Computed tomography (CT) is routinely used for diagnosing lung disease and developing treatment plans using images of intricate lung structure with submillimeter resolution. Automated segmentation of anatomical structures in such images is important to enable efficient processing in clinical and research settings. Convolution neural networks (ConvNets) are largely successful at performing image segmentation with the ability to learn discriminative abstract features that yield generalizable predictions. However, constraints in hardware memory do not allow deep networks to be trained with high-resolution volumetric CT images. Restricted by memory constraints, current applications of ConvNets on volumetric medical images use a subset of the full image; limiting the capacity of the network to learn informative global patterns. Local patterns, such as edges, are necessary for precise boundary localization, however, they suffer from low specificity. Global information can disambiguate structures that are locally similar. The central thesis of this doctoral work is that both local and global information is important for segmentation of anatomical structures in medical images. A novel multi-scale ConvNet is proposed that divides the learning task across multiple networks; each network learns features over different ranges of scales. It is hypothesized that multi-scale ConvNets will lead to improved segmentation performance, as no compromise needs to be made between image resolution, image extent, and network depth. Three multi-scale models were designed to specifically target segmentation of three pulmonary structures: lungs, fissures, and lobes. The proposed models were evaluated on a diverse datasets and compared to architectures that do not use both local and global features. The lung model was evaluated on humans and three animal species; the results demonstrated the multi-scale model outperformed single scale models at different resolutions. The fissure model showed superior performance compared to both a traditional Hessian filter and a standard U-Net architecture that is limited in global extent. The results demonstrated that multi-scale ConvNets improved pulmonary CT segmentation by incorporating both local and global features using multiple ConvNets within a constrained-memory system. Overall, the proposed pipeline achieved high accuracy and was robust to variations resulting from different imaging protocols, reconstruction kernels, scanners, lung volumes, and pathological alterations; demonstrating its potential for enabling high-throughput image analysis in clinical and research settings.

Despite affecting many organ systems, the leading cause of morbidity and mortality in the cystic fibrosis (CF) population is lung disease. For the current studies we investigated elements of CF lung disease in a porcine model of CF and in people with CF. Our primary analysis tool was chest computed tomography (CT). To investigate early CF lung disease we examined three week old CF and non-CF pigs. We found three week old CF pigs to have large, irregular tracheal smooth muscle bundles, airways of reduced size, airways of irregular shape, and airways of abnormal distensibility. Three week old CF pig lung parenchyma was more heterogenous in density than three week non-CF pigs, especially in the right cephalad lung. The degree of lung tissue heterogeneity in CF pigs correlated with the degree of lung infection. Three week old CF pigs also had significantly more air trapping upon exhalation, evidence of airflow obstruction, than non-CF pigs. The degree of air trapping correlated with the degree of mucus accumulation in the airways. These data show that CF pigs spontaneously develop hallmark features of CF lung disease within weeks of birth, and that abnormal airway growth and development in CF may contribute to lung disease. This study helped set the foundation for future comparative studies involving CF therapeutics, for example, antibiotics and mucolytics. In adults with CF we performed a before drug, after drug study. The drug was ivacaftor, and it restores the basic underlying defect in a subset of people with CF: impaired function of a particular anion channel. We hypothesized that abnormal airway smooth muscle behavior in people with CF, known as “CF asthma,” is, in part, a primary pathogenic mechanism of CF lung disease. We tested our hypothesis by assaying smooth muscle tone before and after administration of ivacaftor. We limited the time duration to two days. We reasoned two days was long enough for ivacaftor to become effective, but not long enough to reverse long standing lung infection and inflammation which could affect smooth muscle function independently. The implication being, that observed changes would be directly due to restoration of the CF defect. We found evidence suggesting relaxation of airway and vascular smooth muscle tone. And, the change in airway smooth muscle tone correlated with the change in vascular smooth muscle tone. These data suggest that impaired smooth muscle function is a primary element of CF lung disease. Many of the people in our two day ivacaftor study returned for follow up after one year of ivacaftor therapy. We hypothesized that radiographic features of lung disease would improve following one year of ivacaftor therapy. We observed no change in lung volume upon inspiration, but a reduction in expiratory lung volume, approximately half of which occurred within two days. Our airway measurements were confounded by errors in scan reconstruction, however, other published studies report airway wall thinning over long term ivacaftor administration. Taken together, these studies of pigs with CF and people with CF, help us understand this disease.

The use of computed tomography (CT) in the care of patients has grown dramatically since its introduction over 30 years ago. The vast majority of the utilization research has focused on factors associated with the variable use in the outpatient and emergency department settings. This has left much of the inpatient use and variation understudied. This study has multiple aims. The first is to characterize the inpatient variation across multiple states and markets. The second is to evaluate the relationship between inpatient CT use and commercial payers across these areas. The third is to develop a model to evaluate the relationship between inpatient CT use and the characteristics of markets, hospitals, and patients. The study uses a four-state convenience sample of cross-sectional data for hospitals. It included non-Federal, acute care hospitals that reported the performance of inpatient CT exams during 2015 (N=181). The literature review was used to justify the inclusion of variables in the study. The descriptive analyses were used to justify the appropriateness of the variables and methodology for testing. A comparison of means demonstrated the significant differences for inpatient utilization between states. A univariate general linear model demonstrated a negative relationship with a hospital’s proportion of commercially insured patients and the inpatient utilization rate. An ordinary least squares multivariate linear regression was used to test for variable significance within each of three constructs: markets, hospitals, and patients. The results indicated that inpatient CT rates were positively associated with higher level of insurer concentration (market), positively associated with system centralization (hospitals), and negatively associated with a hospital’s increasing proportion of minority patient discharges (patients). The study serves an important function in identifying varying patterns of CT utilization across the full spectrum of inpatients across multiple states, regardless of payer. It also creates new knowledge about how the characteristics of these markets, hospitals, and patients are related to inpatient use. It also provides implications for administrators, researchers, and policy makers. The additional knowledge and understanding provided by this research have the potential to lead to improvements in the appropriate and equitable use of inpatient CT exams.

<p>The county of Dalarna is at present carrying through a process of digitalization where traditional x-ray film is being replaced with digital detectors. Earlier used methods for dose optimization turned out not being sufficient. This report presents a method to harmonize dose levels between x- ray sites equipped with Fujifilm imaging plate systems. An exposure index, S, related to the dose level of the examination is computed to every x-ray image. S turned out to be inversely proportional to the detector dose used at the examination. Detector dose is also the one simulated factor that doubtlessly affects S the most.There are however a lot of parameters, e.g difference between patients, that are not easily simulated even though they still have considerable affect. The method for harmonizing dose levels between x-ray sites are based on statistics of collected S-values for a kind of examination. The average of the collected S-values levels variations and is a more solid measure of the dose level for the examination. By means of this method the dose level of frontal images of the lungs at a site have been reduced by 30 per cent without endanger the diagnostic security.</p>

<p>The county hospital of Dalarna has for the last couple of years carried through a process of digitalization. The result is that within the county it exists image plate systems from two different manufacturers. In an attempt to create a tool for dose optimization and dose surveillance the county of Dalarna would like to investigate if the exposure index from Agfa and Fuji is suitable as a dose indicator. An investigation of the exposure index, S, from Fuji has already been done. This thesis has been continuing the investigation by evaluating the stability of the exposure index, lgM, from Agfa. Simultaneously an observation if there is a simple relation between the exposure indicators from Agfa and Fuji has been performed. The result showed that the average of lgM, for a set of images from the same type of examination, is appropriate as a dose indicator to the image plate for that specific examination type and X-ray equipment. The usefulness is linked to the same tube voltage and Speed Class for a specific examination, and is to a certain degree restricted by the collimating. There is a relation between the exposure index from Agfa and Fuji, and there is in a simple way possible to transform S-values to lgM-values for comparison. The relation turned out to be examination specific.</p>

<p>The medical imaging system Sectra PACS from Sectra Imtec contains a 3D mode that can be used for visualising image stacks from e.g. computed tomography. Various structures of human anatomy can be visualised in the 3D mode, but visualisations of the bronchial tree of the lungs rarely become good enough to be useful. The goal of this work was to investigate ways of improving such visualisations.</p><p>Various approaches were studied, evaluated and tested. The fact that most effort was needed for small structures with sizes similar to the resolution of the images made things slightly more complicated. A method classifying neighbourhoods based on local structure emerged as most promising, and was used as foundation for a proposed algorithm. It creates a mask representing the presence of bronchi, allowing the hiding of uninteresting structures in its proximity. The algorithm was then implemented so that it could be tested together with the existing system.</p><p>The method was found to work well and was able to detect the smaller tubes of the bronchial tree and output the desired classification mask. Its usefulness was somewhat reduced by issues relating to speed, and the fact that many computed tomography image stacks lack the necessary resolution for visualising the finer details of the bronchial tree.</p>

Objective : To determine the relationship of tongue volume as determined from Cone Beam Computed Tomography (CBCT) scan reconstructions with maxillary and mandibular arch width, axial cross-sectional palatal perimeter, palatal index and axial inclination of upper and lower first premolars and molars. Method: Thirty subjects without prior orthodontic treatment swished barium sulfate to coat the tongue prior to CBCT imaging. The scan reconstructions were analyzed with three after-market softwares and intra-examiner reliability was assessed. Results: Absolute agreement intra-class correlation coefficients were used to determine reliability of the measurements. Pearson correlation coefficients and regression analysis were used to determine relationships. Conclusions: Tongue volume was strongly correlated with upper inter-molar width and palatal perimeter at the molar level, and least correlated with lower inter-molar width and axial inclination of the upper and lower first premolars and molars. The differences in measurements obtained from the three softwares were not statistically significant.

This thesis aims at introducing a methodology for clinical evaluation of orthodontic treatments using three-dimensional dento-maxillofacial images. Since complementary information is achieved by integrating multiple modalities, cone-beam computed tomography (CBCT) and stereophotogrammetry technologies are used to develop a methodology for tracking bone and facial skin variations over time. Our proposed methodology consists of a two-phase registration procedure. In the first phase, the multimodal images are registered using an extrinsic landmark-based registration followed by a robust Iterative Closest Points (ICP) method. In the second phase, by utilizing specific anatomical landmarks, single modal images of the skull and the mandible are registered over time using an intrinsic landmark-based registration method followed by the robust ICP algorithm. The results of registrations show that the signed error distribution of both mandible and skull registrations follow a normal distribution while all the errors fall within the CBCT precision range.

Purpose The purpose of this research is an assessment of whether the noninvasive method of computed tomography angiography of liver preoperatively before liver transplantation does provide sufficient details for the surgeons. At the same time, by comparing the images reconstructed from scanners equipped with and without 64 slices, we can appreciate the limits of each scanner and set up an acceptable standard for each scanner. Methods In a retrospective study from May, 2005 to March, 2007, 52 patients receiving liver transplantation underwent 104 computed tomography studies for evaluation of intra-hepatic arterial anatomy provided by the radiology department in a southern medical center. The reconstructed CT angiograms were compared to see if they were compatible to the conventional digital subtracted angiograms for the demonstration of the anatomy details. The CT images exported from different mutislice count scanner¡]MSCT¡^were reevaluated for the limits and variance of the quality. Results Multi-slices CT with its obvious advantage of the faster speed and thinner scan slices, can distinguish 100¢M(104/104) of the 3rd grade of arteries of liver (segmental branches ), 61.5¢M(64/104) for 4th grade of arteries (sub-segmental branches ), 100¢M (104/104) of the portal vein and 93¢M(97/104) of hepatic vein. The images of the 52 patients provide sufficient quality of intra-hepatic arterial details for liver transplantation.

Larger and stronger muscles are positively associated with bone strength in the growing skeleton; however, less is known about the role of muscle properties on bone strength later in life. The primary objective of this study was to examine the relationship between muscle cross sectional area (MCSA), muscle force and rate of torque development (RTD) with bone strength indices (bone strength index (BSI) and strength strain index (SSI)) in the radius of healthy middle-aged adults. All bone and muscle measurements were determined in the non-dominant forearm in a sample of 40 healthy adults (23 men, 17 women: mean age 49.5, SD 2.3 yrs). Peripheral quantitative computer tomography (pQCT) was used to scan the distal and shaft sites of the radius bone in the forearm. MCSA was determined from the forearm shaft scan. Forearm muscle force was measured by hand grip dynamometry and RTD was obtained from isometric wrist flexion from an isokinetic dynamometry protocol. Hierarchical regression analyses were used to identify whether muscle properties (MCSA, grip force, and RTD) independently predicted radius bone strength indices (BSI and SSI), after adjusting for the confounders of sex, height and weight. Steps of the regression models that included sex, height, weight and a muscle property explained between 66% and 71% of variance in distal radius BSI and between 74% and 78% variance of estimated bone strength (SSI) at the shaft site (all steps p<0.001). MCSA explained a significant amount of variance in BSI (R2=0.08; p<0.01) and SSI (R2=0.04; p<0.05) at the radius. Grip force was also a significant predictor of SSI (R2=0.05; p<0.01) but not distal radius BSI (R2=0.03; p=0.07). Conversely, RTD explained a significant amount of variance in bone strength at the distal radius (R2=0.04; p<0.05), but not at the shaft (R2=0.01; p=0.17). These cross sectional findings support the theory that regional muscle size, force, and rate of torque development are related to estimated bone strength in the forearm at midlife. Further research should focus on targeted interventions to help determine which muscle property elicits a greater osteogenic response to optimize bone strength at distal and shaft sites of the radius.