Brain tumors present the second highest cause of death among pediatric cancers. About 60% are located in the posterior fossa region of the brain; among the most frequent types the ones considered for this project were astrocytomas, medulloblastomas, and ependymomas. Diagnosis can be done either through invasive histopathology exams or by non-invasive magnetic resonance (MR) scans. The tumors listed can be difficult to diagnose, even for trained radiologists, so machine learning methods, in particular deep learning, can be useful in helping to assess a diagnosis. Deep learning has been investigated only in a few other studies.The dataset used included 115 different subjects, some with multiple scan sessions, for which there were 142 T2-w, 119 T1Gd-w, and 89 volumes that presented both MR modalities. 2D slices have been manually extracted from the registered and skull-stripped volumes in the transversal, sagittal, and frontal anatomical plane and have been preprocessed by normalizing them and selecting the slices containing the tumor. The scans employed are T2-w, T1Gd-w, and a combination of the two referred to as multimodal images. The images were divided session-wise into training, validation, and testing, using stratified cross-validation and have also been augmented. The convolutional neural networks (CNN) investigated were ResNet50, VGG16, and MobileNetV2. The model performances were evaluated for two-class and three-class classification tasks by computing the confusion matrix, accuracy, receiver operating characteristic curve (ROC), the area under the curve (AUROC), and F1-score. Moreover,  explanations for the behavior of networks were investigated using GradCAMs and occlusion maps. Preliminary investigations showed that the best plane and modality were the transversal one and T2-w images. Overall the best model was VGG16, for the two-class tasks the best classification was between astrocytomas and medulloblastomas which reached an F1-score of 0.86 for both classes on multimodal images, followed by astrocytomas and ependymomas with an F1-score of 0.76 for astrocytomas and 0.74 for ependymomas on T2-w, and last F1-score of 0.30 for ependymomas and 0.65 for medulloblastomas on multimodal images. The three-class classification reached F1-score values of 0.59 for astrocytomas, 0.46 for ependymomas, and 0.64 for medulloblastomas on T2-w images. GradCAMs and occlusion maps showed that VGG16 was able to focus mostly on the tumor region but that there also seemed to be other information in the background of the images that contributed to the final classification.To conclude, the classification of infratentorial pediatric brain tumors can be achieved with acceptable results by means of deep learning and using a single MR modality, though one might have to account for the dataset size, number of classes and class imbalance. GradCAMs and occlusion maps offer important insights into the decision process of the networks

Brain and nervous system tumors were responsible for around 250,000 deaths in 2020 worldwide. Correctly identifying different tumors is very important, because treatment options largely depend on the diagnosis. This is an expert task, but recently machine learning, and especially deep learning models have shown huge potential in tumor classification problems, and can provide fast and reliable support for pathologists in the decision making process. This thesis investigates classification of two brain tumors, glioblastoma multiforme and lower grade glioma in high-resolution H&E-stained histology images using deep learning. The dataset is publicly available from TCGA, and 220 whole slide images were used in this study. Ground truth labels were only available on whole slide level, but due to their large size, they could not be processed by convolutional neural networks. Therefore, patches were extracted from the whole slide images in two sizes and fed into separate networks for training. Preprocessing steps ensured that irrelevant information about the background was excluded, and that the images were stain normalized. The patch-level predictions were then combined to slide level, and the classification performance was measured on a test set. Experiments were conducted about the usefulness of pre-trained CNN models and data augmentation techniques, and the best method was selected after statistical comparisons. Following the patch-level training, five slide aggregation approaches were studied, and compared to build a whole slide classifier model. Best performance was achieved when using small patches (336 x 336 pixels), pre-trained CNN model without frozen layers, and mirroring data augmentation. The majority voting slide aggregation method resulted in the best whole slide classifier with 91.7% test accuracy and 100% sensitivity. In many comparisons, however, statistical significance could not be shown because of the relatively small size of the test set.

This thesis investigates the use of Convolutional Neural Networks (CNNs) toperform semantic segmentation of feet during endovascular surgery in patientswith Critical Limb Threatening Ischemia (CLTI). It is currently being investigatedwhether objective assessment of perfusion can aid surgeons during endovascularsurgery. By segmenting feet, it is possible to perform automatic analysis of perfusion data which could give information about the impact of the surgery in specificRegions of Interest (ROIs). The CNN was developed in Python with a U-net architecture which has shownto be state of the art when it comes to medical image segmentation. An imageset containing approximately 78 000 images of feet and their ground truth segmentation was manually created from 11 videos taken during surgery, and onevideo taken on three healthy test subjects. All videos were captured with a MultiExposure Laser Speckle Contrast Imaging (MELSCI) camera developed by Hultman et al. [1]. The best performing CNN was an ensemble model consisting of10 sub-models, each trained with different sets of training data. An ROI tracking algorithm was developed based on the Unet output, by takingadvantage of the simplicity of edge detection in binary images. The algorithmconverts images into point clouds and calculates a transformation between twopoint clouds with the use of the Iterative Closest Point (ICP) algorithm. The resultis a system that perform automatic tracking of manually selected ROIs whichenables continuous measurement of perfusion in the ROIs during endovascularsurgery.

Bakgrund: Uträkning av vänsterkammarens (VK) volymer (Enddiastolisk volym, EDV; Endsystolisk volym, ESV; Slagvolym, SV) och ejektionsfraktion (EF) går att göra med elektrokardiografi (EKG)-styrd gating vid positronemissionstomografi (PET) med spårämnet 15O-vatten. Metoden behöver utredas noggrannare och optimeras för att kunna introduceras i klinisk rutinverksamhet. Syfte: Syftet med denna studie var att undersöka bildanalys av PET-rekonstruktioner med olika spatial och temporal upplösning i samband med 15O-vatten-PET utförd med EKG-gating, samt jämföra analysutfallen av VK-volymer och EF mot CMR och sinsemellan, för att utreda möjligheten att optimera metoden. Metod: Totalt 25 patienter som genomgått en 15O-vatten-PET, varav n=11 hade undersökts med CMR samma dag, inkluderades. Olika gating-rekonstruktioner med varierande upplösning utfördes retrospektivt och analyserades automatiskt samt manuellt. Analysutfallen för VK-volymer och EF för PET och CM jämfördes statistiskt. Resultat: I studien fanns en stark till mycket stark korrelation mellan PET och CMR för EDV, stark korrelation för ESV, medel till stark korrelation för SV och svag till medel korrelation för EF. Rekonstruktion med 12 gating-bins och 256x256 matrisstorlek hade starkast korrelation för SV och EF. Samtliga PET-rekonstruktioner korrelerade starkt-till mycket starkt med varandra för VK-volymer och EF. Bland-Altman-analyser visade på en god repeterbarhet, framförallt vid manuell analys, för beräkning av EF med 15O-vatten-PET. Slutsats: VK-volymer och EF kan beräknas med 15O-vatten-PET med en repeterbarhet liknande den för andra, mer använda modaliteter. Att använda en högre upplösning än vad som tidigare testats gav högre värden för EF, och starkare korrelation i jämförelse mot CMR. / Background: Calculation of left ventricle (LV)-volumes (End Diastolic Volume, EDV; End Systolic Volume, ESV; Stroke Volume, SV) and ejection fraction (EF) is possible with electrocardiography (ECG)-gated Positron Emission Tomography (PET) using 15O-water, but the method needs to be further investigated and optimized before clinical routine implementation. Purpose: The purpose of this study was to investigate how altered image resolution affects the analysis and values of LV-volumes and ejection fraction on 15O-water-PET and compare the results against Cardiac Magnetic Resonance imaging (CMR) to enable optimization of the PET-method.  Method: In total, 25 patients who previously underwent a 15O-water-PET, where n=11 also performed CMR on the same day were included in the study. Different gating-reconstructions with varying resolution were performed retrospectively and underwent analysis, both automatically and manually.  Results: Correlation analysis found a strong to very strong correlation comparing PET against CMR for EDV, a strong correlation for ESV, a moderate to strong correlation for SV and a weak to moderate correlation for EF. The reconstruction containing 12 gating-bins and a 256x256 matrix size showed the strongest correlation for SV and EF. All PET-reconstructions correlated strong to very strong against each other for all LV-volumes and EF. Bland-Altman-plots showed good repeatability, especially for manual analysis, when calculating EF on 15O-water-PET.  Conclusion: LV-volumes and EF can be calculated on 15O-water-PET, with repeatability close to that of other modalities. Using an increased resolution than previously tested resulted in higher EF and stronger correlation in comparison with CMR.

Challenges within the field of pathology leads to a high workload for pathologists. Machine learning has the ability to assist pathologists in their daily work and has shown good performance in a research setting. Anomaly detection is useful for preventing machine learning models used for classification and segmentation to be applied on data outside of the training distribution of the model. The purpose of this work was to create an optimal anomaly detection pipeline for digital pathology data using a latent diffusion model and various image similarity metrics. An anomaly detection pipeline was created which used a partial diffusion process, a combined similarity metric containing the result of multiple other similarity metrics and a contrast matching strategy for better anomaly detection performance. The anomaly detection pipeline had a good performance in an out-of-distribution detection task with an ROC-AUC score of 0.90. / <p>Examensarbetet är utfört vid Institutionen för teknik och naturvetenskap (ITN) vid Tekniska fakulteten, Linköpings universitet</p>

In digital pathology, hematoxylin &amp; eosin (H&amp;E) is a routine stain which is performed on most clinical cases and it often provides clinicians with sufficient information for diagnosis. However, when making decisions on how to guide breast cancer treatment, immunohistochemical staining of human epidermal growth factor 2 (HER2 staining) is also needed. Over-expression of the HER2 protein plays a significant role in the progression of breast cancer and is therefore important to consider during treatment planning. However, the downside of HER2 staining is that it is both time consuming and rather expensive. This thesis explores the possibility for H&amp;E-to-HER2 stain translation using generative adversarial networks (GANs). If effective, this has the potential to reduce the costs and time spent on tissue processing while still providing clinicians with the images necessary to make a complete diagnosis. To explore this area two supervised (Pix2Pix, PyramidPix2Pix) and one unsupervised (cycleGAN) GAN structure was implemented and trained on digital pathology images from the MIST dataset. These models were trained two times, with 256x256 and 512x512 patches, to see which effect patch size has on stain translation performance as well. In addition, a methodology for evaluating the quality of the generated HER2 patches was also presented and utilized. This methodology consists of structural similarity index (SSIM) and peak signal to noise ratio (PSNR) comparison to the ground truth, and a HER2 status classification protocol. In the latter a classification tool provided by Sectra was used to assign each patch with a HER2 status of No tumor, 1+, 2+ or 3+ and the statuses of the generated patches were then compared to the statuses of the ground truths. The results show that the supervised Pyramid Pix2Pix model trained on 512x512 patches performs the best according to the SSIM and PSNR metrics. However, the unsupervised cycleGAN model shows more promising results when it comes to both visual assessment and the HER2 status classification protocol. Especially when trained on 256x256 patches for 200 epochs which gave an accuracy of 0.655, F1-score of 0.674 and MCC of 0.490. In conclusion the HER2 status classification protocol is deemed as a suitable way to evaluate H&amp;E-to-HER2 stain translation and thereby the unsupervised method is considered to be better than the supervised. Moreover, it is also concluded that a smaller patch size result in worse translation of cellular structure for the supervised methods. Further studies should focus on incorporating HER2 status classification in the cycleGAN loss function and more extensive training runs to further improve the quality of H&amp;E-to-HER2 stain translation.

Genom att generera mer information om hälsenan i form av dess elasticitet kan förhoppningsvis fler slutsater nås gällande diagnostik och behandling. Elastografi med hjälp av ultraljud skulle kunna vara en metod för att bidra med denna information. För att utföra detta anpassades en mjukvara utifrån ett grundläggande basprogram för elastografimätningar, utvecklat av företaget Verasonics, för att kunna utföra elastografi av hälsenan genom programmering i matlab. Tidigare undersökningsmetoder för elastografi är utvecklade för större organ, varför anpassningen innebar att använda metoder som även ger tillförlitlig information för mindre organ. För att göra detta anpassades först mjukvaran för en mindre fantom med liknande djup som hälsenan. När det konstaterats att skjuvningsvågor genererats på rätt avstånd kunde sedan mätningar göras på hälsenan. Genom att bestämma hastigheten av de genererade skjuvningsvågorna kunde sedan skjuvmodulen, följt av elasticitetsmodulen, beräknas för vävnaden. Denna bestämdes först genom grupphastigheten av skjuvningsvågorna, vilket är den metod som används vid större organ, följt av fashastigheten av skjuvningsvågorna vilket tar hänsyn till vågens dispersion. Detta gav då hälsenans elasticitetsmodul enligt grupphastighet samt fashastighet som sedan kunde jämföras. Slutligen gick det att konstatera att elasticitetsmodulen kommer att variera beroende på vilken typ av hastighet denna härleds från. Detta indikerar då på att sjuvningsvågen interagerar med organets gränsyta vilket orsakar dispersion. / Generating more information about the achilles tendon, such as its elasticity, will hopefully lead to more conclusions and results within both diagnostics as well as treatment. Elastography by ultrasound could be a method to contribute with this information. To do so, a basic software,provided and developed by the company Verasonics for elastography was specialized to fit the achilles tendon by programing in matlab. Earlier methods to perform elastography are developed for larger organs, hence the adjustment will include methods that acquire trustworthy information from smaller organs. To do so the adjustment of the software was first made to work on a smaller phantom with similar symmetry as the achilles tendon. When it was confirmed that shear waves were generated at the correct distance this enabled further measurements on the achilles tendon. By deciding the speed of the generated shear waves the shear modulus, followed by the elastic modulus, could then be estimated for the tissue. This was first decided by the group velocity of the shear waves, as the usual method done on larger organs, followed by the phase velocity that also takes dispersion in mind. The result could then be used to obtain the elastic modulus of the achilles tendon based on group and phase velocity for further comparison.The conclusion was then that the elastic modulus will depend on what kind of velocity it is derived from. This indicates that the shear wave interacts with the organ's boundaries which causes dispersion.

In the past twenty years, the combination of the advances in medical imaging technologies and therapeutic methods had a great impact in developing minimally invasive interventional procedures. Although the use of medical imaging for the surgery and therapy guidance dates back to the early days of x-ray discovery, there is an increasing evidence in using the new imaging modalities such as computed tomography (CT), magnetic reso- nance imaging (MRI) and ultrasound in the operating rooms. The focus of this thesis is on developing image-guided interventional methods and techniques to support the radiation therapy treatment of gynecologic cancers. Gynecologic cancers which involves malignan- cies of the uterus, cervix, vagina and the ovaries are one of the top causes of mortality and morbidity among the women in U.S. and worldwide. The common treatment plan for radiation therapy of gynecologic cancers is chemotherapy and external beam radiation therapy followed by brachytherapy. Gynecological brachytherapy involves placement of interstitial catheters in and around the tumor area, often with the aid of an applicator. The goal is to create an optimal brachytherapy treatment plan that leads to maximal radiation dose to the cancerous tissue and minimal destructive radiation to the organs at risk. The accuracy of the catheter placement has a leading effect in the success of the treatment. However there are several techniques are developed for navigation of catheters and needles for procedures such as prostate biopsy, brain biopsy, and cardiac ablation, it is obviously lacking for gynecologic brachytherapy procedures. This thesis proposes a technique which aims to increase the accuracy and efficiency of catheter placements in gynecologic brachytherapy by guiding the catheters with an electromagnetic tracking system. To increase the accuracy of needle placement a navigation system has been set up and the appropriate software tools were developed and released for the public use as a module in the open-source 3D Slicer software. The developed technology can be translated from benchmark to the bedside to offer the potential benefit of maximizing tumor coverage during catheter placement while avoiding damage to the adjacent organs including bladder, rectum and bowel. To test the designed system two independent experiments were designed and performed on a phantom model in order to evaluate the targeting accuracy of the tracking system and the mean targeting error over all experiments was less than 2.9 mm, which can be compared to the targeting errors in the available commercial clinical navigation systems.

Mammography screenings are performed regularly on women in order to detect early signs of breast cancer, which is the most common form of cancer. During an exam, X-ray images (called mammograms) are taken from two different angles and reviewed by a radiologist. If they find a suspicious lesion in one of the views, they confirm it by finding the corresponding region in the other view. Finding the corresponding region is a non-trivial task, due to the different image projections of the breast and different angles of compression needed during the exam. This thesis explores the possibility of using deep learning, a data-driven approach, to solve the corresponding regions problem. Specifically, a convolutional neural network (CNN) called U-net is developed and trained on scanned mammograms, and evaluated on both scanned and digital mammograms. A model based method called the arc model is developed for comparison. Results show that the best U-net produced better results than the arc model on all evaluated metrics, and succeeded in finding the corresponding area 83.9% of times, compared to 72.6%. Generalization to digital images was excellent, achieving an even higher score of 87.6%, compared to 83.5% for the arc model.

No description available.