In tropical and subtropical countries with bad infrastructure there exists diseases which are often neglected and untreated. Some of these diseases are caused by parasitic intestinal worms which most often affect children severely. The worms spread through parasite eggs in human stool that end up in arable soil and drinking water. Over one billion people are infected with these worms, but medication is available. The problem is the ineffective diagnostic method hindering the medication to be distributed effectively. In the process of designing an automated microscope for increased effectiveness the solution for marking out the stool sample on the microscope slide is important for decreasing the time of diagnosis. This study examined the active contour model and four different semantic segmentation networks for the purpose of delineating the stool sample from the other parts of the microscope slide. The Intersection-over-Union (IoU) measurement was used to measure the performance of the models. Both active contour and the networks increased the IoU compared to the current implementation. The best model was the FCN-32 network which is a fully convolutional network created for semantic segmentation tasks. This network had an IoU of 95.2%, a large increase compared to the current method which received an IoU of 77%. The FCN-32 network showed great potential of decreasing the scanning time while still keeping precision of the diagnosis.

State-of-the-art fluorescent imaging research is strictly limited to eight fluorophore labels duringthe study of intercellular interactions among organelles. The number of excited fluorophore colorsis restricted due to overlap in the narrow spectra of visual wavelength. However, this requires aconsiderable effort of analysis to be able to tell the overlapping signals apart. Significant overlapalready occurs with the use of more than four fluorophores and is leaving researchers limited to asmall number of labels and the hard decision to prioritize between cellular labels to use. Except for the physical limitations of fluorescent labeling, the labeling itself causes behavioralabnormalities due to sample perturbation. In addition to this, the labeling dye or dye-adjacentantibodies are potentially causing phototoxicity and photobleaching thus limiting the timescale oflive cell imaging. Nontoxic imaging modalities such as transmitted-light microscopes, such asbright-field and phase contrast methods, are available but not nearly achieving images of thespecificity as when using fluorophore labeling. An approach that could increase the number of organelles simultaneously studied withfluorophore labels, while being cost-effective and nontoxic as transmitted-light microscopes wouldbe an invaluable tool in the quest to enhance knowledge of cellular studies of organelles. Here wepresent a deep learning solution, using convolutional neural networks built to predict thefluorophore labeling effect on the nucleus, from a transmitted-light input. This solution renders afluorescent channel available for another marker and would eliminate the process of labeling thenucleus with dye or dye-conjugated antibodies by instead using deep convolutional neuralnetworks. / Allra senaste forskningen inom fluorescensmikroskopi är begränsat till upp till åtta fluoroforer förstudier av intracellulära kommunikationer mellan organeller. Antalet fluorescerande färger ärbegränsade till följd av spektralt överlapp i det synliga våglängdsområdet. Överlappande signalerbehöver matematiskt bearbetas vilket innebär ökad arbetsinsats och signifikant överlappning skerredan vid användning av fler än fyra fluoroforer. Denna begräsning innebär i slutändan att forskarehar ett litet antal fluoroforer att arbeta med och behöver därmed prioritera vilka cellulära strukturersom kan märkas samtidigt. Utöver de spektrala begräsningarna med fluorescensmikroskopi, så innebär även själva färgningenav cellulära komponenter en negativ cellulär påverkan i form av avvikande beteende.Fluorescerande färgämnen och märkta antikroppar orsakar potentiellt fototoxicitet ochljusblekning, vilket begränsar tidsrymden vid studier av levande celler. Ljusfältsmikroskop sombright-field and faskontrast har inte en toxisk påverkan men producerar inte i närheten likadetaljerade bilder som fluorescensmikroskop gör. Ett tillvägagångssätt som skulle kunna öka antalet organeller som simultant kan undersökas medfluoroforer, som samtidigt är kostnadseffektiv och inte har en toxisk påverkan somljusfältsmikroskop, skulle vara ett ovärderligt verktyg för utökad kunskap vid cellulära studier avorganeller. Här presenteras en maskininlärningsmetod byggd med artificiella neuronnät för attpredicera fluorescerande infärgningen av cellkärnan i fluorescensmikroskop, med bilder frånljusfältsmikroskop. Denna lösning frigör en fluorofor som kan användas till andra organellersamtidigt som arbetet med fluorescerande infärgning av cellkärnan inte längre är nödvändigt ochersätts med ett artificiellt neuronnät.

In a world reliant on big industries to produce large quantities of more or less every product used, it is of utmost importance that the machines in such industries continue to run with minimum amounts of downtime. One way more and more providers of such industrial machines try to help their customers reduce downtime when a machine stops working or needs maintenance is through the use of remote guidance; a way of knowledge transfer from a technician to a regular employee that aims to allow the regular employee to be guided in real-time by a technician to solve the task himself, thus, not needing the technician to travel to the factory.  One technology that may come to mind if you were to create such a guiding system is to use augmented reality and maybe have a technician record his or her hand and in real-time overlay this upon the videostream the onsite employee sees. This is something available today, however, to separate the hand of the technician from the background can be a complex task especially if the background is not a single colour or the hand has a similar colour to the background. These kinds of limitations to the background separation are what this thesis aims to find a solution to. This thesis addresses this challenge by creating a test dataset containing five different background scenarios that are deemed representative of what a person who would use the product most likely can find something similar to without going out of their way. In each of the five scenarios, there are two videos taken, one with a white hand and one with a hand wearing a black glove. Then a machine learning model is trained in a couple of different configurations and tested on the test scenarios. The best of the models is later also tried to run directly on a mobile phone. It was found that the machine learning model achieved rather promising background segmentation and running on the computer with a dedicated GPU real-time performance was achievable. However, running on the mobile device the processing time proved to be not sufficient.

Automatic segmentation of the lumen of carotid arteries in ultrasound images is a starting step in providing preventive care for patients with atherosclerosis. To perform the segmentation this paper introduces a model utilizing a threshold algorithm. The model was tested with two different threshold algorithms, Otsu and Sauvola, then scored against professionally drawn masks. The scores were calculated with Dice and Jaccard-Needham as well as specificity, recall, and f1-score. The results showed promising mean and median similarity between the predictions and masks. Future work includes either optimizing the current model or augmenting it to give an even better ground to continue the work on providing preventive care for atherosclerosis patients.

The detection and early repair of road damage are paramount for the quality and safety of roads. Current detection efforts typically rely on Deep Learning methods for object detection with bounding boxes, with calculations performed on high-performance hardware. However, semantic segmentation can more accurately express the location of damages on the road, improving the descriptive quality of the detection. In addition, the use of lightweight networks to make these calculations potentially allows the technology to be run entirely on-site, without connecting to remote cloud services. The domain of road damage is inherently challenging. We select and evaluate several techniques for segmenting scarce and small objects; a common problem in traffic scene datasets containing road damage. To evaluate its effectiveness, the most promising method is applied to a new road damage dataset collected in Sweden. We present the results as an early foundation for future studies on segmenting road damage on mobile hardware. / Tidigt upptäckta vägskador och dess reparationer är av stor betydelse för vägens kvalitet och resenärens säkerhet. Aktuella detektionsmetoder förlitar sig vanligtvis på djupinlärnings metoder såsom objektigenkänning, där beräkningarna oftast utförs på högpresterande hårdvara. Med hjälp av semantisk segmentering från beräkningslätta neutrala nätverk kan metoderna förbättras, och potentiellt utföras på plats på mobila enheter utan nätverksankomst. Att segmentera vägskador medför en del utmaningar eftersom skadorna ofta är förhållandevis små och sällan förekommande i dataset. Ett urval av metoder för att bemöta dessa utmaningarna evalueras och den mest välpresterande tekniken tillämpas vidare på ett nyinsamlat dataset från Sverige. Vi presenterar resultatet som grund för framtida studier inom bildsegmentering, och vägskadadetektioner på mobil hårdvara.

The United States Forest Service (USFS) is developing software systems to evaluate forest resources with respect to qualities such as scenic beauty and vegetation structure. Such evaluations usually involve a large amount of human labor. In this thesis, I will discuss the design and implementation of a digital image segmentation system, and how to apply it to analyze forest images so that automated forest resource evaluation can be achieved. The first major contribution of the thesis is the evaluation of various feature design schemes for segmenting forest images. The other major contribution of this thesis is the development of a pattern recognition-based image segmentation algorithm. The best system performance was a 61.4% block classification error rate, achieved by combining color histograms with entropy. This performance is better than that obtained by an ?intelligent? guess based on prior knowledge about the categories under study, which is 68.0%.

No description available.

No description available.

No description available.

No description available.