A crucial step of the radiotherapy workflow is the segmentation of the tumors. Currently, this is done manually, which is very time-consuming and therefore puts a lot of burden in the clinical... Show moreA crucial step of the radiotherapy workflow is the segmentation of the tumors. Currently, this is done manually, which is very time-consuming and therefore puts a lot of burden in the clinical workflow. Deep learning techniques, currently state of the art for computer vision tasks, are a potential solution to speed up the process. Even though they have been applied to segment other structures of interest for the RT workflow, they still underperform for the case of tumors. Furthermore, there is limited research in automatic segmentation of tumors for the particular case of MRI, even though this imaging modality presents better soft tissue contrast and it is therefore ideal to visualize tumors. The goal of this thesis was to develop automatic segmentation techniques for tumors on MRI images that deliver clinically acceptable segmentations. The different automatic segmentation methods were applied in two different tasks: the automatic segmentation of the oropharyngeal primary tumor in multiparametric diagnostic MRI images and the automatic segmentation of the cervical cancer gross tumor volume in the MRI images of the brachytherapy treatment images. Show less
Cerebral amyloid angiopathy (CAA) is a cerebrovascular disease affecting the small arteries in the brain with hallmark depositions of amyloid-β in the vessel wall, leading to cognitive decline and... Show moreCerebral amyloid angiopathy (CAA) is a cerebrovascular disease affecting the small arteries in the brain with hallmark depositions of amyloid-β in the vessel wall, leading to cognitive decline and intracerebral hemorrhage (ICH). An emerging MRI marker for CAA is cortical superficial siderosis (cSS) as it is strongly related to the risk of (recurrent) ICH. Current assessment of cSS is mainly done on T2*- weighted MRI using a qualitative score consisting of 5 categories of severity which is hampered by ceiling effects. Therefore, the need for a more quantitative measurement is warranted to better map disease progression for prognosis and future therapeutic trials. We propose a semi-automated method to quantify cSS burden on MRI and investigated it in 20 patients with CAA and cSS. The method showed excellent inter-observer (Pearson’s 0.991, P < 0.001) and intra-observer reproducibility (ICC 0.995, P < 0.001). Furthermore, in the highest category of the multifocality scale a large spread in the quantitative score is observed, demonstrating the ceiling effect in the traditional score. We observed a quantitative increase in cSS volume in two of the 5 patients who had a 1 year follow up, while the traditional qualitative method failed to identify an increase because these patients were already in the highest category. The proposed method could therefore potentially be a better way of tracking progression.In conclusion, semi-automated segmenting and quantifying cSS is feasible and repeatable and may be used for further studies in CAA cohorts. Show less
MR fingerprinting (MRF) is a promising method for quantitative characterization of tissues. Often, voxel-wise measurements are made, assuming a single tissue-type per voxel. Alternatively, the... Show moreMR fingerprinting (MRF) is a promising method for quantitative characterization of tissues. Often, voxel-wise measurements are made, assuming a single tissue-type per voxel. Alternatively, the Sparsity Promoting Iterative Joint Non-negative least squares Multi-Component MRF method (SPIJN-MRF) facilitates tissue parameter estima-tion for identified components as well as partial volume segmentations. The aim of this paper was to evaluate the accuracy and repeatability of the SPIJN-MRF parameter estimations and partial volume segmentations. This was done (1) through numerical simulations based on the BrainWeb phantoms and (2) using in vivo acquired MRF data from 5 subjects that were scanned on the same week-day for 8 consecutive weeks. The partial volume segmen-tations of the SPIJN-MRF method were compared to those obtained by two conventional methods: SPM12 and FSL. SPIJN-MRF showed higher accuracy in simulations in comparison to FSL-and SPM12-based segmentations: Fuzzy Tanimoto Coefficients (FTC) comparing these segmentations and Brainweb references were higher than 0.95 for SPIJN-MRF in all the tissues and between 0.6 and 0.7 for SPM12 and FSL in white and gray matter and between 0.5 and 0.6 in CSF. For the in vivo MRF data, the estimated relaxation times were in line with literature and minimal variation was observed. Furthermore, the coefficient of variation (CoV) for estimated tissue volumes with SPIJN-MRF were 10.5% for the myelin water, 6.0% for the white matter, 5.6% for the gray matter, 4.6% for the CSF and 1.1% for the total brain volume. CoVs for CSF and total brain volume measured on the scanned data for SPIJN-MRF were in line with those obtained with SPM12 and FSL. The CoVs for white and gray mat-ter volumes were distinctively higher for SPIJN-MRF than those measured with SPM12 and FSL. In conclusion, the use of SPIJN-MRF provides accurate and precise tissue relaxation parameter estimations taking into account intrinsic partial volume effects. It facilitates obtaining tissue fraction maps of prevalent tissues including myelin water which can be relevant for evaluating diseases affecting the white matter. Show less
Neve, O.M.; Chen, Y.J.; Tao, Q.; Romeijn, S.R.; Boer, N.P. de; Grootjans, W.; ... ; Staring, M. 2022
Purpose: To develop automated vestibular schwannoma measurements on contrast-enhanced T1- and T2-weighted MRI scans.Materials and Methods: MRI data from 214 patients in 37 different centers were... Show morePurpose: To develop automated vestibular schwannoma measurements on contrast-enhanced T1- and T2-weighted MRI scans.Materials and Methods: MRI data from 214 patients in 37 different centers were retrospectively analyzed between 2020 and 2021. Patients with hearing loss (134 positive for vestibular schwannoma [mean age 6 SD, 54 years 6 12; 64 men] and 80 negative for vestibular schwannoma) were randomly assigned to a training and validation set and to an independent test set. A convolutional neural network (CNN) was trained using fivefold cross-validation for two models (T1 and T2). Quantitative analysis, including Dice index, Hausdorff distance, surface-to-surface distance (S2S), and relative volume error, was used to compare the computer and the human delineations. An observer study was performed in which two experienced physicians evaluated both delineations.Results: The T1-weighted model showed state-of-the-art performance, with a mean S2S distance of less than 0.6 mm for the whole tumor and the intrameatal and extrameatal tumor parts. The whole tumor Dice index and Hausdorff distance were 0.92 and 2.1 mm in the independent test set, respectively. T2-weighted images had a mean S2S distance less than 0.6 mm for the whole tumor and the intrameatal and extrameatal tumor parts. The whole tumor Dice index and Hausdorff distance were 0.87 and 1.5 mm in the independent test set. The observer study indicated that the tool was similar to human delineations in 85%-92% of cases.Conclusion: The CNN model detected and delineated vestibular schwannomas accurately on contrast-enhanced T1- and T2-weighted MRI scans and distinguished the clinically relevant difference between intrameatal and extrameatal tumor parts. (C) RSNA, 2022 Show less
Outeiral, R.R.; Bos, P.; Hulst, H.J. van der; Al-Mamgani, A.; Jasperse, B.; Simoes, R.; Heide, U.A. van der 2022
Background and purpose: Contouring oropharyngeal primary tumors in radiotherapy is currently done manually which is time-consuming. Autocontouring techniques based on deep learning methods are a... Show moreBackground and purpose: Contouring oropharyngeal primary tumors in radiotherapy is currently done manually which is time-consuming. Autocontouring techniques based on deep learning methods are a desirable alternative, but these methods can render suboptimal results when the structure to segment is considerably smaller than the rest of the image. The purpose of this work was to investigate different strategies to tackle the class imbalance problem in this tumor site.Materials and methods: A cohort of 230 oropharyngeal cancer patients treated between 2010 and 2018 was retrospectively collected. The following magnetic resonance imaging (MRI) sequences were available: T1 -weighted, T2-weighted, 3D T1-weighted after gadolinium injection. Two strategies to tackle the class imbal-ance problem were studied: training with different loss functions (namely: Dice loss, Generalized Dice loss, Focal Tversky loss and Unified Focal loss) and implementing a two-stage approach (i.e. splitting the task in detection and segmentation). Segmentation performance was measured with Sorensen-Dice coefficient (Dice), 95th Hausdorff distance (HD) and Mean Surface Distance (MSD). Results: The network trained with the Generalized Dice Loss yielded a median Dice of 0.54, median 95th HD of 10.6 mm and median MSD of 2.4 mm but no significant differences were observed among the different loss functions (p-value > 0.7). The two-stage approach resulted in a median Dice of 0.64, median HD of 8.7 mm and median MSD of 2.1 mm, significantly outperforming the end-to-end 3D U-Net (p-value < 0.05).Conclusion: No significant differences were observed when training with different loss functions. The two-stage approach outperformed the end-to-end 3D U-Net. Show less
Hesse, L.S.; Aliasi, M.; Moser, F.; Haak, M.C.; Xie, W.D.; Jenkinson, M.; ... ; INTERGROWTH-21st Consortium 2022
The quantification of subcortical volume development from 3D fetal ultrasound can provide important diagnostic information during pregnancy monitoring. However, manual segmentation of subcortical... Show moreThe quantification of subcortical volume development from 3D fetal ultrasound can provide important diagnostic information during pregnancy monitoring. However, manual segmentation of subcortical structures in ultrasound volumes is time-consuming and challenging due to low soft tissue contrast, speckle and shadowing artifacts. For this reason, we developed a convolutional neural network (CNN) for the automated segmentation of the choroid plexus (CP), lateral posterior ventricle horns (LPVH), cavum septum pellucidum et vergae (CSPV), and cerebellum (CB) from 3D ultrasound. As ground-truth labels are scarce and expensive to obtain, we applied few-shot learning, in which only a small number of manual annotations (n = 9) are used to train a CNN. We compared training a CNN with only a few individually annotated volumes versus many weakly labelled volumes obtained from atlas-based segmentations. This showed that segmentation performance close to intra-observer variability can be obtained with only a handful of manual annotations. Finally, the trained models were applied to a large number (n = 278) of ultrasound image volumes of a diverse, healthy population, obtaining novel US-specific growth curves of the respective structures during the second trimester of gestation. Show less
Adapting a radiotherapy treatment plan to the daily anatomy is a crucial task to ensure adequate irradiation of the target without unnecessary exposure of healthy tissue.This adaptation can be... Show moreAdapting a radiotherapy treatment plan to the daily anatomy is a crucial task to ensure adequate irradiation of the target without unnecessary exposure of healthy tissue.This adaptation can be performed by automatically generating contours of the daily anatomy together with fast re-optimization of the treatment plan. These measurescan compensate for the daily variation and ensure the delivery of the prescribed dose distribution at small margins and high robustness settings. In this thesis, we focused on developing a deep learning-based methodology for automatic contouring for real-time adaptive radiotherapy either guided by CT or MR imaging modalities Show less
A dedicated software package that could semi-automatically assess differences in aortic maximal cross-sectional diameters from consecutive CT scans would most likely reduce the post-processing time... Show moreA dedicated software package that could semi-automatically assess differences in aortic maximal cross-sectional diameters from consecutive CT scans would most likely reduce the post-processing time and effort by the physicians. The aim of this study was to present and assess the quality of a new tool for the semi-automatic quantification of thoracic aorta dilation dimensions. Twenty-nine patients with two CTA scans of the thoracic aorta for which the official clinical report indicated an increase in aortic diameters were included in the study. Aortic maximal cross-sectional diameters of baseline and follow-up studies generated semi-automatically by the software were compared with corresponding manual measurements. The semi-automatic measurements were performed at seven landmarks defined on the baseline scan by two operators. Bias, Bland-Altman plots and intraclass correlation coefficients were calculated between the two methods and, for the semi-automatic software, also between two observers. The average time difference between the two scans of a single patient was 1188 +/- 622 days. For the semi-automatic software, in 2 out of 29 patients, manual interaction was necessary; in the remaining 27 patients (93.1%), semi-automatic results were generated, demonstrating excellent intraclass correlation coefficients (all values0.91) and small differences, especially for the proximal aortic arch (baseline: 0.19 +/- 1.30mm; follow-up: 0.44 +/- 2.21mm), the mid descending aorta (0.37 +/- 1.64mm; 0.37 +/- 2.06mm), and the diaphragm (0.30 +/- 1.14mm; 0.37 +/- 1.80mm). The inter-observer variability was low with all errors in diameters1mm, and intraclass correlation coefficients all 0.95. The semi-automatic tool decreased the processing time by 40% (13 vs. 22min). In this work, a semi-automatic software package that allows the assessment of thoracic aorta diameters from baseline and follow-up CTs (and their differences), was presented, and demonstrated high accuracy and low inter-observer variability. Show less
The aim of this thesis is to develop image processing solutions that enable the fully automatic pre-operative planning of aorta-related procedures, such as Trans-catheter aortic valve... Show moreThe aim of this thesis is to develop image processing solutions that enable the fully automatic pre-operative planning of aorta-related procedures, such as Trans-catheter aortic valve replacement and aorta dilatation diagnosis. Hence, the objectives of this thesis are as follows: 1. To fully automatically quantify the aorto-iliac vascular access route, including the aortic root by image processing methods in CTA. 2. To broaden the scope of automatic methods into the detection of aorta dilatation. 3. To integrate the automatic quantification methods into applications which allow manual interactions and the calculation of clinically relevant parameters. 4. To demonstrate the accuracy and feasibility of the fully automatic planning and quantification methods in different patient cohorts. Show less
Image analysis of objects in the microscope scale requires accuracy so that measurements can be used to differentiate between groups of objects that are being studied. This thesis deals... Show more Image analysis of objects in the microscope scale requires accuracy so that measurements can be used to differentiate between groups of objects that are being studied. This thesis deals with measurements in yeast biology that are obtained through microscope images. We study the algorithms and workflow of image analysis of yeast cells in order to understand and improve the measurement accuracy. The Saccharomyces cerevisiae cell is widely used as a model organism in the life sciences. It is essential to study the gene and protein behaviour within these cells, and consequently making it possible to find treatment and solutions for genetic and hereditary diseases. This is possible since many processes that occurs at the molecular level in this organism are similar to those in human cells. In the research group Imaging and Bioinformatics, we have developed a framework for analysis of yeast cells. This framework is intended to serve as a support for research in yeast biology. The framework is integrated in one application and presented via a GUI. The application integrates modules and algorithms including segmentation, measurement, analysis and visualization. Show less
This thesis examines how computer software can be used to analyse medical images of an aseptically loosening hip prosthesis, and subsequently to plan and guide a minimally invasive cement injection... Show moreThis thesis examines how computer software can be used to analyse medical images of an aseptically loosening hip prosthesis, and subsequently to plan and guide a minimally invasive cement injection procedure to stabilize the prosthesis. We addressed the detection and measurement of periprosthetic bone lesions from CT image volumes. Post-operative CTs of patients treated at our institution were analysed. We developed tissue classification algorithms that automatically label periprosthetic bone, cement and fibrous interface tissue. An existing particle-based multi-material meshing algorithm was adapted for improved Finite Element model creation. We then presented HipRFX, a proof-of-concept software tool for planning and guidance during percutaneous cement refixation procedures. Show less
Rudyanto, R.D.; Kerkstra, S.; Rilowort, E.M. van; Fetita, C.; Brillet, P.Y.; Lefevre, C.; ... ; Ginneken, B. van 2014
T2*-weighted imaging provides a non-invasive means to study susceptibility changes of substances such as myelin and iron in the brain. Particularly, phase images show an increased sensitivity to... Show moreT2*-weighted imaging provides a non-invasive means to study susceptibility changes of substances such as myelin and iron in the brain. Particularly, phase images show an increased sensitivity to magnetic susceptibility differences with increased field strengths. The primary goal of the thesis was to develop methods for quantitative analysis of human brain T2*-weighted images at ultrahigh field strength. Additionally, it was also aimed to investigate the use of textural features derived from whole-brain deformation field for classification of Alzheimer__s disease (AD). A framework for the detection of between-group textural differences in 7T T2*-weighted magnitude and phase images of subcortical structures was presented, and its application was demonstrated in Huntington__s disease. A novel algorithm for segmentation of the cerebral cortex from 7T T2*-weighted images was proposed and extensively validated. Subsequently, a highly automated method was proposed for quantification of regional changes in these images in terms of gray matter/white matter contrast and cortical profile. In addition to an analysis of aging effect using data of young and elderly healthy subjects, this method was also applied to compare early- and late- onset AD patients. The analysis techniques presented in this thesis can be useful tools for susceptibility studies using ultrahigh field MR images Show less
The main goal of this thesis was to develop methods for automated segmentation, registration and classification of the carotid artery vessel wall and plaque components using multi-sequence MR... Show moreThe main goal of this thesis was to develop methods for automated segmentation, registration and classification of the carotid artery vessel wall and plaque components using multi-sequence MR vessel wall images to assess atherosclerosis. First, a general introduction into atherosclerosis and different stages of the disease were described including the importance to differentiate between stable and vulnerable plaques. Several non-invasive imaging techniques were discussed and the advantages of multi-sequence MRI were highlighted. Different novel automated image segmentation and registration techniques for analysis of the MRI images have been developed. A 3D vessel model to automatically segment the vessel wall was presented. Automated image registration was applied to correct for patient movement during the acquisition of an MRI scan and between MRI scans. The last topic is the automatic classification of the different plaque components which can be present inside the vessel wall. All techniques were developed and validated using relevant patient data and reference standards. The work presented is an important contribution to the automated analysis of multi-sequence MR vessel wall imaging of the carotid artery. These techniques can speed up the current manual analysis and are potentially more accurate and more reproducible. Show less
Modern radiotherapy requires accurate region of interest (ROI) inputs for plan optimization and delivery. Target delineation, however, remains operator-dependent and potentially serves as a major... Show moreModern radiotherapy requires accurate region of interest (ROI) inputs for plan optimization and delivery. Target delineation, however, remains operator-dependent and potentially serves as a major source of treatment delivery error. In order to optimize this critical, yet observer-driven process, a flexible web-based platform for individual and cooperative target delineation analysis and instruction was developed in order to meet the following unmet needs: (1) an open-source/open-access platform for automated/semiautomated quantitative interobserver and intraobserver ROI analysis and comparison, (2) a real-time interface for radiation oncology trainee online self-education in ROI definition, and (3) a source for pilot data to develop and validate quality metrics for institutional and cooperative group quality assurance efforts. The resultant software, Target Contour Testing/Instructional Computer Software (TaCTICS), developed using Ruby on Rails, has since been implemented and proven flexible, feasible, and useful in several distinct analytical and research applications. Show less
High Throughput (HT) methods are high volume experimental approaches that are common in the fields of the life-sciences. The instrumentation for these methods differs per application. We will focus... Show moreHigh Throughput (HT) methods are high volume experimental approaches that are common in the fields of the life-sciences. The instrumentation for these methods differs per application. We will focus on the HT methods that are concerned with imaging. The aim of this thesis is to find robust methods for object extraction and analysis. We focus on the Computer Science aspects of such analysis, namely pattern recognition. Pattern Recognition can be seen in the context of object recognition and data mining. Both aspects will be described in this thesis. We present a framework for segmenting and recognizing the objects of interest based on Template Matching. This approach was designed for an application in the HT screening of zebrafish embryos. All proposed methods are fully automated. We further elaborate on the segmentation algorithms to apply these in software that can be used in a HT context to derive measurements. Then we apply the software on a real life problem involving zebrafish infected with Mycobacterium marinum. Show less