Medical molecular imaging with positron emission tomography (PET) is sensitive to patient motion since PET scans last several minutes. Despite advancements in PET, such as improved photon-pair time-of-flight (TOF) difference resolution, motion deformations limit image resolution and quantification. Previous research of head motion tracking has produced the data-driven centroid-of-distribution (COD) algorithm. COD generates a 3D center-of-mass (COM) over time via raw list-mode PET data, which can guide motion correction such as gating and event rebinning in non-TOF PET. Knowledge gaps: COD could potentially benefit from sinogram corrections used in image reconstruction, while rebinning has not extended to TOF PET. Methods: This study develops COD with event mass (incorporating random correction and line-of-response (LOR) normalization) and a simplistic TOF rebinner. In scans of phantoms and moving heads with F11 flouro-deoxy-glucose (FDG) tracer, COD alternatives are evaluated with a signal-to-noise ratio (SNR) via linear fit to image COM, while rebinning is evaluated with mean squared error (MSE). Results: COD SNR did not benefit from a corrected event mass. The prototype TOF rebinning reduced MSE, although there were discretization errors and event loss at extreme bins for LOR and TOF due to the simplistic design, which introduced image artifacts. In conclusion, corrected event mass in COD is not promising, while TOF rebinning appears viable if techniques from state-of-the-art LOR rebinning are incorporated.

The relationship between sEMG signals and muscle force, and associated joint torque, is an object of study for clinical applications such as rehabilitation robotics and commercial applications as wearable motion control devices. The information type and quality obtained by sEMG can impact the classification and prediction accuracy of ankle joint torque. In this thesis project, HD-sEMG based data was collected together with ankle joint torque measurements from 5 subjects during MVIC of plantarflexors and dorsiflexors. Machine learning approaches ideally suited for nonlinear regression tasks, such as MLP and LSTM, have been implemented and evaluated to best predict joint torque profiles given extracted features from sEMG data. An evaluation of machine learning performances using HD-sEMG data over bipolar sEMG data has been conducted in intra-session, inter-subjective and intra-subjective study cases.

Ice hockey players are at a high risk to sustain a concussion. Most of the concussion-inducing hits are to the jaw region, nevertheless, most players do not wear any protective gear shielding the jaw.  This parametric study used finite element simulations in LS-Dyna to evaluate the potential of foam materials in a jaw guard that could be attached to a helmet to reduce the concussion risk.  Here, it was investigated how nine different foam materials influence the ability of the jawguard to protect against concussion. Furthermore, aspects like foam thickness, shell thickness, and the impacting object were evaluated. In a second part, the formerly used HIII head model was exchanged with the KTH HBM, a FE model with a detailed representation of a jaw, and it was looked at how a movable jaw affects the head kinematics. Stiffer foams with a certain stress-strain behavior tend to aid best in energy absorption in the simulated crash scenarios and therefore lower the risk of sustaining a concussion. Impact angle and location influence the result significantly. Two simulated impacts show a decrease in concussion risk by up to 8.2\% and 6.9\% respectively when the jawguard was implemented, while the two other impacts resulted in an increase in concussion risk. Shell and foam thickness variation results depend mostly on the impact scenarios. However, it was found that a soft impactor helps tremendously in avoiding concussions. The hits on the KTH HBM tend to produce higher linear and angular accelerations but no significant difference is seen in angular velocity.  In conclusion, using stiff foams in ice hockey jawguards is a promising approach to attenuate impact energy caused by a collision during an ice hockey game. However, the effect of the jawguard on the concussion risk is very sensitive to the impact location and direction.

Ice hockey is a sport with high velocities and a large number of impacts to the head. The high occurrence of concussions is being recognized, and both short and long term consequences have been found. As body checking is the main situation inducing concussions, often in the form of shoulder-to-face impacts, there is thought to be a potential to lowering the rates of concussions with equipment that covers the jaw and chin area. In this study, in-game videos from the Swedish Hockey League were analyzed regarding impact situations and impact locations. The most occurring impact situations were simulated with finite element simulation on a Hybrid-III 50th percentile head and neck model wearing a standard ice hockey helmet. Three jawguard designs were developed and tested with the model, and seven different attachments were tested on two of the designs. The results showed varying effect of the jawguard, depending on the impact situation. In impacts to the side of the chin, all three designs reduced the strains in the brain, by successfully reducing the axial rotation. In impacts to the side of the face/head and to the front of the chin, the jawguard designs produced higher strains in the brain than without any protection. The helmet in this study was attached to the head model with a chin cup, and this may have had a significant effect on the strains of the brain. Other limitations to the simulation set up indicates that the jawguards should be tested on a more realistic model to properly evaluate the jawguard.

Hospital Acquired Infections and equipment contamination are some of the biggest issues faced by the healthcare industry worldwide. These infections generally range from mild to life threatening human infections which lead to increased costs and prolonged hospitalization time. The primary factor which caused these issues were biofilm forming bacteria which are able to withstand medications and defend themselves from various cleaning procedures. Another aspect which make these bacteria troublesome is that they are able to hide inside the biofilm, thus evading a lot of diagnostic tests. The methods used to detect these biofilms are unfortunately toxic to cells and cannot be used in vivo. Though there is enough data on the formation of biofilm on abiotic surfaces, the data present on the biophysical properties, structural organizations within the biofilm or their viscoelastic properties is very limited. In this master’s degree project, a dynamic monitoring platform is made for 2 different strains of the Salmonella Enteritidis bacteria where their structural and biophysical properties was investigated. Each strain lacks either curli or cellulose which are major components responsible for proper biofilm formation so performing these experiments on them gave us important information on how their properties get affected over time. Bacterial growth monitoring for all the strains were performed by measuring the absorbance every hour over a period of 5h and it was observed that all the strains had a very similar growth pattern until the end of the 4th hour after which they showed very mild differences. The next set of experiments involved using an eQCM to monitor the formation of biofilm on the surface of a quartz chip over 48h. There were differences observed in the biofilm formation pattern and mass deposition which provided clues as to how the biofilm formation and their viscoelastic properties are affected due to the mutations. This led to finding further clues regarding which aspect of biofilm formation is targeted by a specific mutant.

This master thesis investigates different approaches for integrating prior information into a neural network for segmentation of medical images. In the study, liver and liver tumor segmentation is performed in a cascading fashion. Context channels in the form of previous segmentations are integrated into a segmentation network at multiple positions and network depths using different integration strategies. Comparisons are made with the traditional integration approach where an input image is concatenated with context channels at a network’s input layer. The aim is to analyze if context information is lost in the upper network layers when the traditional approach is used, and if better results can be achieved if prior information is propagated to deeper layers. The intention is to support further improvements in interactive image segmentation where extra input channels are common. The results that are achieved are, however, inconclusive. It is not possible to differentiate the methods from each other based on the quantitative results, and all the methods show the ability to generalize to an unseen object class after training. Compared to the other evaluated methods there are no indications that the traditional concatenation approach is underachieving, and it cannot be declared that meaningful context information is lost in the deeper network layers.

Stasband används i samband med intravenösa stick inom vården. Ortrud Medical har utvecklat tryckindikerande stasband som används för att optimera förutsättningarna för ett lyckat stick. Banden är utformade med en sensor som påvisar när trycket runt armen befinner sig inom ett visst intervall. Som en del av Ortrud Medicals mättester av stasbandet, har det här examensarbetet haft som syfte att utveckla och utreda noggrannheten i alternativa testmetoder som på ett effektivt sätt kan undersöka bandets sensor. I arbetet utvecklades en testrigg för att undersöka sensorns mätnoggrannhet och vidare användes testriggen för att samla in information om sensorn. Riggens grundstruktur kunde framgångsrikt användas för att genomföra olika typer av tester. Såväl den framtagna riggen som de undersökta testmetoderna utgör bra underlag för vidareutveckling av en standardiserad testmetod med flera påbyggnadsmöjligheter.

Background: Lower limb amputees face many challenges, and most of them prefer to use prosthetics for daily tasks and activities. The prosthesis is usually a combination of connected prosthetic components, and their spatial orientation is called the prosthetic alignment. Proper alignment is essential, as it substantially affects the quality and comfort of a prosthesis.   Objective: The aim of this study was to create a method that could accurately and effectively quantify the alignment of a transtibial and transfemoral prostheses using Vicon optical motion capture system.   Methods: Two experimental series were conducted. The first one was to test the repeatability of the measurement. Three analysts placed retroreflective markers on the prostheses three times, and five measurements were recorded each time. Alignment parameters were calculated in Vicon ProCalc for each measurement, and a standard error of measurement was found for each alignment parameter. The standard error of measurement was calculated from three variance components, between-analyst, within-analyst, and between-trial variability. The second experimental series was conducted to understand the relationship between alignment adjustments and the outcome parameters. The socket height, internal rotation, flexion, adduction, and translation were modified and measured. The socket translation was calculated in three coordinate systems to study how they affect the outcome.   Results: For the first experimental series, the standard error of measurement for every alignment parameter was below 3° and 6 mm. The between-analyst variability was the most prominent, and the parameters calculated in the sagittal plane were more reliable than those calculated in the frontal and transverse plane. In the second experimental series, there was a linear relationship between the modifications and the measured outcome. When a connection between two prosthetic components was changed by turning the screws one round, the average change in angle between them was 2°, and the average translation change was 4.4 mm. Of the three coordinate systems, the translation calculated in ankle coordinates was more reliable than in global coordinates and describe the translation more effectively than in socket coordinates.   Conclusion: The reliability of the measurements was considered good. The standard error of measurement was low, and the main variability resulted from differences in marker placement between the analysts. The results from the measured alignment changes were as expected. All the parameters could be effectively interpreted, and the ankle coordinates were considered advantageous in describing the socket translation.

Independent on where in the world one is, patient safety is regarded as one of the most important aspects in the healthcare industry. On the contrary, depending on where you are, the patient safety will differ and is therefore location dependent. The patient safety in a developing country will therefore be evaluated in a different way compared to a developed country. This study, therefore aimed to identify the patient safety in Puntland, Somalia and with it, its healthcare environment in the hospitals. The goal was to identify the main factors that affected the patient safety. To investigate this, a field study to the region of interest was made and subsequently interviews with staff at the site were conducted as well as observations in the concerned hospitals. The obtained results were analysed using the method of Qualitative Content Analysis. At a later stage, the results could be thematized into four categories; “​Need​”, “​Device​”, “​Training​” and “​Knowledge​”, which pinpointed the main issues. The study show that there was a common transversal issue of a inherent lack of devices, training and knowledge which in turn could severely affect the patients and their safety in ways such as misdiagnosis, delayed treatment and in worst cases death. Furthermore, it was evident that rather than the lack of actual devices, the absence of knowledge was more prevalent. / Oberoende på var än i världen man befinner sig, anses patientsäkerhet vara en av de viktigaste aspekterna i sjukvården. Å andra sidan, helt beroende på var man befinner sig kommer patientsäkerheten skilja sig och är därför lägesberoende. Patientsäkerheten i ett utvecklingsland kommer därför uppfattas på ett annat sätt i jämförelse med ett I-land. Denna studie syftar till att identifiera patientsäkerheten i Puntland, Somalia och med det dess vårdmiljö i sjukhusen. Målet var att identifiera huvudfaktorerna som påverkar patientsäkerheten. För att undersöka detta utfördes en fältstudie i den valda regionen Puntland, därefter gjordes intervjuer med personal på plats i sjukhusen och dessutom utfördes observationer. De erhållna resultaten analyserades med hjälp av metoden “Qualitative Content Analysis”. Vid ett senare skede tematiseras resultaten till fyra kategorier; “​Behov​”, “​Apparat​”, “​Utbildning​” och “​Kunskap​”, vilka visade på de huvudsakliga problemen. Studien visade slutligen på att det fanns ett gemensamt genomgående problem av brist på apparater, utbildning och kunskap, vilket i sin tur skulle kunna påverka patienter och deras säkerhet på sätt såsom feldiagnoser, försenad behandling och i värsta fall döden. Vidare fastställdes att snarare än bristen på apparater, var avsaknaden av kunskap mer påtaglig.

Early detection is crucial for effectively treating glaucoma, a leading cause of irreversible blindness. Diagnosing glaucoma can be challenging due to its subtle early symptoms. This study aims to enhance glaucoma prediction by fine-tuning pre-trained convolutional neural networks. Several networks were re-trained and tested on publicly available retinal image datasets. Additionally, the models were evaluated on fundus images from patients at Region Västernorrland (RVN). The methodology involved exploring how to effectively process and prepare patient data for prediction purposes. The results showed that a majority voting ensemble of the fine-tuned models produced the highest performance, achieving an accuracy of approximately 0.94, with a specificity and sensitivity of 0.97 and 0.90 respectively. The ensemble also identified 0.90 glaucomatous images from RVN correctly. In terms of specificity and sensitivity, all models outperformed the results of ophthalmologist specialists described in a previous study. The findings suggest the effectiveness of transfer learning in enhancing the diagnostic accuracy of glaucoma. It also underscores the importance of proper storage and preparation of medical data for developing predicitive machine learning models. / Glaukom, mer känt som grön starr, är en av de vanligast förekommande ögonsjukdomarna som orsakar blindhet. Det är viktigt att diagnostisera glaukom tidigt i sjukdomsförloppet för att genom behandling, sakta ner eller stoppa ytterligare synförlust. Att diagnostisera glaukom kan vara utmanande, eftersom det vanligtvis inte visar några tidiga symtom. Artificiell intelligens (AI), eller mer specifikt maskininlärning (ML), kan hjälpa läkare att ställa rätt diagnos om det används som ett beslutsstöd. Faltande neurala nätverk (convolutional neural network, CNN) kan lära sig att känna igen mönster i bilder, för att därigenom klassificera bilder till olika kategorier. Ett sätt att diagnostisera glaukom är att studera näthinnan och synnerven i ögats bakre del, som kallas ögonbotten. I denna studie finjusterades redan tränade CNN:s för att prediktera glaukom utifrån ögonbottenbilder. Detta uppnåddes genom att träna om modellerna på publikt tillgängliga ögonbottenbilder. Målet var att jämföra nätverkens noggrannhet på en delmängd av bilderna, samt att evaluera dem på ögonbottenbilder från sjukhus i Region Västernorrland (RVN). För att uppnå detta ingick det även i metodiken att utforska begränsningarna och möjligheterna med hur patientdata får användas, samt att undersöka hur datat bör lagras och tillrättaläggas för att möjliggöra utvecklingen av prediktionsmodeller. Syftet med studien var att öka noggrannheten vid diagnostisering av glaukom. Resultaten visade att en ensemble baserad på majoritetsröstning av alla modeller gav den bästa noggrannheten, ungefär 0.94. Sensitiviteten och specificiteten var 0.90, respektive 0.97. Vidare klassificerades 90% av ögonbottenbilderna från RVN korrekt. Resultaten tyder på att maskininlärning är effektivt för att förbättra den diagnostiska noggrannheten för glaukom. Det understryker också vikten av strategisk lagring och förberedelse av medicinska data för att utveckla prediktiva maskininlärningsmodeller i framtiden.