Automatic extraction of the intracranial volume in fetal and neonatal MR scans using convolutional neural networks.

MR images of infants and fetuses allow non-invasive analysis of the brain. Quantitative analysis of brain development requires automatic brain tissue segmentation that is typically preceded by segmentation of the intracranial volume (ICV). Fast changes in the size and morphology of the developing brain, motion artifacts, and large variation in the field of view make ICV segmentation a challenging task. We propose an automatic method for segmentation of the ICV in fetal and neonatal MRI scans. The method was developed and tested with a diverse set of scans regarding image acquisition parameters (i.e. field strength, image acquisition plane, image resolution), infant age (23-45 weeks post menstrual age), and pathology (posthaemorrhagic ventricular dilatation, stroke, asphyxia, and Down syndrome). The results demonstrate that the method achieves accurate segmentation with a Dice coefficient (DC) ranging from 0.98 to 0.99 in neonatal and fetal scans regardless of image acquisition parameters or patient characteristics. Hence, the algorithm provides a generic tool for segmentation of the ICV that may be used as a preprocessing step for brain tissue segmentation in fetal and neonatal brain MR scans.

Automatic brain tissue segmentation in fetal MRI using convolutional neural networks.

MR images of fetuses allow clinicians to detect brain abnormalities in an early stage of development. The cornerstone of volumetric and morphologic analysis in fetal MRI is segmentation of the fetal brain into different tissue classes. Manual segmentation is cumbersome and time consuming, hence automatic segmentation could substantially simplify the procedure. However, automatic brain tissue segmentation in these scans is challenging owing to artifacts including intensity inhomogeneity, caused in particular by spontaneous fetal movements during the scan. Unlike methods that estimate the bias field to remove intensity inhomogeneity as a preprocessing step to segmentation, we propose to perform segmentation using a convolutional neural network that exploits images with synthetically introduced intensity inhomogeneity as data augmentation. The method first uses a CNN to extract the intracranial volume. Thereafter, another CNN with the same architecture is employed to segment the extracted volume into seven brain tissue classes: cerebellum, basal ganglia and thalami, ventricular cerebrospinal fluid, white matter, brain stem, cortical gray matter and extracerebral cerebrospinal fluid. To make the method applicable to slices showing intensity inhomogeneity artifacts, the training data was augmented by applying a combination of linear gradients with random offsets and orientations to image slices without artifacts. To evaluate the performance of the method, Dice coefficient (DC) and Mean surface distance (MSD) per tissue class were computed between automatic and manual expert annotations. When the training data was enriched by simulated intensity inhomogeneity artifacts, the average achieved DC over all tissue classes and images increased from 0.77 to 0.88, and MSD decreased from 0.78 mm to 0.37 mm. These results demonstrate that the proposed approach can potentially replace or complement preprocessing steps, such as bias field corrections, and thereby improve the segmentation performance.

Volumetric assessment of extracranial carotid artery aneurysms.

The extracranial carotid artery aneurysm (ECAA) is a rare pathology for which clinical treatment guidelines are lacking. In general, symptoms or growth of the aneurysm sac are thought to indicate intervention. ECAAs may present in a large variety of shapes and sizes, and conventional diameter measurements fail to indicate geometrical differences. Therefore, we propose a protocol to measure ECAA size by 3D volumetric assessment. The volumes of 40 ECAAs in computed tomography angiography (CTA) images were measured through manual segmentation, by two independent operators. Volumes of the entire internal carotid artery (ICA) and the ECAA were measured separately. Excellent inter- and intraoperator reliability was found for both ICA and ECAA volumes, with all intraclass correlation coefficients above 0.94. Bland-Altman analysis revealed normal differences for both inter- and intraoperator agreement. For all volumes, similarity of the segmentations was excellent. Outliers were explained by presence of intraluminal ECAA thrombus, which hampered identification of the aneurysm outer wall. These results implicate robustness of our protocol, which is designed as a step-up towards (semi)automatic volumetric measurements to monitor patients with ECAA. Future (semi)automatic volumetric assessments are recommended and such techniques can be developed and validated using the proposed protocol and manual reference segmentations.

MR-based source localization for MR-guided HDR brachytherapy.

For the purpose of MR-guided high-dose-rate (HDR) brachytherapy, a method for real-time localization of an HDR brachytherapy source was developed, which requires high spatial and temporal resolutions. MR-based localization of an HDR source serves two main aims. First, it enables real-time treatment verification by determination of the HDR source positions during treatment. Second, when using a dummy source, MR-based source localization provides an automatic detection of the source dwell positions after catheter insertion, allowing elimination of the catheter reconstruction procedure. Localization of the HDR source was conducted by simulation of the MR artifacts, followed by a phase correlation localization algorithm applied to the MR images and the simulated images, to determine the position of the HDR source in the MR images. To increase the temporal resolution of the MR acquisition, the spatial resolution was decreased, and a subpixel localization operation was introduced. Furthermore, parallel imaging (sensitivity encoding) was applied to further decrease the MR scan time. The localization method was validated by a comparison with CT, and the accuracy and precision were investigated. The results demonstrated that the described method could be used to determine the HDR source position with a high accuracy (0.4-0.6 mm) and a high precision (⩽0.1 mm), at high temporal resolutions (0.15-1.2 s per slice). This would enable real-time treatment verification as well as an automatic detection of the source dwell positions.

Effects of hunger state on the brain responses to food cues across the life span.

The abundant exposure to food cues in our environment is one of the main drivers of overconsumption. Food evaluation is important for the regulation of food intake by the brain and it's interaction with hunger state. Children are especially susceptible to food cues. Understanding the mechanisms behind this regulation in healthy individuals across the life span can help to elucidate the mechanisms underlying overconsumption and aid the development of future obesity prevention strategies. Few functional neuroimaging studies have been done in children and elderly. Furthermore, it is unknown how hunger state affects neural food cue reactivity in these groups, since this has not been examined consistently. We examined the effects of hunger state and age on the brain responses to low- and high calorie foods. On two mornings, 122 participants (17 children; 38 teens; 36 adults; 31 elderly) performed a food image viewing task while being scanned using fMRI, either fasted or sated. Hunger induced greater activation during high versus low calorie food image viewing than satiety in the bilateral dorsomedial (dmPFC) and in the right dorsolateral prefrontal cortex (dlPFC) across all age groups. There was no significant main effect of age group on high versus low calorie food image viewing and no interaction between age group and hunger state. The greater activation of the dlPFC across all age groups during high calorie food image viewing in a fasted state might reflect increased inhibitory control in response to these foods. This may underlie the ability to resist overconsumption of high calorie foods. Furthermore, increased medial prefrontal cortex activation during hunger might reflect increased reward value of high calorie foods, which declines with satiation. Further studies are needed to better understand these results. Notably, overweight and obese individuals should be included to examine whether these responses are altered by weight status across the life span.

Preoperative indication for systemic therapy extended to patients with early-stage breast cancer using multiparametric 7-tesla breast MRI.

PURPOSE: To establish a preoperative decision model for accurate indication of systemic therapy in early-stage breast cancer using multiparametric MRI at 7-tesla field strength. MATERIALS AND METHODS: Patients eligible for breast-conserving therapy were consecutively included. Patients underwent conventional diagnostic workup and one preoperative multiparametric 7-tesla breast MRI. The postoperative (gold standard) indication for systemic therapy was established from resected tumor and lymph-node tissue, based on 10-year risk-estimates of breast cancer mortality and relapse using Adjuvant! Online. Preoperative indication was estimated using similar guidelines, but from conventional diagnostic workup. Agreement was established between preoperative and postoperative indication, and MRI-characteristics used to improve agreement. MRI-characteristics included phospomonoester/phosphodiester (PME/PDE) ratio on 31-phosphorus spectroscopy (31P-MRS), apparent diffusion coefficients on diffusion-weighted imaging, and tumor size on dynamic contrast-enhanced (DCE)-MRI. A decision model was built to estimate the postoperative indication from preoperatively available data. RESULTS: We included 46 women (age: 43-74yrs) with 48 invasive carcinomas. Postoperatively, 20 patients (43%) had positive, and 26 patients (57%) negative indication for systemic therapy. Using conventional workup, positive preoperative indication agreed excellently with positive postoperative indication (N = 8/8; 100%). Negative preoperative indication was correct in only 26/38 (68%) patients. However, 31P-MRS score (p = 0.030) and tumor size (p = 0.002) were associated with the postoperative indication. The decision model shows that negative indication is correct in 21/22 (96%) patients when exempting tumors larger than 2.0cm on DCE-MRI or with PME>PDE ratios at 31P-MRS. CONCLUSIONS: Preoperatively, positive indication for systemic therapy is highly accurate. Negative indication is highly accurate (96%) for tumors sized ≤2,0cm on DCE-MRI and with PME≤PDE ratios on 31P-MRS.

Improved estimation of MR relaxation parameters using complex-valued data.

PURPOSE: In MR image analysis, T1 , T2 , and T2* maps are generally calculated using magnitude MR data. Without knowledge of the underlying noise variance, parameter estimates at low signal to noise ratio (SNR) are usually biased. This leads to confounds in studies that compare parameters across SNRs and or across scanners. This article compares several estimation techniques which use real or complex-valued MR data to achieve unbiased estimation of MR relaxation parameters without the need for additional preprocessing. THEORY AND METHODS: Several existing and new techniques to estimate relaxation parameters using complex-valued data were compared with widely used magnitude-based techniques. Their bias, variance and processing times were studied using simulations covering various aspects of parameter variations. Validation on noise-degraded experimental measurements was also performed. RESULTS: Simulations and experiments demonstrated the superior performance of techniques based on complex-valued data, even in comparison with magnitude-based techniques that account for Rician noise characteristics. This was achieved with minor modifications to data modeling and at computational costs either comparable to or higher ( ≈two fold) than magnitude-based estimators. Theoretical analysis shows that estimators based on complex-valued data are statistically efficient. CONCLUSION: The estimation techniques that use complex-valued data provide minimum variance unbiased estimates of parametric maps and markedly outperform commonly used magnitude-based estimators under most conditions. They additionally provide phase maps and field maps, which are unavailable with magnitude-based methods. Magn Reson Med 77:385-397, 2017. © 2016 Wiley Periodicals, Inc.

Monte Carlo study of the impact of a magnetic field on the dose distribution in MRI-guided HDR brachytherapy using Ir-192.

In the process of developing a robotic MRI-guided high-dose-rate (HDR) prostate brachytherapy treatment, the influence of the MRI scanner's magnetic field on the dose distribution needs to be investigated. A magnetic field causes a deflection of electrons in the plane perpendicular to the magnetic field, and it leads to less lateral scattering along the direction parallel with the magnetic field. Monte Carlo simulations were carried out to determine the influence of the magnetic field on the electron behavior and on the total dose distribution around an Ir-192 source. Furthermore, the influence of air pockets being present near the source was studied. The Monte Carlo package Geant4 was utilized for the simulations. The simulated geometries consisted of a simplified point source inside a water phantom. Magnetic field strengths of 0 T, 1.5 T, 3 T, and 7 T were considered. The simulation results demonstrated that the dose distribution was nearly unaffected by the magnetic field for all investigated magnetic field strengths. Evidence was found that, from a dose perspective, the HDR prostate brachytherapy treatment using Ir-192 can be performed safely inside the MRI scanner. No need was found to account for the magnetic field during treatment planning. Nevertheless, the presence of air pockets in close vicinity to the source, particularly along the direction parallel with the magnetic field, appeared to be an important point for consideration.

Dual-head gamma camera system for intraoperative localization of radioactive seeds.

Breast-conserving surgery is a standard option for the treatment of patients with early-stage breast cancer. This form of surgery may result in incomplete excision of the tumor. Iodine-125 labeled titanium seeds are currently used in clinical practice to reduce the number of incomplete excisions. It seems likely that the number of incomplete excisions can be reduced even further if intraoperative information about the location of the radioactive seed is combined with preoperative information about the extent of the tumor. This can be combined if the location of the radioactive seed is established in a world coordinate system that can be linked to the (preoperative) image coordinate system. With this in mind, we propose a radioactive seed localization system which is composed of two static ceiling-suspended gamma camera heads and two parallel-hole collimators. Physical experiments and computer simulations which mimic realistic clinical situations were performed to estimate the localization accuracy (defined as trueness and precision) of the proposed system with respect to collimator-source distance (ranging between 50 cm and 100 cm) and imaging time (ranging between 1 s and 10 s). The goal of the study was to determine whether or not a trueness of 5 mm can be achieved if a collimator-source distance of 50 cm and imaging time of 5 s are used (these specifications were defined by a group of dedicated breast cancer surgeons). The results from the experiments indicate that the location of the radioactive seed can be established with an accuracy of 1.6 mm ± 0.6 mm if a collimator-source distance of 50 cm and imaging time of 5 s are used (these experiments were performed with a 4.5 cm thick block phantom). Furthermore, the results from the simulations indicate that a trueness of 3.2 mm or less can be achieved if a collimator-source distance of 50 cm and imaging time of 5 s are used (this trueness was achieved for all 14 breast phantoms which were used in this study). Based on these results we conclude that the proposed system can be a valuable tool for (real-time) intraoperative breast cancer localization.

Comparison of DCE-CT models for quantitative evaluation of K(trans) in larynx tumors.

Dynamic contrast enhanced CT (DCE-CT) can be used to estimate blood perfusion and vessel permeability in tumors. Tumor induced angiogenesis is generally associated with disorganized microvasculature with increased permeability or leakage. Estimated vascular leakage (K(trans)) values and their reliability greatly depend on the perfusion model used. To identify the preferred model for larynx tumor analysis, several perfusion models frequently used for estimating permeability were compared in this study. DCE-CT scans were acquired for 16 larynx cancer patients. Larynx tumors were delineated based on whole-mount histopathology after laryngectomy. DCE-CT data within these delineated volumes were analyzed using the Patlak and Logan plots, the Extended Tofts Model (ETM), the Adiabatic Approximation to the Tissue Homogeneity model (AATH) and a variant of AATH with fixed transit time (AATHFT). Akaike's Information Criterion (AIC) was used to identify the best fitting model. K(trans) values from all models were compared with this best fitting model. Correlation strength was tested with two-tailed Spearman's rank correlation and further examined using Bland-Altman plots. AATHFT was found to be the best fitting model. The overall median of individual patient medians K(trans) estimates were 14.3, 15.1, 16.1, 2.6 and 22.5 mL/100 g min( - 1) for AATH, AATHFT, ETM, Patlak and Logan, respectively. K(trans) estimates for all models except Patlak were strongly correlated (P < 0.001). Bland-Altman plots show large biases but no significant deviating trend for any model other than Patlak. AATHFT was found to be the preferred model among those tested for estimation of K(trans) in larynx tumors.

Registration of structurally dissimilar images in MRI-based brachytherapy.

A serious challenge in image registration is the accurate alignment of two images in which a certain structure is present in only one of the two. Such topological changes are problematic for conventional non-rigid registration algorithms. We propose to incorporate in a conventional free-form registration framework a geometrical penalty term that minimizes the volume of the missing structure in one image. We demonstrate our method on cervical MR images for brachytherapy. The intrapatient registration problem involves one image in which a therapy applicator is present and one in which it is not. By including the penalty term, a substantial improvement in the surface distance to the gold standard anatomical position and the residual volume of the applicator void are obtained. Registration of neighboring structures, i.e. the rectum and the bladder is generally improved as well, albeit to a lesser degree.

CT brain perfusion protocol to eliminate the need for selecting a venous output function.

BACKGROUND AND PURPOSE: In CTP, an arterial input function is used for cerebral blood volume measurement. AIFs are often influenced by partial volume effects resulting in overestimated CBV. A venous output function is manually selected to correct for partial volume. This can introduce variability. Our goal was to develop a CTP protocol that enables AIF selection unaffected by partial volume. MATERIALS AND METHODS: First, the effects of partial volume on artery sizes/types including the MCA were estimated by using a CTP phantom with 9 protocols (section thicknesses of 1, 1.8, and 5 mm and image resolutions of 0.5, 1, and 1.5 mm). Next, these protocols were applied to clinical CTP studies from 6 patients. The influence of the partial volume effect was measured by comparison of the time-attenuation curves from different artery locations with reference veins. RESULTS: AIFs from MCAs were unaffected by partial volume effects when using high image resolution (1 mm) and medium section thickness (1.8 mm). For the clinical data, a total of 104 arteries and 60 veins was selected. The data confirmed that high image resolution and thin section thickness enable selection of MCAs for AIFs free of partial volume influences. In addition, we found that large veins were not insusceptible to partial volume effects relative to large arteries, questioning the use of veins for partial volume correction. CONCLUSIONS: A CTP protocol with 1.8-mm section thickness and 1-mm image resolution allows AIF selection unaffected by partial volume effects in MCAs.

Visual cerebral microbleed detection on 7T MR imaging: reliability and effects of image processing.

SUMMARY: MR imaging at 7T has a high sensitivity for cerebral microbleed detection. We identified mIP processing conditions with an optimal balance between the number of visually detected microbleeds and the number of sections on 7T MR imaging. Even with optimal mIP processing, the limited size of some of the microbleeds and the susceptibility effects of other adjacent structures were a challenge for visual detection, which led to a modest inter-rater agreement, mainly due to missed microbleeds. Automated lesion-detection techniques may be required to optimally benefit from the increased spatial resolution offered by 7T MR imaging.

Spatially regularized region-based perfusion estimation in peripherals using angiographic C-arm systems.

The outcome assessment of endovascular revascularization procedures in the lower limbs is currently carried out by x-ray digital subtraction angiography (DSA). Due to the two-dimensional nature of this technique, only visual assessment of arterial blood flow is possible and no tissue blood flow information (i.e. perfusion) is available to assess the effective restoration of blood supply to the tissue. In this work, we propose a method for interventional perfusion estimation in peripherals using C-arms which is based on DSA and two additional 3D images reconstructed from rotational scans. The method assumes spatial homogeneity of contrast within multiple regions identified by segmentation of the reconstructed 3D images. A dedicated segmentation method which relies on local contrast homogeneity and connectivity of anatomical structures is introduced. Region-based perfusion is obtained by mapping the 2D blood flow information from DSA to the 3D segments by solving an inverse problem. Instability of the solution due to the spatial overlap of the regions is addressed by applying spatial and temporal regularizations. The method was evaluated on data simulated from CT perfusion scans of the lower limb. Blood flow values estimated with the optimal number of segmented regions exhibited errors of 1 ± 4 and 2 ± 11 ml/100 ml min(-1) for the two analyzed cases, respectively, which showed to be sufficient to differentiate hypoperfused and normally perfused areas. The use of spatial and temporal regularization proved to be an effective way to limit inaccuracies due to instability in the solution of the inverse problem. Results in general proved the feasibility of C-arm interventional perfusion imaging by a combination of temporal information derived from DSA and spatial information derived from 3D reconstructions.

Improved arterial visualization in cerebral CT perfusion-derived arteriograms compared with standard CT angiography: a visual assessment study.

BACKGROUND AND PURPOSE: Invasive cerebral DSA has largely been replaced by CTA, which is noninvasive but has a compromised arterial view due to superimposed bone and veins. The purpose of this study was to evaluate whether arterial visualization in CTPa is superior to standard CTA, which would eliminate the need for an additional CTA scan to assess arterial diseases and therefore reduce radiation dose. MATERIALS AND METHODS: In this study, we included 24 patients with subarachnoid hemorrhage for whom CTA and CTP were available. Arterial quality and presence of superimposed veins and bone in CTPa were compared with CTA and scored by 2 radiologists by using a VAS (0%-100%). Average VAS scores were determined and VAS scores per patient were converted to a 10-point NRS. Arterial visualization was considered to be improved when the highest rate (NRS 10, VAS > 90%) was scored for arterial quality, and the lowest rate (NRS 1, VAS < 10%), for the presence of superimposed veins and bone. A sign test with continuity correction was used to test whether the number of cases with these rates was significant. RESULTS: Average VAS scores in the proximal area were 94% (arterial quality), 4% (presence of bone), and 7% (presence of veins). In this area, the sign test showed that a significant number of cases scored NRS 10 for arterial quality (P < .02) and NRS 1 for the presence of superimposed veins and bone (P < .01). CONCLUSIONS: Cerebral CTPa shows improved arterial visualization in the proximal area compared with CTA, with similar arterial quality but no superimposed bone and veins.

Influence of variation in semiflexed knee positioning during image acquisition on separate quantitative radiographic parameters of osteoarthritis, measured by Knee Images Digital Analysis.

OBJECTIVE: The clinical application of quantitative measurement of separate radiographic parameters of knee osteoarthritis (OA) might be hampered by a lack of reproducible semiflexed joint positioning during acquisition of radiographs. The influence of systematic variations in knee positioning on measurement of separate quantitative radiographic parameters was studied. METHODS: Five components of knee position during radiographic acquisition (beam height, lower and upper leg extension, internal rotation, and lateral shift) were systematically varied within a clinically relevant range, using three cadaver legs. The influence of these variations on the measurement of the separate quantitative radiographic parameters by Knee Images Digital Analysis (KIDA) was evaluated. Significant changes were validated in vivo. Changes were compared with differences during 2-year follow-up in a radiographic progression cohort of early OA. RESULTS: Systematic variation in upper and lower leg extension induced changes in the measurement of joint space width (JSW). Lower leg extension also influenced osteophyte area and eminence height measurement. Also bone density measurement was influenced by variation in all five position components. Variations were of clinical relevance compared with 2-year differences in knees with radiographic progression, and were confirmed in vivo. CONCLUSIONS: Variations in semiflexed knee positioning, which are considered to occur easily during image acquisition in trials and clinical practice despite standardization, are of significant influence on the quantitative measurement of most separate radiographic parameters of OA using KIDA. The additional value of quantitative measurement might improve significantly by better standardization during radiographic acquisition; with radiography still being the gold standard for structure-modification in OA.

Evaluation of separate quantitative radiographic features adds to the prediction of incident radiographic osteoarthritis in individuals with recent onset of knee pain: 5-year follow-up in the CHECK cohort.

OBJECTIVE: Detailed radiographic evaluation might enable the identification of osteoarthritis (OA) earlier in the disease. This study evaluated whether and which separate quantitative features on knee radiographs of individuals with recent onset knee pain are associated with incidence of radiographic OA and persistence and/or progression of clinical OA during 5-year follow-up. METHOD: From the Cohort Hip & Cohort Knee study participants with knee pain at baseline were evaluated. Radiographic OA development was defined as Kellgren & Lawrence (K&L) grade ≥ II at 5-year follow-up. Clinical OA was defined as persistent knee pain and as progression of Westen Ontario & McMaster Universities Osteoarthritis index (WOMAC) pain and function score during follow-up. At baseline radiographic damage was determined by quantitative measurement of separate features using Knee Images Digital Analysis, and by K&L-grading. RESULTS: Measuring osteophyte area [odds ratio (OR) =7.0] and minimum joint space width (OR=0.7), in addition to demographic and clinical characteristics, improved the prediction of radiographic OA 5 years later [area under curve receiver operating characteristic=0.74 vs 0.64 without radiographic features]. When the predictive score (based on multivariate regression coefficients) was larger than the cut-off for optimal specificity, the chance of incident radiographic OA was 54% instead of the prior probability of 19%. Evaluating separate quantitative features performed slightly better than K&L-grading (AUC=0.70). Radiographic characteristics hardly added to prediction of clinical OA. CONCLUSION: In individuals with onset knee pain, radiographic characteristics added to the prediction of radiographic OA development 5 years later. Quantitative radiographic evaluation in individuals with suspected OA is worthwhile when determining treatment strategies and designing clinical trials.

Connectivity and phase coherence in neural network models of interconnected Z(4)-bi-stable units.

A phenomenological neural network model with bi-stable oscillatory units is used to model up- and down-states. These states have been observed in vivo in biological neuronal systems and feature oscillatory, limit cycle type of behavior in the up-states. A network is formed by a set of interconnected units. Two different types of network layouts are considered in this work: networks with hierarchical connections and hubs and networks with random connections. The phase coherence between the different units is analyzed and compared to the connectivity distance between nodes. In addition the connectivity degree of a node is associated to the average phase coherence with all other units. The results show that we may be able to identify the set of hubs in a network based on the phase coherence estimates between the different nodes. If the network is very dense or randomly connected, the underlying network structure, however, can not be derived uniquely from the phase coherence.

Feasibility of bone density evaluation using plain digital radiography.

OBJECTIVE: For the radiographic evaluation of subchondral bone changes (sclerosis) in osteoarthritis (OA), bone density (BD) is commonly subjectively assessed. BD evaluation using plain digital radiography might be influenced by acquisition and post-processing (PP) settings. Objective of this study was to evaluate the effects of these settings on the measurement of BD using digital radiographs. METHODS: A bone density standard (BDS) of hydroxyapatite (HA) mimicked a BD range of 1.0-5.75 g/cm(2). Digital radiographs were acquired with variation in acquisition settings, and with clinical and minimal PP. An aluminum step wedge served as an internal reference to express the gray values of the BDS in mm aluminum equivalents (mmAl). The relation (R(2)) between actual BD and BD normalized to the reference wedge was evaluated with linear regression analyses for radiographs with variations in PP and acquisition settings. Precision of BD measurement of the BDS was evaluated for application in clinical practice. RESULTS: The correlation between actual BD and BD normalized to the reference was improved by changing PP from clinical (R(2)=0.96) to minimal (R(2)=0.98). Higher tube voltage [kilovolt (kV)] improved the correlation further. Even for clinical PP, average standard deviation (SD) was 0.97 mmAl, much smaller than the change of 2.51 mmAl clinically observed in early OA, which implies the feasibility of BD measurements on digital radiographs. CONCLUSION: Changing PP and acquisition settings in clinical practice can have profound effect on outcome. If done with care, accurate BD measurement is feasible using plain digital radiography.