Deformable image registration is usually performed manually by clinicians,which is time-consuming and costly, or using optimization-based algorithms, which are not always optimal for registering images of different modalities. In this work, a deep learning-based method for MR-CT deformable image registration is presented. In the first place, a neural network is optimized to register CT pelvic image pairs. Later, the model is trained on MR-CT image pairs to register CT images to match its MR counterpart. To solve the unavailability of ground truth data problem, two approaches were used. For the CT-CT case, perfectly aligned image pairs were the starting point of our model, and random deformations were generated to create a ground truth deformation field. For the multi-modal case, synthetic CT images were generated from T2-weighted MR using a CycleGAN model, plus synthetic deformations were applied to the MR images to generate ground truth deformation fields. The synthetic deformations were created by combining a coarse and fine deformation grid, obtaining a field with deformations of different scales. Several models were trained on images of different resolutions. Their performance was benchmarked with an analytic algorithm used in an actual registration workflow. The CT-CT models were tested using image pairs created by applying synthetic deformation fields. The MR-CT models were tested using two types of test images. The first one contained synthetic CT images and MR ones deformed by synthetically generated deformation fields. The second test set contained real MR-CT image pairs. The test performance was measured using the Dice coefficient. The CT-CT models obtained Dice scores higherthan 0.82 even for the models trained on lower resolution images. Despite the fact that all MR-CT models experienced a drop in their performance, the biggest decrease came from the analytic method used as a reference, both for synthetic and real test data. This means that the deep learning models outperformed the state-of-the-art analytic benchmark method. Even though the obtained Dice scores would need further improvement to be used in a clinical setting, the results show great potential for using deep learning-based methods for multi- and mono-modal deformable image registration.

Renal phase-contrast magnetic resonance imaging (PC-MRI) is an MRI modality where the phase component of the MR signal is made sensitive to the velocity of water molecules in the kidneys. PC-MRI is able to assess the Renal Blood Flow (RBF), which is an important biomarker in the development of kidney disease. RBF is analyzed with the manual or semi-automatic delineation by experts of the renal arteries in PC-MRI. This is a time-consuming and operator-dependent process. We have therefore trained, validated and tested a fully-automated deep learning model for faster and more objective renal artery segmentation. The PC-MRI data used in model training, validation and testing come from four studies (N=131 subjects). Images were acquired from three manufacturers with different imaging parameters. The best deep learning model found consists of a deeply-supervised 2D attention U-Net with residual skip connections. The output of this model was re-introduced as an extra channel in a second iteration to refine the segmentation result. The flow values in the segmented regions were integrated to provide a quantification of the mean arterial flow in the segmented renal arteries. The automated segmentation was evaluated in all the images that had manual segmentation ground-truths that come from a single operator. The evaluation was completed in terms of a segmentation accuracy metric called Dice Coefficient. The mean arterial flow values that were quantified from the auto-mated segmentation were also evaluated against ground-truth flow values from semi-automatic software. The deep learning model was trained and validated on images with segmentation ground-truths with 4-fold cross-validation. A Dice segmentation accuracy of 0.71±0.21 was achieved (N=73 subjects). Although segmentation results were accurate for most arteries, the algorithm failed in ten out of 144arteries. The flow quantification from the segmentation was highly correlated without significant bias in comparison to the ground-truth flow measurements. This method shows promise for supporting RBF measurements from PC-MRI and can likely be used to save analysis time in future studies. More training data has to be used for further improvement, both in terms of accuracy and generalizability.

The procedure of cardiographic magnetic resonance imaging requires patients to hold their breath for up to twenty seconds, creating an uncomfortable situation for many patients. It is proposed that an acquisition scheme using spiral trajectories is preferable due to their much shorter total scan time; however, spiral trajectories suffer from a blurring effect caused by off-resonance frequencies in the image area. There are several methods for reconstructing images with reduced blur and Conjugate Phase Reconstruction has been chosen as a method for implementation into Matlab-script for evaluation in regards to image reconstruction quality and computation time. This method finds a conjugate to the off-resonance from a field map to demodulate the image and an algorithm for frequency‑segmented Conjugate Phase Reconstruction is implemented along with an improvement called Multi-frequency Interpolation. The implementation is tested through simulation of spiral magnetic resonance imaging using a Shepp‑Logan phantom. Different off-resonance frequencies and field maps are used to provide a broad view of the functionality of the code. The two algorithms are then compared to each other in terms of computation speed and image quality. It is concluded that this implementation might reconstruct images well but that further testing on actual scan sequences is required to determine the usefulness. The Multi-frequency Interpolation algorithm yields images that are not useful in a clinical context. Further study of other methods not requiring a field map is suggested for comparison.

Background Positron emission tomography scanners can be used together with the PET tracer 15O-water to image perfusion in the body. It has mainly been used to image perfusion in the brain and heart. The new total-body PET scanners have a larger axial field-of-view, making imaging of the whole body possible without moving the patient. Objectives The objective of this study was to use data from a total-body PET scanner to map perfusion in the whole body all at once. Methods Cluster analysis was used to identify time-activity curves which were used in a compartment model describing the tracer kinetics. The compartment model equations were solved using linear regression and whole-body parametric images were generated. The results were validated using nonlinear regression. Result and conclusions The results showed that the method is promising but in need of adjustments. The method needs to be tested on more data and improvement in calculations for the lungs and for time delay is needed. With improvements, this method could be used for mapping perfusion in the whole body.

Purpose: Arterial stiffening poses an increased risk of cerebrovascular diseases, cognitive impairments, and even dementia as cardiac pulsations reach further into the brain causing white matter hyperintensities and microbleeds. Therefore it is of interest to obtain methods to estimate and map cardiac related pulsatility in the brain. Improvements of functional magnetic resonance imaging (fMRI) sequences is potentially allowing detection of rapid physiological processes in the echo-planar imaging (EPI) signal in the brainthrough a higher sampling rate. Specifically in this thesis, estimation and localization of cardiac pulsation and respiration is conducted through analysis of resting state data obtained with a multiband EPI sequence that permits whole brain imaging at a shorter repetition time (TR) than conventional EPI. The origin of these physiological signals are likely a mixture of inflow and compartment volume shifts during the cardiac- and respiratory cycles. As the amount of physiologically related signal in the multiband sequence used at the Biomedical Engineering Dept. R&D, Umeå University Hospital is unknown, the aim of this project is to find and map cardiac pulsatility and respiration for future research. Methods: Multiband fMRI data from 8 subjects was used, collected in a 3 Tesla scanner using a 32-channel head coil. The physiological signals were estimated through an algorithm that was developed to down-sample and temporally shift copies of simultaneous recordings of pulse and respiration. These signals were obtained using the scanner’s built-in pulse oximeter and a respiration belt. The shifted copies were voxel-wise, and slice by slice, correlated to the fMRI data using Pearson correlation. The time shift yielding maximum mean correlation within the brain was, for each slice, used to create statistical maps for significant voxels to show the localization and magnitude of correlation for cardiac pulsation andrespiration. Results: Many voxels around and nearby larger vessels and ventricles were highly correlated with the down-sampled, time shifted signals of the cardiac pulsation for all subjects. The cardiac pulsation maps resembled cerebral vasculature and were mostly localized around the Circle of Willis, brainstem, and the ventricles. Respiration signal was also highly correlated, and spatially located at the sides of the brain although mostly concentrated at the parietal- and occipital lobes. Conclusion: The results demonstrated that many voxels in the brain were highly correlated with cardiac pulsation and respiration using multiband EPI, and the statistical maps revealed distinct patterns for both of the physiological signals. This method and results for mapping cardiac related pulsatility, and respiration could be used for future research in order to better understand cerebral diseases and impairments, and alsoto improve fMRI filtering. Keywords: Arterial stiffness, Functional magnetic resonance imaging, Resting state, Multiband, CardiacPulsation, Respiration, Correlation analysis / Syfte: Arteriell förstyvning medför en ökad risk för cerebrovaskulära sjukdomar, kognitiva störningar och till och med demens då hjärtpulsationer når längre in i hjärnan orsakar vit materia hyperintensiteter och mikroblödningar. Av detta skäl är det därför av intresse att ta fram metoder för att estimera och kartlägga hjärtrelaterad pulsationer i hjärnan. Förbättringar av funktionella magnetresonanstomografi (fMRI) sekvenser kan möjliggöra detektering av snabba fysiologiska processer i den eko-planära (EPI) signalen i hjärnan genom en högre samplingsfrekvens. Specifikt i denna uppsats, utförs en skattning och lokalisering av hjärtpulsation och respiration genom analys av ’resting state’ data erhållet av en multiband-EPI sekvens som tillåter bildgivning av hela hjärnan med en kortare repetitionstid (TR) än konventionell EPI. Ursprunget avdessa fysiologiska signaler är sannolikt från en blandning av flöde- och volymsförändringar under hjärt- och respirationscyklerna. Då mängden av fysiologiskt relaterad signaler i multiband sekvensen, som används på Biomedicinska avdelningen, FoU Umeå Universitetssjukhust, är okänd så är målet med projektet att hitta och kartlägga hjärtpulsation och respiration för framtida forskning. Metod: Multiband fMRI data från 8 personer användes, insamlade från en 3 Tesla scanner med en 32-kanals huvudspole. De fysiologiska signalerna uppskattades genom en algoritm som utveckades för att sampla ned och tidsförskjuta kopior av simultant tagna signaler av puls och respiration. Dessa signaler samlades in med skannerns inbyggda pulsoximeter och andningsband. De förskjutna kopiorna var voxelvis, snitt för snitt, korrelerade med fMRI datat med användning av Pearson-korrelation. Det tidsskift somför varje snitt resulterade i maximal medelkorrelation i hjärnan användes för att skapa statistiska kartor, med endast signifikanta voxlar, för att visa var och hur mycket korrelation av hjärtpulsation och respiration som finns. Resultat: Många voxlar runt och nära större kärl och ventriklar var för alla personer starkt korrelerade medde samtidigt tagna, och tidsförskjutna signalerna av hjärtpulsation. Pulsationskartorna liknade cerebral vaskulatur och var mestadels lokaliserade kring Willis ring, hjärnstammen och ventriklar. Respirationssignalen var även starkt korrelerad och lokaliserad på sidorna av hjärnan, mestadels koncentrerat vid parietal- och occipital loberna. Slutsats: Resultaten visade att många voxlar i hjärnan var starkt korrelerade med hjärtpulsation och respiration vid användning av multiband EPI, och de statistiska kartorna avslöjade distinkta mönster för de båda fysiologiska signalerna. Den framtagna metoden och dess resultat för kartläggning av hjärtrelaterade pulsationer och respiration kan användas i framtida forskning i syfte att bättre förstå cerebrala sjukdomar och nedsättning, även för att förbättre fMRI filtrering. Nyckelord: Arteriell förstyvning, Funktionell magnetresonanstomografi, Resting state, Multiband, Hjärtpulsation, Andning, Korrelationsanalys

The aim of this thesis is to investigate an ultrasound transducer testing system calledProbeHunter and determine the limits of the device. This device requires custom configuratedfiles to adapt to and be applicable for new ultrasound probes. The task was to investigate theimprovement possibilities for the file creation and thus broaden the use of the system.The study shows that file creation is theoretically possible for certain models of probes. Ingeneral, there is a lack of files in Västmanland and that needs to improve to keep up with thelarge amounts of probes. Multiplexed array probes are, however, not possible to create files forat the state the system is in currently. / Introduktion: Sjukvården söker ständigt utveckling i sitt dagliga arbete för patientsäkerhet. Föratt uppnå detta behövs ett samarbete med externa företag som utvecklar system för att utföradessa kvalitetssäkringar. Medicinsk teknik i region Västmanland har under en längre tid varit påjakt efter en bra utrustning för test av ultraljudsprober och har använt sig av ett flertalundersökningsmetoder. ProbeHunter var en kandidat till flertal alternativ och kom att utnyttjasför att förbättra sjukvården. Syfte: Syftet med arbetet är att undersöka så många prob modeller med ProbeHunter sommöjligt. Identifiera vart gränserna med ProbeHunter går och därmed bredda på användningen avsystemet. Detta innebär inte nödvändigtvis att alla prober måste gå att testa men att man har gjorten rimlig procentökning. Metod: En mindre del av rapporten består av litteraturstudie tillsammans med mest praktisktarbete. Större del av rapportens källor för det primära arbetet består av intervjuer ochProbeHunters egna manualer. Som utgångspunkt hade det praktiska arbetet mycket fokus på atttesta så många prober som möjligt med systemet för att uppfylla de krav som var satta på arbetet. Resultat: Detta projekt gav möjligheten till en ökning av antal prober som kan testas på 11% avde 240 prober som finns i Västmanland sjukhus. Där 11% (= 27 prober) motsvarar de tester somhar blivit godkända. Medan de totalt skapade och editerade probespecifika filerna motsvara testav 27% (=65 prober av 240). Dessa filer inkluderar både fungerande och icke fungerande filer. Slutsats: ProbeHunter behöver förbättras ytterligare i några aspekter. Det är inte på något sättdet perfekta instrumentet men lyckas ändå ge bra resultat om rätt material är närvarande vidtestning. Aktiv utveckling krävs för att eliminera vissa nackdelar med systemet. ProbeHunter ärfortfarande en bra konkurrent till andra liknande system och kan komma att bli ännu bättre.

Advances of magnetic resonance imaging (MRI) techniques enable visualguidance to identify the anatomical target of interest during the image guidedintervention(IGI). Non-rigid image registration is one of the crucial techniques,aligning the target tissue with the MRI preoperative image volumes. As thegrowing demand for the real-time interaction in IGI, time used for intraoperativeregistration is increasingly important. This work implements 3D diffeomorphicdemons algorithm on Nvidia GeForce GTX 1070 GPU in C++ based on CUDA8.0.61 programming environment, using which the average registration time hasaccelerated to 5s. We have also extensively evaluated GPU accelerated 3D diffeomorphicregistration against both CPU implementation and Matlab codes, and theresults show that GPU implementation performs a much better algorithm efficiency.

Ultrasound imaging is a safe and popular imaging technique that relies on received radio frequency (RF) echos to show the internal organs and tissue. B-mode (Brightness mode) is the typical mode of ultrasound images generated from RF signals. In practice, the real processing algorithms from RF signals to B-mode images in ultrasound machines are kept confidential by the manufacturers. The thesis aims to estimate the process and reproduce the same results as the Ultrasonix One ultrasound machine does using deep learning. 11 scalar parameters including global gain, time-gain-compensation (TGC1-8), dynamic range and reject affect the transformation from RF signals to B-mode images in the machine. Data generation strategy was proposed. Two network architectures adapted from U-Net and Tiramisu Net were investigated and compared. Results show that a deep learning network is able to translate RF signals to B-mode images with respect to the controlling parameters. The best performance is achieved by adapted U-Net that reduces per pixel error to 1.325%. The trained model can be used to generate images for other experiments.

Malignant melanoma (MM) is a type of skin cancer that is associated with a very poor prognosis and can often lead to death. Early detection is crucial in order to administer the right treatment successfully but currently requires the expertise of a dermatologist. In the past years, studies have shown that automatic detection of MM is possible through computer vision and machine learning methods. Skin lesion segmentation and classification are the key methods in supporting automatic detection of different skin lesions. Compared with traditional computer vision as well as other machine learning methods, deep neural networks currently show the greatest promise both in segmentation and classification. In our work, we have implemented several deep neural networks to achieve the goals of skin lesion segmentation and classification. We have also applied different training schemes. Our best segmentation model achieves pixel-wise accuracy of \textbf{0.940}, Dice index of \textbf{0.867} and Jaccard index of \textbf{0.765} on the ISIC 2017 challenge dataset. This surpassed the official state of the art model whose pixel-wise accuracy was 0.934, Dice index 0.849 and Jaccard Index 0.765. We have also trained a segmentation model with the help of adversarial loss which improved the baseline model slightly. Our experiments with several neural network models for skin lesion classification achieved varying results. We also combined both segmentation and classification in one pipeline meaning that we were able to train the most promising classification model on pre-segmented images. This resulted in improved classification performance. The binary (melanoma or not) classification from this single model trained without extra data and clinical information reaches an area under the curve (AUC) of 0.684 on the official ISIC test dataset. Our results suggest that automatic detection of skin cancers through image analysis shows significant promise in early detection of malignant melanoma.

This thesis investigates the potential of federated learning (FL) in medical image analysis, addressing the challenges posed by data privacy regulations in accessing medical datasets. The motivation stems from the increasing interest in artificial intelligence (AI)research, particularly in medical imaging for tumor detection using magnetic resonance imaging (MRI) and computer tomography (CT) scans. However, data accessibility remains a significant hurdle due to privacy regulations like the General Data Protection Regulation (GDPR). FL emerges as a solution by focusing on sharing network parameters instead of raw medical data, thus ensuring patient confidentiality. The aims of the study are to understand the requirements for FL models to perform comparably to centrally trained models, explore the impact of different aggregation functions, assess dataset heterogeneity, and evaluate the generalization of FL models. To achieve these goals, this thesis uses the BraTS 2021 dataset, which contains 1251 cases of brain tumor volumes from 23 distinct sites, with different distributions of the data across 3-8 nodes in a federation. The federation is set up to perform brain tumor segmentation, using different forms of aggregationfunctions (FedAvg. FedOpt, and FedProx) to finalize a global model. The final FL models demonstrate similar performance to that of centralized and local models, with minor variations. However, FL models’ performance varies depending on the dataset distribution and aggregation method used. Additionally, this study explores the impact of privacy-preserving techniques, such as differential privacy (DP), on FL model performance. While DP methods generally result in lower performance compared to non-DP methods, their effectiveness varies across different data distributions, and aggregation functions.