This thesis presents techniques constructed to aid the radiologists in detecting breast cancer, the second largest cause of cancer deaths for western women. In the first part of the thesis, a computer-aided detection (CAD) system constructed for the detection of stellate lesions is presented. Different segmentation methods and an attempt to incorporate contra-lateral information are evaluated. In the second part, a new method for evaluating such CAD systems is presented based on constructing credible regions for the number of false positive marks per image at a certain desired target sensitivity. This method shows that the resulting regions are rather wide and this explains some of the difficulties encountered by other researchers when trying to compare CAD algorithms on different data sets. In this part an attempt to model the clinical use of CAD as a second look is also made and it shows that applying CAD in sequence to the radiologist in a routine manner, without duly altering the decision criterion of the radiologist, might very well result in suboptimal operating points. Finally, in the third part two dual-energy imaging methods optimized for contrast-enhanced imaging of breast tumors are presented. The first is based on applying an electronic threshold to a photon-counting digital detector to discriminate between high- and low-energy photons. This allows simultaneous acquisition of the high- and low-energy images. The second method is based on the geometry of a scanned multi-slit system and also allows single-shot contrast-enhanced dual-energy mammography by filtering the x-ray beam that reaches different detector lines differently. / QC 20100819

Purpose: The aim of this thesis is to study the European (ESTRO/EAU/EORTC) and American (ABS) guidelines how to report the permanent seed implant and the most significant dosimetric parameters. It will also report on the permanent seed implant at Södersjukhuset, Karolinska University Hospital according to the guidelines. A large number of studies on pre- and post-implant dosimetry on permanent seed implants have recently been published but none is considered a standard. This makes it difficult, if not impossible, to compare data from different centres. The differences in reporting will also be discussed in this thesis. Another part of the study is to investigate how the morbidity correlates with the dose. The results in this report will give an overview of the experience at Södersjukhuset. Matherials and Methods: This study includes 198 patients who received implants between 2004-2007 with I-125 seeds under transperineal ultrasound at Södersjukhuset (to a prescribed dose of 145 Gy). The dose-planning system VariSeed 7.1 was used with an online connection to the ultrasound system with real-time verification. Dose constraints for the planning system are V(100)>99%, V(150)>60%, V(200)>25%, UrD(10)<130% and UrD(30)<125%. Outer and inner wall of rectum was outlined for 55 patients as recommended by ESTRO/EAU/EORTC and doses to rectum were also computed. Results: The median value for dosimetric parameters at Södersjukhuset, Karolinska University Hospital are for the prostate; D(90)=174Gy (153-194Gy), V(100)= 99% (93-100%), V(150)= 57% (40-74%), for the urethra; UrD(30) = 130% (112-147%), UrD(10) = 124% (107-142%) and for the rectum; RD2cc= 98Gy (73-128Gy), RD0.1cc=164Gy (119-240Gy), RV(100)=0.3cc (0.0-1.3cc), RV(150)=0.0cc (0.0-0.2cc). These values correspond to recommended data, except for the V(150) value. Regarding the clinically observed results, 3 patients had a relapse in their cancer, 2 patients had mild proctitis and 15 patients had urinary problems. Discussion and Conclusions: The significant dosimetric parameters for reporting according to ESTRO/EAU/EORTC and ABS for prostate are D90[Gy], V(100)[%] and V(150)[%], for urethra are D(30) and D(10), and for rectum RD2cc and RD0.1cc. These parameters consider as a minimum to use and they further recommend secondary parameters to report. Other authors have also recommended to report RV(100) and RV(150) for rectum. This study did not show any relationship between UrD(10), UrD(30) and urinary morbidity. According to the recommendations every patient should undergo a CT-based evaluation. Further investigations are needed on whether a post-implant CT-study is necessary for real-time implantation, as there is not enough published data on this aspect.

One of the most used irradiation techniques in modern radiation therapy is step-and-shoot IMRT. The accuracy of this technique when delivering complex dose distributions strongly depends on the size of the subfields. The aims of this study is to determine the minimum size of subfields that can be used efficiently in Step-and-Shoot IMRT, to investigate the validation process for beam delivery and treatment planning dose calculations, and to find recommendations for practical clinical implementations. Two different detectors, a CC04 ion chamber and a SFD stereotactic diode, have been used for measuring head scatter factors in air (Sc), total output factors (Scp) and dose profiles in water for a wide range of field sizes. The measurements were compared to calculations done with a pre-release version of the Nucletron MasterPlanTM v 3.1 treatment planning system that employs a novel, high resolution fluence modelling for both its pencil beam and collapsed cone dose calculation algorithms. Collimator settings were explicitly checked using FWHM film measurements with a build-up sheet of tungsten placed close to the treatment head to reduce the influence from lateral electron transport and geometrical penumbra. An analysis of the influence and sensitivity of Scp for small fields with respect to the linear accelerator source size and shape was also made. The measurements with the ionization chamber and the stereotactic diode showed good agreements with each other and with the treatment planning system calculations for field sizes larger than 2×2 cm2. For small field sizes, measurements with different detectors yielded different results. Calculations showed agreements with measurements with the smallest detector, provided careful field size calibration and commissioning of calculation parameters. Uncertainties in collimator settings and source characteristics were shown to yield large uncertainties in Scp for fields smaller than 2×2 cm2. The treatment planning system was found to properly handle small subfields but results were very sensitive to uncertainties in source size, as well as calibration and reproducibility of the collimator settings. Therefore if subfields smaller than 2×2 cm2 are to be used in IMRT extra care should be taken to determine the source characteristics and to calibrate the collimators. The volume of the detectors used for validation of such small fields and the loss of charged particle equilibrium conditions also have to be taken into consideration.

Patient dose in diagnostic radiology is usually expressed in terms of organ dose and effective dose. The latter is used as a measure of the stochastic risk. Determinations of these doses are obtained by measurements (Thermoluminescent dosemeters) or by calculations (Monte Carlo simulation). Conversion factors for the calculation of effective dose from dose-area product (DAP) values are commonly used to determine radiation dose in conventional x-ray imaging to realize radiation risks for different investigations, and for different ages. The exposure can easily be estimated by converting the DAP into an effective dose. The aim of this study is to determine the conversion factor in procedures by computing the ratio between effective dose and DAP for fluoroscopic cardiac procedures in adults and for conventional lung and urography examinations in children. Thermoluminescent dosemeters (TLD) were placed in an anthropomorphic phantom (Alderson Rando phantom) and child phantom (one year old) in order to measure the organ dose and compute the effective dose. A DAP meter was used to measure dose-area product. MC calculations of radiation transport in mathematical anthropomorphic phantoms were used to obtain the effective dose for the same conditions with DAP as input data. The deviation between the measured and calculated data was less than 10 %. The conversion factor for cardiac procedures varies between 0.19 mSvGy-1 cm-2 and 0.18 mSvGy-1 cm-2, for TLD respective MC. For paediatric simulation of a one year old phantom the average conversion factor for urography was 1.34 mSvGy-1 cm-2 and 1,48 mSvGy-1cm-2 for TLD respective MC. This conversion factor will decrease to 1.07 mSvGy-1 cm-2 using the TLD method, if the new ICRP (ICRP Publication 103) weighting factors were used to calculate the effective dose. For lung investigations, the conversion factor for children was 1.75 mSvGy-1 cm-2 using TLD, while this value was 1.62 mSvGy-1 cm-2 using MC simulation. The conversion value increased to 2.02 mSvGy-1 cm-2 using ICRP’s new recommendation for tissue weighting factors and child phantom.

Abstract   Background Semi-automated segmentation using deformable registration of atlases consisting of pre-segmented patient images can facilitate the tedious task of delineating structures and organs in patients subjected to radiotherapy planning. However, a generic atlas based on a single patient may not function well enough due to the anatomical variation between patients. Fusion of segmentation proposals from multiple atlases has the potential to provide a better segmentation due to a more complete representation of the anatomical variation. Purpose The main goal of the present study was to investigate potential operator timesavings from use of atlas-based segmentation compared to manual segmentation of patients with prostate cancer. It was also anticipated that, and evaluated if, the use of semi-automated segmentation workflows would reduce the operator dependent variations in delineation. Materials and Methods A commercial atlas-based segmentation software (VelocityAI from Nucletron AB) was used with several atlases of consistently, protocol based, delineated CT images to create multiple-atlas segmentation proposals through deformable registration. The atlas that was considered most representative was selected to construct single generic atlas segmentation proposals. For fusion of the multiple-atlas segmentations an in-house developed algorithm, which includes information of local registration success was used in a MATLAB-environment[1]. The algorithm used weighted distance map calculations where weights represent probabilities of improving the segmentation results. Based on results from Sjöberg and Ahnesjö the probabilities were estimated using the cross correlation image similarity measure evaluated over a region within a certain distance from the segmentation. 10 patients were included in the study. Each patient was delineated three times, (a) manually by the radiation oncologist, (b) with a generic single-atlas segmentation and (c) with a fusion of multiple-atlas segmentations. For the methods (b) and (c) the radiation oncologist corrected the proposed segmentations blindly without using the result from method (a) as reference. The total number of atlases used for case (c) was 15. The operator time spent by the radiation oncologist was recorded separately for each method. In addition a grading was used to score how helpful the segmentation proposals were for the delineations. The Dice Similarity Coefficient, the Hausdorff distance and the segmented volumes were used to evaluate the similarity between the delineated structures and organs. Results An average time reduction of 26% was found when the radiation oncologist corrected the multiple atlas-based segmentation proposals as compared to manual segmentations. Due to more accurate segmentations and more time saved, segmentation with fused multiple-atlases (c) was superior to the generic single-atlas (b) method, which showed a time reduction of 17%. Hints of an affected intra- and inter-operator variability were seen. Conclusions Atlas-based segmentation saves time for the radiation oncologist but the segmentation proposals always need editing to be approved for dose planning. The atlases, the fusion of these and the software implementation needs to be improved for optimal results and to extend the clinically usefulness.

Background Semi-automated segmentation using deformable registration of atlases consisting of pre-segmented patient images can facilitate the tedious task of delineating structures and organs in patients subjected to radiotherapy planning. However, a generic atlas based on a single patient may not function well enough due to the anatomical variation between patients. Fusion of segmentation proposals from multiple atlases has the potential to provide a better segmentation due to a more complete representation of the anatomical variation. Purpose The main goal of the study was to investigate potential operator timesaving from editing of atlas-based segmentation compared to manual segmentation for head & neck cancer. Materials and Methods A commercial atlas-based segmentation software (VelocityAI from Nucletron AB) was used together with several expert generated and protocol-based atlases of delineated CT images to create multiple atlas segmentations through deformable registration. The atlas that was considered most universal was selected to construct single atlas segmentation proposals. For fusion of the multiple atlas segmentations an in-house developed algorithm, including information of local registration success was used in a MATLAB-environment1. The algorithm uses weighted distance map calculations where weights represent probabilities of improving the segmentation results. Based on previous results1 the probabilities were estimated using the cross correlation image similarity measure evaluated over a region within a certain distance from the segmentation. Ten patients were incorporated in the study. Each patient was delineated three times, (a) manually by the radiation oncologist, (b) with a single atlas segmentation and (c) with a fusion of multiple atlas segmentations. For the methods (b) and (c) the radiation oncologist corrected the proposed segmentations blindly without using the result from method (a) as reference. For case (c) a total number of 11 atlas segmentations were used. The time spent for segmenting or editing the segmentation proposals by the radiation oncologist was recorded separately for each method and each individual ROI. In addition a grading was used to score how helpful the candidate segmentation proposals were for the structure delineations. The Dice Similarity Coefficient, the Hausdorff distance and the volume were used to evaluate the similarity between the delineated structures. Results The results show a time reduction in the order of 40% when the radiation oncologist only has to correct the multiple atlas-based segmentation proposal compared to manual segmentation. When using single atlas the corresponding figure is 21%. Conclusions Using atlas-based segmentation can reduce the time needed for delineation in the head and neck area of patients admitted for radiotherapy. 1C. Sjöberg and A. Ahnesjö, Evaluation of atlas-based segmentation using probabilistic weighted distance maps, Manuscript, Uppsala University, 2011 / Bakgrund Atlasbaserad, semiautomatisk segmentering skulle kunna användas för att underlätta den för onkologen tidskrävande uppgiften med att manuellt segmentera strukturer och organ i patienter vid behandlingsplanering inför strålbehandling. Tidigare segmenterade atlaspatienter ger med hjälp av deformeringsalgoritmer segmenteringsförslag för strukturer i den aktuella patienten. Dessa kan sedan kontrolleras och editeras av onkologen med en tidsbesparing gentemot manuell segmentering som följd. En atlas som baserats på en enstaka individ (singelatlas) kan dock ha begränsningar när det gäller att täcka de anatomiska variationer som finns mellan olika patienter. Därför har metoder med fusionering av multipla segmenteringsförslag från en databas bestående av ett antal sedan tidigare segmenterade patienter (fusionerad multipelatlas) potential att ge ett bättre segmenteringsresultat. Syfte Huvudsyftet med arbetet var att undersöka de möjliga tidsbesparingar för onkologen som kan åstadkommas när editering av atlasbaserad segmentering används vid planering inför strålbehandling i huvud- och halsområdet istället för manuell segmentering Material och metoder En kommersiell, atlasbaserad segmenteringsprogramvara (VelocityAI från Nucletron AB) användes i studien. Genom att låta en erfaren onkolog segmentera ett antal CT-studier (11 st) enligt ett vedertaget protokoll skapades en databas av atlaser som sedan, via deformerbara registreringar, kunde generera lika många segmenteringsförslag för en nytillkommen patient. Den enskilda atlas som ansågs mest representativ valdes till att framställa segmenteringsförslaget för metoden med singelatlas. Till metoden med fusionerade multipla atlaser användes en lokalt utvecklad MATLAB-algoritm baserad på viktade    distansmappar. Vikterna representerar sannolikheten för förbättrat segmenteringsresultat och baseras på tidigare resultat1 där sannolikheterna bestämts utifrån en beräkning av likheterna mellan bilderna i ett visst område från den specifika segmenteringen. Tio patienter har inkluderats i studien. Varje patient segmenterades tre gånger, (a) manuellt, (b) med singelatlas och (c) med fusionerade multipla atlaser. För metoderna (b) och (c) editerades sedan segmenteringsförslagen av onkologen utan att denne fick använda resultatet från metod (a) som referens. För fallet med fusionerade multipla atlaser, (c), användes databasen med 11 atlaser. Tiden onkologen behövde för segmentering respektive editering av segmenteringsförslaget uppmättes i varje enskilt fall för jämförelse. Onkologen fick även göra en bedömning av hur hjälpsamt segmenteringsförslaget var i samband med editeringen. För utvärdering av resultaten användes Dice’s similaritetskoefficient, Hausdorff’s distansmått samt strukturernas volym. Resultat Resultaten visar på att en tidsbesparing i storleksordningen 40 % är rimlig när onkologen editerar förslag från fusioneringen av multipla atlassegmenteringar i jämförelse med manuell segmentering. Vid användning av singelatlas är motsvarande siffra 21 %. Slutsatser Användandet av atlasbaserad segmentering kan reducera tidsåtgången för segmentering av patienter inför strålbehandling i huvud-halsområdet. 1C. Sjöberg and A. Ahnesjö, Evaluation of atlas-based segmentation using probabilistic weighted distance maps, Manuscript, Uppsala University, 2011

This thesis describes an information extraction system that is designed to process free text Turkish radiology reports in order to extract and convert the available information into a structured information model. The system uses natural language processing techniques together with domain ontology in order to transform the verbal descriptions into a target information model, so that they can be used for computational purposes. The developed domain ontology is effectively used in entity recognition and relation extraction phases of the information extraction task. The ontology provides the flexibility in the design of extraction rules, and the structure of the ontology also determines the information model that describes the structure of the extracted semantic information. In addition, some of the missing terms in the sentences are identified with the help of the ontology. One of the main contributions of this thesis is the usage of ontology in information extraction that increases the expressive power of extraction rules and helps to determine missing items in the sentences. The system is the first information extraction system for Turkish texts. Since Turkish is a morphologically rich language, the system uses a morphological analyzer and the extraction rules are also based on the morphological features. TRIES achieved 93% recall and 98% precision results in the performance evaluations.

In the event that a radiological dispersal device (RDD) is detonated in the U.S. or near U.S. interests overseas, it will be crucial that the actors involved in the event can be identified quickly. If irradiated nuclear fuel is used as the dispersion material for the RDD, it will be beneficial for law enforcement officials to quickly identify where the irradiated nuclear fuel originated. One signature which may lead to the identification of the spent fuel origin is the isotopic composition of the RDD debris. The objective of this research was to benchmark a forward model methodology for predicting isotopic composition of spent nuclear fuel used in an RDD while at the same time optimizing the fidelity of the model to reduce computational time. The code used in this study was Monteburns-2.0. Monteburns is a Monte Carlo based neutronic code utilizing both MCNP and ORIGEN. The size of the burnup step used in Monteburns was tested and found to converge at a value of 3,000 MWd/MTU per step. To ensure a conservative answer, 2,500 MWd/MTU per step was used for the benchmarking process. The model fidelity ranged from the following: 2-dimensional pin cell, multiple radial-region pin cell, modified pin cell, 2D assembly, and 3D assembly. The results showed that while the multi-region pin cell gave the highest level of accuracy, the difference in uncertainty between it and the 2D pin cell (0.07% for 235U) did not warrant the additional computational time required. The computational time for the multiple radial-region pin cell was 7 times that of the 2D pin cell. For this reason, the 2D pin cell was used to benchmark the isotopics with data from other reactors. The reactors from which the methodology was benchmarked were Calvert Cliffs Unit #1, Takahama Unit #3, and Trino Vercelles. Calvert Cliffs is a pressurized water reactor (PWR) using Combustion Engineering 14??14 assemblies. Takahama is a PWR using Mitsubishi Heavy Industries 17??17 assemblies. Trino Vercelles is a PWR using non-standard lattice assemblies. The measured isotopic concentrations from all three of the reactors showed good agreement with the calculated values.

The air kerma area product PKA is an important quantity used by hospital physicists in quality assurance and optimization processes in diagnostic radiology and is recommended by national authorities for setting of diagnostic reference levels. PKA can be measured using a transmission ionization chamber (kerma area product (KAP) meter) mounted on the collimator housing. Its signal QKAP must be calibrated to give values of PKA. The objective of this thesis is to analyze the factors influencing the accuracy of the calibration coefficients k= PKA/QKAP and of reported PKA-values. Due to attenuation and scatter in the KAP-meter and presence of extra-focal radiation, values of PKA depend on the choice of integration area A and the distance of the reference plane from the focal spot yielding values of PKA that may differ by as much as 23% depending on this choice. The two extremes correspond to (1) PKA=PKA,o integrated over the exit surface of the KAP-meter resulting in geometry independent calibration coefficients and (2) PKA=PKA,Anom integrated over the nominal beam area in the patient entrance plane resulting in geometry dependent calibration coefficients. Three calibration methods are analysed. Method 1 aims at determine PKA,Anom, for clinical use at the patient entrance plane. At standard laboratories, the method is used to calibrate with respect to radiation incident on the KAP-meter. Problems with extra-focal and scattered radiation are then avoided resulting in calibration coefficients with low standard uncertainty (±1.5 %, coverage factor 2). Method 2 was designed in this work to approach determination of PKA,o using thermoluminescent detectors to monitor contributions from extra-focal radiation and account for the heel effect. The uncertainty in derived calibration coefficients was ± 3% (coverage factor 2). Method 3 uses a Master KAP-meter calibrated at a standard laboratory for incident radiation to calibrate clinical KAP-meters. It has potential to become the standard method in the future replacing the tedious method 2 for calibrations aiming at determination of PKA,o. Commercially available KAP-meters use conducting layers of indium oxide causing a strong energy dependence of their calibration coefficients. This dependence is investigated using Monte Carlo simulations and measurements. It may introduce substantial uncertainties in reported PKA– values since calibration coefficients as obtained from standard laboratories are often available only at one filtration (2.5 mm Al) as function of tube voltage or HVL. This is not sufficient since higher filtrations are commonly used in practice, including filters of Cu. In extreme cases, calibration coefficients for the same value of HVL but using different tube voltages and filtrations can deviate by as much as 30%. If standardised calibration methods are not used and choice of calibration coefficients not carefully chosen with respect to beam quality, the total uncertainty in reported PKA–values may be as large as 40-45%. Conversion of PKA-values to risk related quantities is briefly discussed. The large energy dependence of the conversion coefficients, ε/PKA, for determination of energy imparted,ε, to the patient reduces to a lower energy dependence of calibration coefficients CQ,ε = ε/QKAP for determination of ε from the KAP-meter signal.

The dosimetric properties of fast-neutron beams with energies ≤80 MeV were explored using Monte Carlo techniques. Taking into account transport of all relevant types of released charged particles (electrons, protons, deuterons, tritons, 3He and α particles) pencil-beam dose distributions were derived and used to calculate absorbed dose distributions. Broad-beam depth doses in phantoms of different materials were calculated and compared and the scaling factors required for converting absorbed dose in one material to absorbed dose in another derived. The scaling factors were in good agreement with available published data and show that water is a good substitute for soft tissue even at neutron energies as high as 80 MeV. The inherent penumbra and the fraction of absorbed dose due to photon interactions were also studied, and found to be consistent with measured values reported in the literature. Treatment planning in fast-neutron therapy is commonly performed using dose calculation algorithms designed for photon beam therapy. When applied to neutron beams, these algorithms have limitations arising from the physical models used. Monte Carlo derived neutron pencil-beam kernels were parameterized and implemented in the photon dose calculation algorithms of the TMS (MDS Nordion) treatment planning system. It was shown that these algorithms yield good results in homogeneous water media. However, the method used to calculate heterogeneity corrections in the photon dose calculation algorithm did not yield correct results for neutron beams in heterogeneous media. To achieve results with adequate accuracy using Monte Carlo simulations, fundamental cross-section data are needed. Neutron cross-sections are still not sufficiently well known. At the The Svedberg Laboratory in Uppsala, Sweden, an experimental facility has been designed to measure neutron-induced charged-particle production cross-sections for (n,xp), (n,xd), (n,xt), (n,x3He) and (n,xα) reactions at neutron energies up to 100 MeV. Depending on neutron energy, these generated particles account for up to 90% of the absorbed dose. In experimental determination of the cross-sections, measured data have to be corrected for the energies lost by the charged particles before leaving the target in which they were generated. To correct for the energy-losses, a computational code (CRAWL) was developed. It uses a stripping method. With the limitation of reduced energy resolution, spectra derived using CRAWL compares well with those derived using other methods. In fast-neutron therapy, the relative biological effectiveness (RBE) varies from 1.5 to 5, depending on neutron energy, dose level and biological end-point. LET and other physical quantities, developed within the field of microdosimetry over the past couple of decades, have been used to describe RBE variations between different fast-neutron beams as well as within a neutron irradiated body. In this work, a Monte Carlo code (SHIELD-HIT) capable of transporting all charged particles contributing to absorbed dose, was used to calculate energy-differential charged particle spectra. Using these spectra, values of the RBE related quantities LD, γD, γ* and R were derived and studied as function of neutron energy, phantom material and position in a phantom. Reasonable agreement with measured data in the literature was found and indicates that the quantities may be used to predict RBE variations in an arbitrary fast-neutron beam.