Table of contents

Robotic radiosurgery using more than one circular collimator can improve treatment plan quality and reduce total monitor units (MU). The rationale for an iris collimator that allows the field size to be varied during treatment delivery is to enable the benefits of multiple-field-size treatments to be realized with no increase in treatment time due to collimator exchange or multiple traversals of the robotic manipulator by allowing each beam to be delivered with any desired field size during a single traversal. This paper describes the Iris™ variable aperture collimator (Accuray Incorporated, Sunnyvale, CA, USA), which incorporates 12 tungsten–copper alloy segments in two banks of six. The banks are rotated by 30° with respect to each other, which limits the radiation leakage between the collimator segments and produces a 12-sided polygonal treatment beam. The beam is approximately circular, with a root-mean-square (rms) deviation in the 50% dose radius of <0.8% (corresponding to <0.25 mm at the 60 mm field size) and an rms variation in the 20–80% penumbra width of about 0.1 mm at the 5 mm field size increasing to about 0.5 mm at 60 mm. The maximum measured collimator leakage dose rate was 0.07%. A commissioning method is described by which the average dose profile can be obtained from four profile measurements at each depth based on the periodicity of the isodose line variations with azimuthal angle. The penumbra of averaged profiles increased with field size and was typically 0.2–0.6 mm larger than that of an equivalent fixed circular collimator. The aperture reproducibility is ⩽0.1 mm at the lower bank, diverging to ⩽0.2 mm at a nominal treatment distance of 800 mm from the beam focus. Output factors (OFs) and tissue-phantom-ratio data are identical to those used for fixed collimators, except the OFs for the two smallest field sizes (5 and 7.5 mm) are considerably lower for the Iris Collimator. If average collimator profiles are used, the assumption of circular symmetry results in dose calculation errors that are <1 mm or <1% for single beams across the full range of field sizes; errors for multiple non-coplanar beam treatment plans are expected to be smaller. Treatment plans were generated for 19 cases using the Iris Collimator (12 field sizes) and also using one and three fixed collimators. The results of the treatment planning study demonstrate that the use of multiple field sizes achieves multiple plan quality improvements, including reduction of total MU, increase of target volume coverage and improvements in conformality and homogeneity compared with using a single field size for a large proportion of the cases studied. The Iris Collimator offers the potential to greatly increase the clinical application of multiple field sizes for robotic radiosurgery.

This study investigated the compatibility of a head and neck immobilization device with magnetic resonance imaging (MRI). The immobilization device is used to position a patient in the same way as when receiving a computed tomography (CT) scan for radiotherapy planning and radiation treatment. The advantage of using immobilization in MR is improved accuracy in CT/MR image registration enabling greater confidence in the delineation of structures. The main practical difficulty in using an immobilization device in MRI is that physical constraints make their use incompatible with head imaging coils. Within this paper we describe a method for MR imaging of the brain which allows the use of head and neck immobilization devices. By a series of image quality tests we obtained the same or better image quality as a multi-channel head coil.

The non-invasive localization of focal heart activity via body surface potential measurements (BSPM) could greatly benefit the understanding and treatment of arrhythmic heart diseases. However, the in vivo validation of source localization algorithms is rather difficult with currently available measurement techniques. In this study, we used a physical torso phantom composed of different conductive compartments and seven dipoles, which were placed in the anatomical position of the human heart in order to assess the performance of the Recursively Applied and Projected Multiple Signal Classification (RAP-MUSIC) algorithm. Electric potentials were measured on the torso surface for single dipoles with and without further uncorrelated or correlated dipole activity. The localization error averaged 11 ± 5 mm over 22 dipoles, which shows the ability of RAP-MUSIC to distinguish an uncorrelated dipole from surrounding sources activity. For the first time, real computational modelling errors could be included within the validation procedure due to the physically modelled heterogeneities. In conclusion, the introduced heterogeneous torso phantom can be used to validate state-of-the-art algorithms under nearly realistic measurement conditions.

Radiotherapy using kilovoltage x-rays in conjunction with contrast agents incorporated into the tumor, gold nanoparticles in particular, could represent a potential alternative to current techniques based on high-energy linear accelerators. In this paper, using the voxelized Zubal phantom in conjunction with the Monte Carlo code PENELOPE to model a prostate cancer treatment, it is shown that in combination with a 360° arc delivery technique, tumoricidal doses of radiation can be delivered to deep-seated tumors while still providing acceptable doses to the skin and other organs at risk for gold concentrations in the tumor within the range of 7–10 mg-Au per gram of tissue. Under these conditions and using a x-ray beam with 90% of the fluence within the range of 80–200 keV, a 72 Gy physical absorbed dose to the prostate can be delivered, while keeping the rectal wall, bladder, skin and femoral heads below 65 Gy, 55 Gy, 40 Gy and 30 Gy, respectively. However, it is also shown that non-uniformities in the contrast agent concentration lead to a severe degradation of the dose distribution and that, therefore, techniques to locally quantify the presence of the contrast agent would be necessary in order to determine the incident x-ray fluence that best reproduces the dosimetry obtained under conditions of uniform contrast agent distribution.

This work reports on the development and performance evaluation of the VrPET/CT, a new multimodality scanner with coplanar geometry for in vivo rodent imaging. The scanner design is based on a partial-ring PET system and a small-animal CT assembled on a rotatory gantry without axial displacement between the geometric centers of both fields of view (FOV). We report on the PET system performance based on the NEMA NU-4 protocol; the performance characteristics of the CT component are not included herein. The accuracy of inter-modality alignment and the imaging capability of the whole system are also evaluated on phantom and animal studies. Tangential spatial resolution of PET images ranged between 1.56 mm at the center of the FOV and 2.46 at a radial offset of 3.5 cm. The radial resolution varies from 1.48 mm to 1.88 mm, and the axial resolution from 2.34 mm to 3.38 mm for the same positions. The energy resolution was 16.5% on average for the entire system. The absolute coincidence sensitivity is 2.2% for a 100–700 keV energy window with a 3.8 ns coincident window. The scatter fraction values for the same settings were 11.45% for a mouse-sized phantom and 23.26% for a rat-sized phantom. The peak noise equivalent count rates were also evaluated for those phantoms obtaining 70.8 kcps at 0.66 MBq/cc and 31.5 kcps at 0.11 MBq/cc, respectively. The accuracy of inter-modality alignment is below half the PET resolution, and the image quality of biological specimens agrees with measured performance parameters. The assessment presented in this study shows that the VrPET/CT system is a good performance small-animal imager, while the cost derived from a partial ring detection system is substantially reduced as compared with a full-ring PET tomograph.

In epicardial potential imaging, the epicardial potential is reconstructed computationally from the measured body surface potential. The transfer function that relates the heart and body surface potentials is commonly constructed with some point-collocation-weighted boundary element technique, assuming an electrically homogeneous volume conductor. This assumption causes modeling errors. In this study, the system of surface integral equations that describes the relationship between the heart and body surface potentials is thoroughly derived in a piece-wise homogeneous volume conductor. The equations are discretized with the method of weighted residuals, enabling the use of Galerkin weighting in the numerical solution of the equations. The construction of the transfer matrix is described in detail for constant and linear collocation and Galerkin methods, and the resulting forward transfer matrices are validated via simple numerical simulations. The linear Galerkin method is found to generate the smallest errors. The presented method increases the accuracy of the forward-computed body surface potential and thus prepares the way for more accurate inverse reconstructions of epicardial potential.

In this paper, we present a series of phantom experiments that aim to systematically evaluate the possibility of quantitative three-dimensional (3D) photoacoustic imaging of the finger joints in a spherical scanning geometry. The 3D absorption coefficient images of the joint-mimicking phantoms are obtained using our finite-element-based photoacoustic image reconstruction algorithm coupled with the photon diffusion equation. The results show that a 1 mm thick 'cartilage' can be accurately differentiated from the 'bones' with a 1 MHz transducer. In addition, the absorption coefficient of the 'cartilage' can be effectively recovered when this optical property is varied from 0.015 to 0.04 mm⁻¹. The phantom images shown in this paper suggest that 3D photoacoustic tomography in a spherical scanning geometry has the potential to image a realistic human joint.

For the treatment of conjunctival lymphoma in the early stages, external beam radiotherapy offers a curative approach. Such treatment requires the use of highly conformed small radiation beams. The beam size is so small that even advanced treatment planning systems have difficulties in calculating dose distributions. One possible approach for optimizing the treatment technique and later performing treatment planning is by means of full Monte Carlo (MC) simulations. In this paper, we compare experimental absorbed dose profiles obtained with a collimator used at the University Hospital Essen, with MC simulations done with the general-purpose radiation transport code PENELOPE. The collimator is also simulated with the hybrid MC code electron Monte Carlo (eMC) implemented in the commercial treatment planning system Eclipse (Varian). The results obtained with PENELOPE have a maximum difference with experimental data of 2.3%, whereas the eMC code differs systematically from the experimental data about 7% in the penumbra tails. We also show that PENELOPE simulations are able to obtain absorbed dose maps with an equivalent statistical uncertainty to the one found with eMC in similar CPU times.

An accurate assessment of the extent of the tumor is critical for successful local treatment of lung cancer by surgery and/or radiotherapy. Guidelines to establish the extent of treatment margins may be derived from correlation studies between pre-treatment imaging and histopathology. Deformations occur, however, between in-vivo CT imaging and ex-vivo pathology due to the softness of lung tissue and pathology processing. The first aim of this study was to quantify these deformations in tissue around non-small cell lung cancer. The second aim was to explore factors associated with the magnitude of the deformations. The study was performed in 25 patients who underwent lobectomy after preoperative CT. Non-rigid registration was employed to evaluate tissue deformations around the gross tumor volume (GTV), taking into account potential differences in elasticity between tumor and healthy lung tissue. Tissue was found to be compacted by approximately 60% depending on circularity of the tumor and orientation of the specimen on the pathology table during processing. The deformations give rise to potential underestimation of the treatment margins in pathology studies that do not take this aspect into account.

In this study, the radiation emission from mobile phones when used with wireless and wired hands-free kits (HFK) was evaluated to determine the necessity for a dedicated compliance procedure and the extent to which the use of wired and wireless HFK can reduce human exposure. The specific absorption rates (SAR) from wireless HFK were determined experimentally. Wired HFK were evaluated dosimetrically while connected to mobile phones (GSM900/1800, UMTS1950) under maximized current coupling onto the HFK cable and various wire routing configurations. In addition, experimentally validated simulations of a wired HFK and a mobile phone operating on anatomical whole-body models were performed. The maximum spatial peak SAR in the head when using wired HFK was more than five times lower than ICNIRP limits. The SAR in the head depends on the output power of the mobile phone, the coupling between the antenna and cable, external attenuation and potential cable specific attenuation. In general, a wired HFK considerably reduces the exposure of the entire head region compared to mobile phones operated at the head, even under unlikely worst-case coupling scenarios. However, wired HFK may cause a localized increase of the exposure in the region of the ear inside the head under worst-case conditions. Wireless HFK exhibit a low but constant exposure.

Dose distributions generated on a static anatomy may differ significantly from those delivered to temporally varying anatomy such as for abdominal and thoracic tumors, due largely in part to the unavoidable organ motion and deformation effects stemming from respiration. In this work, the degree of such variation for three treatment techniques, namely static conventional, gating and target tracking radiotherapy, was investigated. The actual delivered dose was approximated by planning all the phases of a 4DCT image set. Data from six (n = 6) previously treated lung cancer patients were used for this study with tumor motion ranging from 2 to 10 mm. Complete radiobiological analyses were performed to assess the clinical significance of the observed discrepancies between the 3D and 4DCT image-based dose distributions. Using the complication-free tumor control probability (P₊) objective, we observed small differences in P₊ between the 3D and 4DCT image-based plans (<2.0% difference on average) for the gating and static conventional regimens and higher differences in P₊ (4.0% on average) for the tracking regimen. Furthermore, we observed, as a general trend, that the 3D plan underestimated the P₊ values. While it is not possible to draw any general conclusions from a small patient cohort, our results suggest that there exists a patient population in which 4D planning does not provide any additional benefits beyond that afforded by 3D planning for static conventional or gated radiotherapy. This statement is consistent with previous studies based on physical dosimetric evaluations only. The higher differences observed with the tracking technique suggest that individual patient plans should be evaluated on a case-by-case basis to assess if 3D or 4D imaging is appropriate for the tracking technique.

The PET tracer O-(2-[¹⁸F]Fluoroethyl)-l-tyrosine (FET) has been shown to be valuable for different roles in the management of brain tumours. The aim of this study was to evaluate several quantitative measures of dynamic FET PET imaging in patients with resected glioblastoma. We evaluated dynamic FET PET in nine patients with histologically confirmed glioblastoma. Following FET PET, all subjects had radiation and chemotherapy. Tumour ROIs were defined by a threshold-based region-growing algorithm. We compared several standard measures of tumour uptake and uptake kinetics: SUV, SUV/background, distribution volume ratio (DVR), weighted frame differences and compartment model parameters. These measures were correlated with disease-free and overall survival, and analysed for statistical significance. We found that several measures allowed robust quantification. SUV and distribution volume did not correlate with clinical outcome. Measures that are based on a background region (SUV/BG, Logan-DVR) highly correlated with disease-free survival (r = −0.95, p < 0.0001), but not overall survival. Some advanced measures also showed a prognostic value but no improvement over the simpler methods. We conclude that FET PET probably has a prognostic value in patients with resected glioblastoma. The ratio of SUV to background may provide a simple and valuable predictive measure of the clinical outcome. Further studies are needed to confirm these explorative results.

We have modeled, by finite element analysis, the process of heating of a spherical gold nanoparticle by nanosecond laser pulses and of heat transfer between the particle and the surrounding medium, with no mass transfer. In our analysis, we have included thermal conductivity changes, vapor formation, and changes of the dielectric properties as a function of temperature. We have shown that such changes significantly affect the temperature reached by the particle and surrounding microenvironment and therefore the thermal and dielectric properties of the medium need to be known for a correct determination of the temperature elevation. We have shown that for sufficiently low intensity and long pulses, it is possible to establish a quasi-steady temperature profile in the medium with no vapor formation. As the intensity is increased, a phase-change with vapor formation takes place around the gold nanoparticle. As phase-transition starts, an additional increase in the intensity does not significantly increase the temperature of the gold nanoparticle and surrounding environment. The temperature starts to rise again above a given intensity threshold which is particle and environment dependent. The aim of this study is to provide useful insights for the development of molecular targeting of gold nanoparticles for applications such as remote drug release of therapeutics and photothermal cancer therapy.

The aim of this study was to investigate the utility of consistency metrics, such as inverse consistency, in contour-based deformable registration error analysis. Four images were acquired of the same phantom that has experienced varying levels of deformation. The deformations were simulated with deformable image registration. Using calculated deformation maps, the inconsistencies within the algorithm were investigated. This can be done, for example, by calculating deformation maps both in forward and reverse directions and applying them subsequently to an image. If the algorithm is not inverse consistent, then this final image will not be the same as the original, as it should be. Other consistency tests were done, for example by comparing different algorithms or by applying the deformation maps to a circular set of multiple deformations, whereby the original and final images are in fact the same. The resulting composite deformation map in this case contains a combination of the errors within those maps, because if error free, the resulting deformation map should be zero everywhere. We have termed this the generalized inverse consistency error map (). The correlation between the consistency metrics and registration error varied considerably depending on the registration algorithm and type of consistency metric. There was also a trend for the actual registration error to be larger than the consistency metrics. A disadvantage of these techniques is that good performance in these consistency checks is a necessary but not sufficient condition for an accurate deformation method.

A non-destructive, diffuse reflectance near infrared spectroscopy (DR-NIRS) approach is considered as a potential tool for determining the component-level structural properties of articular cartilage. To this end, DR-NIRS was applied in vitro to detect structural changes, using principal component analysis as the statistical basis for characterization. The results show that this technique, particularly with first-derivative pretreatment, can distinguish normal, intact cartilage from enzymatically digested cartilage. Further, this paper establishes that the use of DR-NIRS enables the probing of the full depth of the uncalcified cartilage matrix, potentially allowing the assessment of degenerative changes in joint tissue, independent of the site of initiation of the osteoarthritic process.

This work investigates variation in the volume of the prostate measured at different stages through the prostate brachytherapy procedure for 30 patients treated with I-125 radioactive seeds. The implanted seeds were localized on post-implantation ultrasound (US) images and the effect of prostate enlargement due to edema on dose coverage for 15 patients was studied. The volume of the prostate was measured at four stages as follows: (a) 2–3 weeks prior to implantation using US imaging, (b) then at the start of the intra-operative prostate brachytherapy procedure on the day of the implant, (c) immediately post-implantation using US imaging in the operating room and (d) finally by CT imaging at nearly 4 weeks post-implantation. Comparative prostate volume studies were performed using US imaging stepper and twister modes. For the purpose of this study, the implanted seeds were localized successfully on post-implant ultrasound twister images, retrospectively. The plans using post-implant US imaging were compared with intra-operative plans on US and plans created on CT images. The prostate volume increases about 10 cm³ on average due to edema induced by needle insertion and seed loading during implantation. The visibility of the implanted seeds on US twister images acquired post-implantation is as good as those on CT images and can be localized and used for dose calculation. The dose coverage represented by parameters such as D90 (dose covering 90% of the volume) and V100 (volume covered by 100% dose) is poorer on plans performed on post-implantation twister US studies than on the intra-operative live plan or the CT scan performed 4 weeks post-operatively. For example, the mean D90 difference on post-implantation US is lower by more than 15% than that on pre-implantation US. The volume enlargement of the prostate due to edema induced by needle insertion and seed placement has a significant effect on the quality of dosimetric coverage in brachytherapy prostate seed implants. Here, we introduced a new approach based on the use of post-implant US twister images to correct for prostate enlargement intra-operatively. Besides the ability to localize the seeds and superior soft tissue visibility, the twister US images include effects of the enlargement of the prostate gland and seed migration during the implantation procedure.

Previous studies have demonstrated that simple intensity-modulated radiotherapy (IMRT) plans can be produced with a series of rectangular segments formed by conventional jaws. This study investigates whether complex IMRT plans for head and neck cancer can be delivered with the conventional jaws efficiently. Six nasopharyngeal cancer patients, previously treated with multi-leaf collimator (MLC)-IMRT plans, were re-planned using conventional jaw delivery options. All IMRT plans were subject to the plan acceptance criteria of the RTOG-0225 protocol. For a selected patient, the maximum number of segments varied from five to nine per beam, and was tested for both jaws-only IMRT (JO-IMRT) plans and MLC-IMRT plans. Subsequently, JO-IMRT plans and MLC-IMRT on the same treatment planning system were attempted for all patients with identical beams. The dose distribution, dose volume histograms (DVH), the conformal index (COIN), the uniformity index and delivery efficiency were compared between the MLC-IMRT and JO-IMRT plans. All JO-IMRT plans met the RTOG-0225 criteria for tumor coverage and sensitive structures sparing. The corresponding MLC-IMRT and JO-IMRT plans show comparable conformality and uniformity, with average COINs of the planning gross tumor volume(pGTV) 37.7% ± 18.7% versus 37.9% ± 18.1%, and the average uniformity index 82.8% ± 2.5% versus 83.6% ± 3.1%, respectively. The average monitor unit for JO-IMRT plans was about twice that of MLC-IMRT plans. In conclusion, conventional jaws can be used solely to deliver complex IMRT plans for patients with nasopharyngeal cancer yet still within a practical delivery time.

Atherosclerotic plaque rupture is the leading cause of mortality in cardiovascular disease. Intravascular ultrasound (IVUS) imaging is a powerful clinical technique that provides real-time cross-sectional images of the arterial wall and atherosclerotic plaques. However, it does not provide sufficient information about the histological composition of plaques to characterize their vulnerability. Arterial wall strain measurements may provide insights into plaque composition and vulnerability, complementing the information directly available in the IVUS echogram. We have developed a method to measure the transverse arterial wall strain tensor in response to luminal pressure change, by registering IVUS images acquired at different pressures. This method has been validated by using IVUS images with simulated motion and IVUS images of a vessel phantom. In this study, we further evaluate the method by assessing the correspondence of the calculated strain distribution and the histological composition of atherosclerotic coronary arteries from Sinclair miniature pigs following 12 months of a high fat diet. The images were acquired in situ using a clinical IVUS system and under computer-controlled pressurization. After image acquisition, the artery segments were fixed for histology to identify plaque components. The strain distributions were aligned with the corresponding histological sections. The stiffness of various components of the lesion, inferred from the wall strain distribution, was consistent with the tissue composition seen in the histological cross-sections. These findings suggest that strain measurements from IVUS are promising for assessing plaque vulnerability.

In practice, gated cardiac SPECT images suffer from a number of degrading factors, including distance-dependent blur, attenuation, scatter and increased noise due to gating. Recently, we proposed a motion-compensated approach for four-dimensional (4D) reconstruction for gated cardiac SPECT and demonstrated that use of motion-compensated temporal smoothing could be effective for suppressing the increased noise due to lowered counts in individual gates. In this work, we further develop this motion-compensated 4D approach by also taking into account attenuation and scatter in the reconstruction process, which are two major degrading factors in SPECT data. In our experiments, we conducted a thorough quantitative evaluation of the proposed 4D method using Monte Carlo simulated SPECT imaging based on the 4D NURBS-based cardiac-torso (NCAT) phantom. In particular, we evaluated the accuracy of the reconstructed left ventricular myocardium using a number of quantitative measures including regional bias-variance analyses and wall intensity uniformity. The quantitative results demonstrate that use of motion-compensated 4D reconstruction can improve the accuracy of the reconstructed myocardium, which in turn can improve the detectability of perfusion defects. Moreover, our results reveal that while traditional spatial smoothing could be beneficial, its merit would become diminished with the use of motion-compensated temporal regularization. As a preliminary demonstration, we also tested our 4D approach on patient data. The reconstructed images from both simulated and patient data demonstrated that our 4D method can improve the definition of the LV wall.

We describe a phantom for simultaneous electroencephalography (EEG) and near-infrared imaging which consists of a solid, optically turbid and electrically conducting interface enclosing a tissue-mimicking aqueous scattering solution. The interface provides an electrical contact impedance comparable to that of the human scalp while the phantom as a whole has optical properties and electrical conductivity equivalent to that of head tissue. The construction is described and our design is evaluated experimentally using an optically absorbing target which also provides an EEG-equivalent electric field source. The results of this simultaneous EEG and near-infrared imaging experiment are presented.

In the UMC Utrecht a prototype MRI accelerator has been installed to investigate the feasibility of real-time, MRI guided radiotherapy. The system consists of a 6 MV Elekta (Crawley, UK) accelerator and a 1.5 T Philips (Best, The Netherlands) MRI system. The system is installed in a standard radiotherapy bunker. The bunker is at the corner of a block of six bunkers, so there are three neighbouring clinical Elekta accelerators. During ramping of the magnet, the magnetic fringe field in the two nearest bunkers was measured as a function of the magnetic field strength of the MRI magnet. At 8 m, a maximum increase of 1.5 G was measured, at 12 m, 0.6 G. This is up to three times the earth's magnetic field. The clinical accelerators are needed to be re-calibrated in order to operate in such an external magnetic field. The resulting radiation field flatness of the clinical accelerators was measured and was similar to the situation before ramping the magnet.

In magnetic drug targeting a chemotherapeutic agent is bound to coated magnetic nanoparticles, which are administered to the blood vessel system and subsequently focused by an external applied magnetic field. The optimization of intra-arterial magnetic drug targeting (MDT) requires detailed knowledge about the biodistribution of particles in the artery and the respective surrounding after the application. Here, we demonstrate the potential of magnetorelaxometry for quantifying the distribution of magnetic nanoparticles in the artery. To this end, we present a magnetorelaxometry investigation of a MDT study in an artery model. In particular, the absolute magnetic nanoparticle accumulation along the artery as well as the uptake profile along the region around the MDT-magnet position was quantified.

Recently, in the compressed sensing framework we proved that an interior ROI can be exactly reconstructed via the total variation minimization if the ROI is piecewise constant. In the proofs, we implicitly utilized the property that if an artifact image assumes a constant value within the ROI, then this constant must be zero. Here we prove this property in the space of square integrable functions.

The authors would like to thank Drs Kneschaurek and Wehrmann for pointing out a regrettable omission in our paper. We would have gladly quoted their superb paper (Kneschaurek and Wehrmann 1987 Med. Phys.14 602–607) had it not escaped our computer-aided literature search. We are happy that the two different mathematical procedures led to the same basic results.

In past years Physics in Medicine and Biology (PMB) published some special 'conference issues' given over to papers from a specific event. This policy changed some years ago for several reasons. Firstly, IOP Publishing launched its own conference publication journal (the open access Journal of Physics: Conference Series). Secondly, producing special issues of PMB tended to adversely affect the scheduling of regular peer-reviewed papers. Thirdly, the growing number of requests to accommodate such special issues (presumably because of the popularity of the journal) led to an unworkable situation where worthy conferences had to be turned down. Fourthly, there was some concern about the uniformity of refereeing standards when conference organizers themselves refereed papers.

However, it may not be so well known that the journal is very happy to consider grouping a number of research papers from a specific conference, subject to certain conditions of standardization. The papers must be taken through the usual (rigorous) refereeing process managed in-house by PMB. The conference organizer would be responsible for pre-filtering the papers so only the cream are submitted; they are also responsible for managing the close timing of submissions. In this issue we feature three of the best papers from ICMLA '08.

The 2008 International Conference on Machine Learning and Applications (ICMLA '08) was held in San Diego, California, USA on 11–13 December 2008. The aim of the conference was to bring researchers working in the areas of machine learning and applications together. The conference covered both theoretical and experimental research results on machine learning applications in fields like medicine, biology, industry, virtual environments, game playing and problem solving. As part of the conference, a special session on 'Applications of Machine Learning in Radiotherapy' was organized by Dr Steve Jiang of University of California San Diego and Dr Martin Murphy of Virginia Commonwealth University. The aim was to provide a platform to present and discuss recent advancements in the application of machine learning methods in radiotherapy, specifically in helping accurately localize tumours in fluoroscopic images, precisely target the radiation to the tumours, analyze treatment outcomes, and improve treatment quality and patient safety.

The journal welcomes papers from similar high-quality conferences and would organize to take them through this route.

Steve WebbEditor-in-Chief

Simon HarrisPublisher

We propose a novel approach for potential online treatment verification using cine EPID (electronic portal imaging device) images for hypofractionated lung radiotherapy based on a machine learning algorithm. Hypofractionated radiotherapy requires high precision. It is essential to effectively monitor the target to ensure that the tumor is within the beam aperture. We modeled the treatment verification problem as a two-class classification problem and applied an artificial neural network (ANN) to classify the cine EPID images acquired during the treatment into corresponding classes—with the tumor inside or outside of the beam aperture. Training samples were generated for the ANN using digitally reconstructed radiographs (DRRs) with artificially added shifts in the tumor location—to simulate cine EPID images with different tumor locations. Principal component analysis (PCA) was used to reduce the dimensionality of the training samples and cine EPID images acquired during the treatment. The proposed treatment verification algorithm was tested on five hypofractionated lung patients in a retrospective fashion. On average, our proposed algorithm achieved a 98.0% classification accuracy, a 97.6% recall rate and a 99.7% precision rate.

Radiotherapy outcomes are determined by complex interactions between treatment, anatomical and patient-related variables. A common obstacle to building maximally predictive outcome models for clinical practice is the failure to capture potential complexity of heterogeneous variable interactions and applicability beyond institutional data. We describe a statistical learning methodology that can automatically screen for nonlinear relations among prognostic variables and generalize to unseen data before. In this work, several types of linear and nonlinear kernels to generate interaction terms and approximate the treatment-response function are evaluated. Examples of institutional datasets of esophagitis, pneumonitis and xerostomia endpoints were used. Furthermore, an independent RTOG dataset was used for 'generalizabilty' validation. We formulated the discrimination between risk groups as a supervised learning problem. The distribution of patient groups was initially analyzed using principle components analysis (PCA) to uncover potential nonlinear behavior. The performance of the different methods was evaluated using bivariate correlations and actuarial analysis. Over-fitting was controlled via cross-validation resampling. Our results suggest that a modified support vector machine (SVM) kernel method provided superior performance on leave-one-out testing compared to logistic regression and neural networks in cases where the data exhibited nonlinear behavior on PCA. For instance, in prediction of esophagitis and pneumonitis endpoints, which exhibited nonlinear behavior on PCA, the method provided 21% and 60% improvements, respectively. Furthermore, evaluation on the independent pneumonitis RTOG dataset demonstrated good generalizabilty beyond institutional data in contrast with other models. This indicates that the prediction of treatment response can be improved by utilizing nonlinear kernel methods for discovering important nonlinear interactions among model variables. These models have the capacity to predict on unseen data.

Computer-aided diagnosis (CAD) has been an active area of study in medical image analysis. A filter for the enhancement of lesions plays an important role for improving the sensitivity and specificity in CAD schemes. The filter enhances objects similar to a model employed in the filter; e.g. a blob-enhancement filter based on the Hessian matrix enhances sphere-like objects. Actual lesions, however, often differ from a simple model; e.g. a lung nodule is generally modeled as a solid sphere, but there are nodules of various shapes and with internal inhomogeneities such as a nodule with spiculations and ground-glass opacity. Thus, conventional filters often fail to enhance actual lesions. Our purpose in this study was to develop a supervised filter for the enhancement of actual lesions (as opposed to a lesion model) by use of a massive-training artificial neural network (MTANN) in a CAD scheme for detection of lung nodules in CT. The MTANN filter was trained with actual nodules in CT images to enhance actual patterns of nodules. By use of the MTANN filter, the sensitivity and specificity of our CAD scheme were improved substantially. With a database of 69 lung cancers, nodule candidate detection by the MTANN filter achieved a 97% sensitivity with 6.7 false positives (FPs) per section, whereas nodule candidate detection by a difference-image technique achieved a 96% sensitivity with 19.3 FPs per section. Classification-MTANNs were applied for further reduction of the FPs. The classification-MTANNs removed 60% of the FPs with a loss of one true positive; thus, it achieved a 96% sensitivity with 2.7 FPs per section. Overall, with our CAD scheme based on the MTANN filter and classification-MTANNs, an 84% sensitivity with 0.5 FPs per section was achieved.