Haralick texture features are common texture descriptors in image analysis. To compute the Haralick features, the image gray-levels are reduced, a process called quantization. The resulting features depend heavily on the quantization step, so Haralick features are not reproducible unless the same quantization is performed. The aim of this work was to develop Haralick features that are invariant to the number of quantization gray-levels. By redefining the gray-level co-occurrence matrix (GLCM) as a discretized probability density function, it becomes asymptotically invariant to the quantization. The invariant and original features were compared using logistic regression classification to separate two classes based on the texture features. Classifiers trained on the invariant features showed higher accuracies, and had similar performance when training and test images had very different quantizations. In conclusion, using the invariant Haralick features, an image pattern will give the same texture feature values independent of image quantization.

PurposeThe role of hybrid imaging using F-18-fluoro-2-deoxy-D-glucose positron-emission tomography (FDG-PET), computed tomography (CT) and magnetic resonance imaging (MRI) to improve preoperative evaluation of rectal cancer is largely unknown. To investigate this, the RECTOPET (REctal Cancer Trial on PET/MRI/CT) study has been launched with the aim to assess staging and restaging of primary rectal cancer. This report presents the study workflow and the initial experiences of the impact of PET/CT on staging and management of the first patients included in the RECTOPET study.MethodsThis prospective cohort study, initiated in September 2016, is actively recruiting patients from Region Vasterbotten in Sweden. This pilot study includes patients recruited and followed up until December 2017. All patients had a biopsy-verified rectal adenocarcinoma and underwent a minimum of one preoperative FDG-PET/CT and FDG-PET/MRI examination. These patients were referred to the colorectal cancer multidisciplinary team meeting at Umea University Hospital. All available data were evaluated when making management recommendations. The clinical course was noted and changes consequent to PET imaging were described; surgical specimens underwent dedicated MRI for anatomical matching between imaging and histopathology.ResultsTwenty-four patients have so far been included in the study. Four patients were deemed unresectable, while 19 patients underwent or were scheduled for surgery; one patient was enrolled in a watch-and-wait programme after restaging. Consequent to taking part in the study, two patients were upstaged to M1 disease: one patient was diagnosed with a solitary hepatic metastasis detected using PET/CT and underwent metastasectomy prior to rectal cancer surgery, while one patient with a small, but metabolically active, lung nodulus experienced no change of management. PET/MRI did not contribute to any recorded change in patient management.ConclusionsThe RECTOPET study investigating the role of PET/CT and PET/MRI for preoperative staging of primary rectal cancer patients will provide novel data that clarify the value of adding hybrid to conventional imaging, and the role of PET/CT versus PET/MRI.Trial registrationNCT03846882.

Aim: To evaluate if quantifying proton density fat fraction (PDFF) would be useful in separating lipoma, atypical lipomatous tumour (ALT) and liposarcoma in the extremities and trunk. In addition, differentiating ALT versus non-classical lipomas using magnetic resonance imaging (MRI)-based fatty acidcomposition (FAC) and three-dimensional (3D) texture analysis was tested.

Material and methods: This prospective study (undertaken between 2014–2017; comprising 20 women, 21 men) was approved by the Regional Ethical Review Board and informed consent was obtained from all participants. For PDFF and FAC 3D spoiled gradient multi-echo images were acquired. PDFF was analysed in 16 lipomas (25–76 years), 14 ALTs (42–78 years) and 11 myxoid liposarcomas (31–68 years). The difference of mean PDFF was tested with one-way analysis of variance. A support vector machine algorithm was used to find the separating mean PDFF values.

Results: Mean PDFF for lipomas was 90% (range 76–98%), for ALT 83% (range 62–91%), and for liposarcoma 4% (range 0–21%). The difference of mean PDFF for liposarcomas versus ALT and lipoma was significant (p=0.0001, for both), and for ALT versus lipoma (p=0.021). The optimal threshold for separating liposarcoma from ALT and lipoma was 41.5%, and for ALT and lipoma 85%. Texture analysis could not separate ALT and non-classical lipomas, while the difference for FAC unsaturation degree was significant (p=0.013).

Conclusion: Measuring PDFF is a promising complement to standard MRI, to separate liposarcomas from ALT and lipomas. Lipomas that are not solely composed of fat cannot confidently be separated from ALT using PDFF, FAC, or texture analysis.

The Haralick texture features are common in the image analysis literature, partly because of their simplicity and because their values can be interpreted. It was recently observed that the Haralick texture features are very sensitive to the size of the GLCM that was used to compute them, which led to a new formulation that is invariant to the GLCM size. However, these new features still depend on the sample size used to compute the GLCM, i.e. the size of the input image region-of-interest (ROI).

The purpose of this work was to investigate the performance of density estimation methods for approximating the GLCM and subsequently the corresponding invariant features.

Three density estimation methods were evaluated, namely a piece-wise constant distribution, the Parzen-windows method, and the Gaussian mixture model. The methods were evaluated on 29 different image textures and 20 invariant Haralick texture features as well as a wide range of different ROI sizes.

The results indicate that there are two types of features: those that have a clear minimum error for a particular GLCM size for each ROI size, and those whose error decreases monotonically with increased GLCM size. For the first type of features, the Gaussian mixture model gave the smallest errors, and in particular for small ROI sizes (less than about 20×20).

In conclusion, the Gaussian mixture model is the preferred method for the first type of features (in particular for small ROIs). For the second type of features, simply using a large GLCM size is preferred.

Purpose: Simultaneous collection of PET and MR data for radiotherapy purposes are useful for, for example, target definition and dose escalations. However, a prerequisite for using PET/MR in the radiotherapy workflow is the ability to image the patient in treatment position. The aim of this work was to adapt a GE SIGNA PET/MR scanner to image patients for radiotherapy treatment planning and evaluate the impact on signal-to-noise (SNR) of the MR images, and the accuracy of the PET attenuation correction. Method: A flat tabletop and a coil holder were developed to image patients in the treatment position, avoid patient contour deformation, and facilitate attenuation correction of flex coils. Attenuation corrections for the developed hardware and an anterior array flex coil were also measured and implemented to the PET/MR system to minimize PET quantitation errors. The reduction of SNR in the MR images due to the added distance between the coils and the patient was evaluated using a large homogenous saline-doped water phantom, and the activity quantitation errors in PET imaging were evaluated with and without the developed attenuation corrections. Result: We showed that the activity quantitation errors in PET imaging were within ±5% when correcting for attenuation of the flat tabletop, coil holder, and flex coil. The SNR of the MRI images were reduced to 74% using the tabletop, and 66% using the tabletop and coil holders. Conclusion: We present a tabletop and coil holder for an anterior array coil to be used with a GE SIGNA PET/MR scanner, for scanning patients in the radiotherapy work flow. Implementing attenuation correction of the added hardware from the radiotherapy setup leads to acceptable PET image quantitation. The drop in SNR in MR images may require adjustment of the imaging protocols.

Magnetic resonance imaging, MRI, offers a vast range of imaging methods that can be employed in the characterization of tumors. MRI is generally used in a qualitative way, where radiologists interpret the images for e.g. diagnosis, follow ups, or assessment of treatment response. In the past decade, there has been an increasing interest for quantitative imaging, which give repeatable measurements of the anatomy. Quantitative imaging allows for objective analysis of the images, which are grounded in physical properties of the underlying tissues. The aim of this thesis was to improve quantitative measurements of Dynamic contrast enhanced MRI (DCE-MRI), and the texture analysis of diffusion weighted MRI (DW-MRI).

DCE-MRI measures perfusion, which is the delivery of blood, oxygen and nutrients to the tissues. The exam involves continuously imaging the region of interest, e.g. a tumor, while injecting a contrast agent (CA) in the blood stream. By analyzing how fast and how much CA leaks out into the tissues, the cell density and the permeability of the capillaries can be estimated. Tumors often have an irregular and broken vasculature, and DCE-MRI can aid in tumor grading or treatment assessment. One step is crucial when performing DCE-MRI analysis, the quantification of CA in the tissue. The CA concentration is difficult to measure accurately due to uncertainties in the imaging, properties of the CA, and physiology of the patient. Paper I, the possibility of using two aspects of the MRI data, phase and magnitude, for improved CA quantification, is explored. We found that the combination of phase and magnitude information improved the CA quantification in regions with high CA concentration, and was more advantageous for high field strength scanners.

DW-MRI measures the diffusion of water in and between cells, which reflects the cell density and structure of the tissue. The structure of a tumor can give insights into the prognosis of the disease. Tumors are heterogeneous, both genetically and in the distribution of cells, and tumors with high intratumoral heterogeneity have poorer prognosis. This heterogeneity can be measured using texture analysis. In 1973, Haralick et al. presented a texture analysis method using a gray level co-occurrence matrix, GLCM, to gauge the spatial distribution of gray levels in the image. This method of assessing texture in images has been successfully applied in many areas of research, from satellite images to medical applications. Texture analysis in treatment outcome assessment is studied in Paper II, where we showed that texture can distinguish between groups of patients with different survival times, in images acquired prior to treatment start.

However, this type of texture analysis is not inherently quantitative in the way it is calculated today. This was studied in Paper III, where we investigated how texture features were affected by five parameters related to image acquisition and pre-processing. We found that the texture feature values were dependent on the choice of these imaging and preprocessing parameters. In Paper IV, a novel method for calculating Haralick texture features was presented, which makes the texture features asymptotically invariant to the size of the GLCM. This method allows for comparison of textures between images that have been analyzed in different ways.

In conclusion, the work in this thesis has been aimed at improving quantitative analysis of tumors using MRI and texture analysis.

In recent years, texture analysis of medical images has become increasingly popular in studies investigating diagnosis, classification and treatment response assessment of cancerous disease. Despite numerous applications in oncology and medical imaging in general, there is no consensus regarding texture analysis workflow, or reporting of parameter settings crucial for replication of results. The aim of this study was to assess how sensitive Haralick texture features of apparent diffusion coefficient (ADC) MR images are to changes in five parameters related to image acquisition and pre-processing: noise, resolution, how the ADC map is constructed, the choice of quantization method, and the number of gray levels in the quantized image. We found that noise, resolution, choice of quantization method and the number of gray levels in the quantized images had a significant influence on most texture features, and that the effect size varied between different features. Different methods for constructing the ADC maps did not have an impact on any texture feature. Based on our results, we recommend using images with similar resolutions and noise levels, using one quantization method, and the same number of gray levels in all quantized images, to make meaningful comparisons of texture feature results between different subjects.

Objective: Multiparametric magnetic resonance imaging (MRI) and PI-RADS (Prostate Imaging - Reporting and Data System) has become the standard to determine a probability score for a lesion being a clinically significant prostate cancer. T2-weighted and diffusion-weighted imaging (DWI) are essential in PI-RADS, depending partly on visual assessment of signal intensity, while dynamic-contrast enhanced imaging is less important. To decrease inter-rater variability and further standardize image evaluation, complementary objective measures are in need. Methods: We here demonstrate a sequence enabling simultaneous quantification of apparent diffusion coefficient (ADC) and T2-relaxation, as well as calculation of the perfusion fraction f from low b-value intravoxel incoherent motion data. Expandable wait pulses were added to a FOCUS DW SE-EPI sequence, allowing the effective echo time to change at run time. To calculate both ADC and f, b-values 200 s/mm(2) and 600 s/mm(2) were chosen, and for T2-estimation 6 echo times between 64.9 ms and 114.9 ms were used. Results: Three patients with prostate cancer were examined and all had significantly decreased ADC and T2 values, while f was significantly increased in 2 of 3 tumors. T2 maps obtained in phantom measurements and in a healthy volunteer were compared to T2 maps from a SE sequence with consecutive scans, showing good agreement. In addition, a motion correction procedure was implemented to reduce the effects of prostate motion, which improved T2-estimation. Conclusions: This sequence could potentially enable more objective tumor grading, and decrease the inter-rater variability in the PI-RADS classification.