedVersio
Synthesizing skin lesion images using CycleGANs – a case study
No full textGenerative adversarial networks (GANs) have seen some success as a way to synthesize training data for supervised machine learning models. In this work, we design two novel approaches for synthetic image generation based on CycleGANs, aimed at generating realistic-looking, class-specific dermoscopic skin lesion images. We evaluate the images’ usefulness as additional training data for a convolutional neural network trained to perform a difficult lesion classification task. We are able to generate visually striking images, but their value for augmenting the classifier’s training data set is low. This is in-line with other researcher’s investigations into similar GAN models, indicating the need for further research into forcing GAN models to produce samples further from the training data distribution, and to find ways of guiding the image generation using feedback from the ultimate classification objective
Lars Amund Vaage's sorrowsong read with Lacan
No full textMasteroppgåve i nordisk litteratur Institutt for lingvistiske, litterære og estetiske studium Universitetet i Bergen Våren 2018 Student: Sondre Johan Chesney Kvamme Rettleiar: Christine Hamm Tittel: Lars Amund Vaage sin sorgsong lesen med Lacan Undertittel: Ei psykoanalytisk lesing av Den framande byen (1999), Syngja(2012), og Sorg og song (2016) I denne masteroppgåva analyserer eg Lars Amund Vaage sin sorgsong i dei tre bøkene Den framande byen (1999), Syngja (2012) og Sorg og song (2016). Eg tolkar og analyserer forteljarteknikken og strukturen i bøkene i dialog med Lacan sin psykoanalyse, med eit særleg fokus på repetisjonar og erindringar. I oppgåva argumenterer eg for at det Vaage kallar for «sorgsong», er ein metode for å omarbeida og behandla eit traume. Gjentakingar av minne er ein sentral funksjon til sorgsongen for at subjektet skal kunne handtera traumet. I Den framande byen fylgjer me psykoanalytikaren Wilhelm Reich på reisa hans til Noreg. Romanpersonen er basert på den verkelege Wilhelm Reich. I Syngja fylgjer lesaren ein eg-forteljar, som også er forfattar, som har eit autistisk barn, mykje basert på forfattaren Lars Amund Vaage sitt eige liv. Sorg og song er eit essay skrive i etterkant av Syngja, der forfattaren Lars Amund Vaage skriv om skriveprosessen knytt til Syngja. Eg analyserer dei tre tekstane strukturelt og narratologisk, og deretter tolkar eg dei i ljos av Lacan sine psykoanalytiske skrifter. Eg har funne ut at det er ein stor likskap mellom dei tre romanpersonane Reich, eg-forteljaren i Syngja og Vaage. Alle tre har falt tilbake til det som Lacan kallar Det imaginære stadiet. Som eit resultat av dette har dei blitt handlingslamma. Dei tre personane lir av ei form for nevrose lik den som Lacan diagnostiserer Hamlet med i si tolking av Hamlet (Lacan 1977). Eg har funne ut at tekstane kan bli forstått som ein type sorgsong, og at utfordringa ved å akseptera tav av ulik art har satt sitt preg på Vaage sine romanar og romanpersonar. Eg har også konkludert med at Wilhelm Reich sin psykoanalyse truleg har påverka Vaage sin litterære stil meir enn han sjølv trur. Strukturen på romanane er eit direkte resultat av denne påverknaden.MA thesis in Nordic Literature Department of Linguistic, Literary and Aesthetic Studies University of Bergen, May 2018 Student: Sondre Johan Chesney Kvamme Tutor: Christine Hamm Title: Lars Amund Vaage sin sorgsong lesen med Lacan [lit. Lars Amund Vaage’s sorrowsong read with Lacan] Subtitle: Ei psykoanalytisk lesing av Den framande byen (1999), Syngja(2012), og Sorg og song (2016) [lit. A psychoanalytical reading of Den framande byen (1999, Syngja (2012) and Sorg og song (2016)] This MA thesis analyses Lars Amund Vaage’s «sorgsong» and its development in his three books Den framande byen (1999), Syngja (2012) and Sorg og song (2016). I interpret and analyse the narrative techniques and structures found in the books in dialog with Jacques Lacan’s psychoanalysis, with a focus on repetitions and recollections, or memories. In the thesis, I argue that the «sorgsong» is a method for treating and overcoming a trauma. Repetitions of memories is a central function in the “sorgsong”, as it attempts to overcome the trauma. Den framande byen is a novel about the psychoanalyst Wilhelm Reich, following him on his journey and stay in Norway. The fictional person is based on the real Wilhelm Reich. Syngja is a book about an author and his autistic daughter. The book is inspired by Vaage’s own life, as he also has a child with autism. Sorg og song is a book that reflects upon the writing processes that Vaage underwent whilst writing Syngja. In the thesis, I have found a striking similarity between the three persons Reich, the father in Syngja, and Vaage. They have all fallen back to Lacan’s Imaginary state and have thus lost parts of their language and ability to act. They all suffer from a form of neurosis similar to the one Lacan diagnoses Hamlet with in his analysis of Hamlet. (Lacan 1977) I have found that the three texts can be understood as a type of “sorgsong”. The challenges of accepting a loss, often a traumatic loss, has had a great impact on Vaage’s books and fictional persons. I have also concluded that Wilhelm Reich’s psychoanalysis possibly has affected Vaage’s literary style more than he is aware of himself. The way the books are structured is a direct result of this.Mastergradsoppgåve i nordiskNOLISP35
Kirkwood-Buff integrals from molecular simulation
No full textThe Kirkwood-Buff (KB) theory provides a rigorous framework to predict thermodynamic properties of isotropic liquids from the microscopic structure. Several thermodynamic quantities relate to KB integrals, such as partial molar volumes. KB integrals are expressed as integrals of RDFs over volume but can also be obtained from density fluctuations in the grand-canonical ensemble. Various methods have been proposed to estimate KB integrals from molecular simulation. In this work, we review the available methods to compute KB integrals from molecular simulations of finite systems, and particular attention is paid to finite-size effects. We also review various applications of KB integrals computed from simulations. These applications demonstrate the importance of computing KB integrals for relating findings of molecular simulation to macroscopic thermodynamic properties of isotropic liquids.Accepted Author ManuscriptEngineering Thermodynamic
Kollektivt auteurskap i Scorsese-filmen
No full textHvordan fungerer samarbeidet i Scorsese-filmen? Denne teksten bruker auteurteori for å undersøke auteurskap i filmene regissert av Martin Scorsese og diskuterer Thelma Schoonmakers mulige bidrag til dem. Til tross for deres langvarige samarbeid på over 40 år, blir alle filmene deres markedsført som "A Martin Scorsese Picture". Ved å analysere hva dette navnet egentlig betyr, argumenterer jeg for at det implisitt også refererer til henne.How is the nature of collaboration in "A Martin Scorsese Picture"? This text uses author theory to examine authorship in the movies directed by Martin Scorsese and discusses Thelma Schoonmaker’s possible contributions to them. Despite their collaboration of more than forty years, their films are presented as "A Martin Scorsese Picture". By deciphering what this name really means, I argue that it implicitly also refers to her
Effect of tunnel wash water on survival, growth and migration of Atlantic salmon (Salmo salar) and brown trout (Salmo trutta) in river Årungselva
No full textMaster Thesis, Department of Ecology and Natural Resource Management (INA) at the Norwegian University of Life Sciences (NMBU)Tunnel wash water is frequently released to the river Årungselva through a sedimentation pond. This tunnel wash water may cause harm to the fish through intrinsic and extrinsic toxic effects. A reduction in growth have previously been observed for 0+ brown trout at downstream locations of the sedimentation pond, where fish located below outlet point of the sedimentation pond had a 21 % lower length than fish located above the sedimentation pond. As no reduction in growth were observed prior to the establishment of the sedimentation pond, the author suggested that the reduced growth could be due to the toxic effect from the tunnel wash water. Since migration and density was not accounted for, it remains enigmatic whether the observed difference was solely due to the suggested toxic effect. The aim of this study was to estimate differences in survival, growth and migration of Atlantic salmon (Salmo salar) and brown trout (Salmo trutta) caught above and below the outlet point of the Vassum sedimentation pond in Årungselva. The study was conducted using capture-mark-recapture methodology in combination with Passive Integrated Transponders (PIT) telemetry with two antennas. This set-up allowed for estimation of survival, individual growth and migration. In total, 520 individuals were caught by electric fishing. Out of these, 253 individuals were PITtagged from which 75 were resighted at least once during the November 2014-October 2015 study period. The results show a lower size-adjusted survival in both species for individuals caught below the outlet point compared to those caught above the outlet point. Further, a lower length-at-age among 0+ parr of brown trout and 1+ parr of Atlantic salmon, as well as lower length at first-winter for Atlantic salmon parr, were observed for below-individuals compared to above-individuals. The reduced survival and growth rate observed in this study occur despite the fact that fish density is lower at below-sites of the outlet point of the sedimentation pond. Water chemistry variables generally did not vary between above and below sites, apart from chloride, sulphate and uranium that all attained higher below-values. However, other physiochemical variables that vary between above and below outlet point sites may influence growth and survival. In conclusion, fish at below-sites of the outlet point may experience a higher exposure to pollution due to the release of tunnel wash water based on these results.Norwegian Public Roads Administratio
The nature of resources for tourism and recreation
No full textThe first part of the article is devoted to discussing the concept of resources and the professional
approach to this phenomenon. The author goes on to put into effect the traditional theory which
divides resources into land, work force and capital. A central point in the article is that resources
consist not only of material things and conditions (concrete resources) but equally of nonmaterial
conditions (abstract/non-physical conditions) and that many resources are culturally
determined.
Then follows a section dealing with the relationship between resource quality, knowledge and
awareness on the one hand and competitive power on the other. The concepts of knowledge and
awareness are key factors, but must also be further divided into the knowledge and awareness
imparted by schools and universities as opposed to the experience and skills rooted in tradition.
The facts and ideas taught in our schools and universities can be termed institutionalized
knowledge and are common throughout large areas and cultural spheres. The experience and
skills rooted in tradition, however, are more likely to be confined to smaller areas. They are
transmitted and maintained by practical participation and can often be extremely important for
tourism.
Finally the question of resource analysis is addressed. The enclosed report is a description of an
analysis programme developed here at the college in Lillehammer and made use of by several
Norwegian municipalities and regions. Several of these analyses have been posted on the Internet by the college library but unfortunately they are all in Norwegian
Trajectory Following for Formula Student Driverless Vehicle - For Revolve NTNU
No full textThis master s thesis is written as part of a position in Revolve NTNU Team 2018, as part of the Driverless team. The author s area of responsibility in the project is Trajectory Following. The main objective is to design, implement and test a motion control system for the driverless vehicle, along with any peripherals required to interface with neighbouring systems. Additionally it is the author s responsibility to maintain and adapt the existing driver assistance control systems and the accompanying state estimations, in the transition to a driverless vehicle.
A simple minimum viable prototype system was implemented, consisting of a Line-Of-Sight guidance controller for lateral control and a PI feedback controller with a quadratic drag feedforward term for longitudinal control. This system was very robust at low speeds.
In addition, a high-performance system was implemented in software, but was never tested on the vehicle because the upstream modules never progressed to the point where this was a sensible course of action. This system consists of a nonlinear feedback linearization lateral controller, and a PI longitudinal controller with feedforward terms for aerodynamic drag and reference speed gradient. The interface to the vehicles tractive system control system was augmented to directly accept a reference yaw rate and a longitudinal force request, which are produced by the lateral and longitudinal controllers, respectively. A longitudinal force controller was implemented to handle the latter.
A parameter estimation algorithm for the Magic Formula tire model, using the gradient search method, was implemented in order to estimate the longitudinal and lateral tire-road forces and coefficients of friction, for use in the tractive system control system. This showed good results in simulations, but requires good estimates for the tire slip ratios, which are not readily available on the physical vehicle.
Everything presented in this thesis is the author's own work, unless stated otherwise. No other party was direct involved in the development. The relevant resources at disposal for the project was software licenses for MATLAB and IPG CarMaker, the tractive system control systems already implemented by Revolve NTNU students, including the modelling of Eld and the track used for simulations, in addition to the RC-car and the Gazebo simulation
