Voxel Values Research Articles

White matter biomarker for predicting de novo Parkinson's disease using tract-based spatial statistics: a machine learning-based model.

Parkinson's disease (PD) is an irreversible, chronic degenerative disease of the central nervous system, potentially associated with cerebral white matter (WM) lesions. Investigating the microstructural alterations within the WM in the early stages of PD can help to identify the disease early and enable intervention to reduce the associated serious threats to health. This study selected 227 cases from the Parkinson's Progression Markers Initiative (PPMI) database, including 152 de novo PD patients and 75 normal controls (NC). Whole-brain voxel analysis of the WM was performed using the tract-based spatial statistics (TBSS) method. The WM regions with statistically significant differences (P<0.05) between the PD and NC groups were identified and used as masks. The mask was applied to each case's fractional anisotropy (FA) image to extract voxel values as feature vectors. Geometric dimensionality reduction was then applied to eliminate redundant values in the feature vectors. Subsequently, the cases were randomly divided into a training group (158 cases, including 103 PD patients and 55 NC) and a test group (69 cases, including 49 PD patients and 20 NC). The least absolute shrinkage and selection operator (LASSO) regression algorithm was employed to extract the minimal set of relevant features, then the random forest (RF) algorithm was utilized for classification using 5-fold cross validation. The resulting model was further integrated with clinical factors to create a comprehensive prediction model. In comparison to the NC group, the FA values in PD patients exhibited a statistically significant decrease (P<0.05), indicating the presence of widespread WM lesions across multiple brain regions. Moreover, the PD prediction model, constructed based on these WM lesion regions, yielded prediction accuracy (ACC) and area under the receiver operating characteristic (ROC) curve (AUC) values of 0.778 and 0.865 in the validation set, and 0.783 and 0.831 in the test set, respectively. Furthermore, the performance of the integrated model showed some improvement, with ACC and AUC values in the test set reaching 0.804 and 0.844, respectively. The quantitative calculation of WM lesion area on FA images using the TBSS method can serve as a neuroimaging biomarker for diagnosing and predicting early PD at the individual level. When integrated with clinical variables, the predictive performance improves.

Type A standard uncertainty evaluation in one measurement through uncertainty propagation from voxel values’ distribution for computed tomography metrology

According to the guide to the expression of uncertainty in measurement, ‘type A evaluation’ generally requires repeated measurements, which are time-consuming for CT scans. To solve this problem, we developed a method for estimating the standard deviation of measurement results in one measurement through uncertainty propagation, which can be regarded as repeatability standard deviation to evaluate the type A standard uncertainty. The method first fits the CT voxel value distribution, uses the ISO50 method to determine the spatial distribution of surface points from the voxel value distribution and edge shape interpolation, and then derives the measurement results by fitting geometric parameters with the least square algorithm. Finally, the standard deviation of the measurement results is evaluated according to the distribution of the surface point position through uncertainty propagation. We performed simulations and experiments using the hole-plate with 28 holes to compare the uncertainty evaluated by our method and the type A standard uncertainty evaluated on the basis of a series of observations obtained under repeatability conditions. Both simulation and experimental results show that these two uncertainties follow the same statistical variation pattern. The Pearson correlation coefficients of the two uncertainties in simulation and experiment are 0.79 and 0.33, respectively, indicating that the uncertainty evaluated by the proposed method can directly replace the type A uncertainty or provide a reference similar to type A uncertainty for the evaluation of the combined uncertainty.

Radiomics-Based Prediction Model for Outcome of Radioembolization in Metastatic Colorectal Cancer.

To evaluate the benefit of a contrast-enhanced computed tomography (CT) radiomics-based model for predicting response and survival in patients with colorectal liver metastases treated with transarterial Yttrium-90 radioembolization (TARE). Fifty-one patients who underwent TARE were included in this single-center retrospective study. Response to treatment was assessed using the Response Evaluation Criteria in Solid Tumors (RECIST 1.1) at 3-month follow-up. Patients were stratified as responders (complete/partial response and stable disease, n = 24) or non-responders (progressive disease, n = 27). Radiomic features (RF) were extracted from pre-TARE CT after segmentation of the liver tumor volume. A model was built based on a radiomic signature consisting of reliable RFs that allowed classification of response using multivariate logistic regression. Patients were assigned to high- or low-risk groups for disease progression after TARE according to a cutoff defined in the model. Kaplan-Meier analysis was performed to analyze survival between high- and low-risk groups. Two independent RF [Energy, Maximal Correlation Coefficient (MCC)], reflecting tumor heterogeneity, discriminated well between responders and non-responders. In particular, patients with higher magnitude of voxel values in an image (Energy), and texture complexity (MCC), were more likely to fail TARE. For predicting treatment response, the area under the receiver operating characteristic curve of the radiomics-based model was 0.75 (95% CI 0.48-1). The high-risk group had a shorter overall survival than the low-risk group (3.4 vs. 6.4months, p < 0.001). Our CT radiomics model may predict the response and survival outcome by quantifying tumor heterogeneity in patients treated with TARE for colorectal liver metastases.

Abstract TMP80: Hematoma Radiomic Markers of Survival After Supratentorial Intracerebral Hemorrhage on Admission Non-Contrast Head CT

Background: Radiomic features provide quantitative information about lesions morphology and texture on medical images. Recent studies have shown that intracerebral hemorrhage (ICH) radiomics on admission non-contrast head CT are associated with severity of symptoms at baseline and clinical outcomes - providing prognostic information beyond hematoma volume. In this study, we determined hematoma radiomic markers of post-ICH survival on admission head CT. Methods: Using the ATACH-2 trial dataset, we manually segmented ICH on admission head CTs and extracted 1130 radiomic features. Univariate and multivariate survival analyses were performed using the Cox Proportional Hazards model, Kaplan Meier Analysis, and logistic regression. Results: We split our dataset (n=871) to training (n=580) and testing (n=291) cohorts. Increased “original first order Energy” radiomic feature was associated with a higher probability of death (Hazard Ratio=1.64, p &lt;0.0001). LASSO Cox Proportional Hazards modeling also identified “original first order Energy” alongside age, NIHSS, and baseline INR as significant predictors of death. The ROC analysis confirmed the prognostic capability of this feature at 7 days, 30 days, and 90 days post-ICH in training and test cohorts. Kaplan Meier analysis demonstrated that increased “original first order Energy” is associated with a higher probability of death (p &lt; 0.05 at all time points). Logistic regression also showed that this feature is an independent predictor of post-ICH mortality (Odds ratios=2.04, p &lt; 0.001). Conclusions: The “original first order Energy” radiomic feature of hematoma on admission non-contrast head CT scans is a main predictor of survival in supratentorial ICH. This radiomic feature represents the magnitude of voxel values confounded with lesion volume - i.e. brighter larger hematomas on head CT have higher " original first order Energy ", and are associated with higher mortality rates after supratentorial ICH.

Reducing False-Positives in CT Perfusion Infarct Core Segmentation Using Contralateral Local Normalization.

The established global threshold of rCBF <30% for infarct core segmentation can lead to false-positives, as it does not account for the differences in blood flow between GM and WM and patient-individual factors, such as microangiopathy. To mitigate this problem, we suggest normalizing each voxel not only with a global reference value (ie, the median value of normally perfused tissue) but also with its local contralateral counterpart. We retrospectively enrolled 2830 CTP scans with suspected ischemic stroke, of which 335 showed obvious signs of microangiopathy. In addition to the conventional, global normalization, a local normalization was performed by dividing the rCBF maps with their mirrored and smoothed counterpart, which sets each voxel value in relation to the contralateral counterpart, intrinsically accounting for GM and WM differences and symmetric patient individual microangiopathy. Maps were visually assessed and core volumes were calculated for both methods. Cases with obvious microangiopathy showed a strong reduction in false-positives by using local normalization (mean 14.7 mL versus mean 3.7 mL in cases with and without microangiopathy). On average, core volumes were slightly smaller, indicating an improved segmentation that was more robust against naturally low blood flow values in the deep WM. The proposed method of local normalization can reduce overestimation of the infarct core, especially in the deep WM and in cases with obvious microangiopathy. False-positives in CTP infarct core segmentation might lead to less-than-optimal therapy decisions when not correctly interpreted. The proposed method might help mitigate this problem.

Medical image synthesis via conditional GANs: Application to segmenting brain tumours

Accurate brain tumour segmentation is critical for tasks such as surgical planning, diagnosis, and analysis, with magnetic resonance imaging (MRI) being the preferred modality due to its excellent visualisation of brain tissues. However, the wide intensity range of voxel values in MR scans often results in significant overlap between the density distributions of different tumour tissues, leading to reduced contrast and segmentation accuracy. This paper introduces a novel framework based on conditional generative adversarial networks (cGANs) aimed at enhancing the contrast of tumour subregions for both voxel-wise and region-wise segmentation approaches. We present two models: Enhancement and Segmentation GAN (ESGAN), which combines classifier loss with adversarial loss to predict central labels of input patches, and Enhancement GAN (EnhGAN), which generates high-contrast synthetic images with reduced inter-class overlap. These synthetic images are then fused with corresponding modalities to emphasise meaningful tissues while suppressing weaker ones. We also introduce a novel generator that adaptively calibrates voxel values within input patches, leveraging fully convolutional networks. Both models employ a multi-scale Markovian network as a GAN discriminator to capture local patch statistics and estimate the distribution of MR images in complex contexts. Experimental results on publicly available MR brain tumour datasets demonstrate the competitive accuracy of our models compared to current brain tumour segmentation techniques

Quantifying 3D MR fingerprinting (3D-MRF) reproducibility across subjects, sessions, and scanners automatically using MNI atlases.

Quantitative MRI techniques such as MR fingerprinting (MRF) promise more objective and comparable measurements of tissue properties at the point-of-care than weighted imaging. However, few direct cross-modal comparisons of MRF's repeatability and reproducibility versus weighted acquisitions have been performed. This work proposes a novel fully automated pipeline for quantitatively comparing cross-modal imaging performance in vivo via atlas-based sampling. We acquire whole-brain 3D-MRF, turbo spin echo, and MPRAGE sequences three times each on two scanners across 10 subjects, for a total of 60 multimodal datasets. The proposed automated registration and analysis pipeline uses linear and nonlinear registration to align all qualitative and quantitative DICOM stacks to Montreal Neurological Institute (MNI) 152 space, then samples each dataset's native space through transformation inversion to compare performance within atlas regions across subjects, scanners, and repetitions. Voxel values within MRF-derived maps were found to be more repeatable (σT1  = 1.90, σT2  = 3.20) across sessions than vendor-reconstructed MPRAGE (σT1w  = 6.04) or turbo spin echo (σT2w  = 5.66) images. Additionally, MRF was found to be more reproducible across scanners (σT1  = 2.21, σT2  = 3.89) than either qualitative modality (σT1w  = 7.84, σT2w  = 7.76). Notably, differences between repeatability and reproducibility of in vivo MRF were insignificant, unlike the weighted images. MRF data from many sessions and scanners can potentially be treated as a single dataset for harmonized analysis or longitudinal comparisons without the additional regularization steps needed for qualitative modalities.

Fuzzy Attention Neural Network to Tackle Discontinuity in Airway Segmentation.

Airway segmentation is crucial for the examination, diagnosis, and prognosis of lung diseases, while its manual delineation is unduly burdensome. To alleviate this time-consuming and potentially subjective manual procedure, researchers have proposed methods to automatically segment airways from computerized tomography (CT) images. However, some small-sized airway branches (e.g., bronchus and terminal bronchioles) significantly aggravate the difficulty of automatic segmentation by machine learning models. In particular, the variance of voxel values and the severe data imbalance in airway branches make the computational module prone to discontinuous and false-negative predictions, especially for cohorts with different lung diseases. The attention mechanism has shown the capacity to segment complex structures, while fuzzy logic can reduce the uncertainty in feature representations. Therefore, the integration of deep attention networks and fuzzy theory, given by the fuzzy attention layer, should be an escalated solution for better generalization and robustness. This article presents an efficient method for airway segmentation, comprising a novel fuzzy attention neural network (FANN) and a comprehensive loss function to enhance the spatial continuity of airway segmentation. The deep fuzzy set is formulated by a set of voxels in the feature map and a learnable Gaussian membership function. Different from the existing attention mechanism, the proposed channel-specific fuzzy attention addresses the issue of heterogeneous features in different channels. Furthermore, a novel evaluation metric is proposed to assess both the continuity and completeness of airway structures. The efficiency, generalization, and robustness of the proposed method have been proved by training on normal lung disease while testing on datasets of lung cancer, COVID-19, and pulmonary fibrosis.

Deep learning approach using SPECT-to-PET translation for attenuation correction in CT-less myocardial perfusion SPECT imaging

ObjectiveDeep learning approaches have attracted attention for improving the scoring accuracy in computed tomography-less single photon emission computed tomography (SPECT). In this study, we proposed a novel deep learning approach referring to positron emission tomography (PET). The aims of this study were to analyze the agreement of representative voxel values and perfusion scores of SPECT-to-PET translation model-generated SPECT (SPECTSPT) against PET in 17 segments according to the American Heart Association (AHA).MethodsThis retrospective study evaluated the patient-to-patient stress, resting SPECT, and PET datasets of 71 patients. The SPECTSPT generation model was trained (stress: 979 image pairs, rest: 987 image pairs) and validated (stress: 421 image pairs, rest: 425 image pairs) using 31 cases of SPECT and PET image pairs using an image-to-image translation network. Forty of 71 cases of left ventricular base-to-apex short-axis images were translated to SPECTSPT in the stress and resting state (stress: 1830 images, rest: 1856 images). Representative voxel values of SPECT and SPECTSPT in the 17 AHA segments against PET were compared. The stress, resting, and difference scores of 40 cases of SPECT and SPECTSPT were also compared in each of the 17 segments.ResultsFor AHA 17-segment-wise analysis, stressed SPECT but not SPECTSPT voxel values showed significant error from PET at basal anterior regions (segments #1, #6), and at mid inferoseptal regions (segments #8, #9, and #10). SPECT, but not SPECTSPT, voxel values at resting state showed significant error at basal anterior regions (segments #1, #2, and #6), and at mid inferior regions (segments #8, #9, and #11). Significant SPECT overscoring was observed against PET in basal-to-apical inferior regions (segments #4, #10, and #15) during stress. No significant overscoring was observed in SPECTSPT at stress, and only moderate over and underscoring in the basal inferior region (segment #4) was found in the resting and difference states.ConclusionsOur PET-supervised deep learning model is a new approach to correct well-known inferior wall attenuation in SPECT myocardial perfusion imaging. As standalone SPECT systems are used worldwide, the SPECTSPT generation model may be applied as a low-cost and practical clinical tool that provides powerful auxiliary information for the diagnosis of myocardial blood flow.

Auto-Classification of Parkinson's Disease with Different Motor Subtypes Using Arterial Spin Labelling MRI Based on Machine Learning.

The purpose of this study was to automatically classify different motor subtypes of Parkinson's disease (PD) on arterial spin labelling magnetic resonance imaging (ASL-MRI) data using support vector machine (SVM). This study included 38 subjects: 21 PD patients and 17 normal controls (NCs). Based on the Unified Parkinson's Disease Rating Scale (UPDRS) subscores, patients were divided into the tremor-dominant (TD) subtype and the postural instability gait difficulty (PIGD) subtype. The subjects were in a resting state during the acquisition of ASL-MRI data. The automated anatomical atlas 3 (AAL3) template was registered to obtain an ASL image of the same size and shape. We obtained the voxel values of 170 brain regions by considering the location coordinates of these regions and then normalized the data. The length of the feature vector depended on the number of voxel values in each brain region. Three binary classification models were utilized for classifying subjects' data, and we applied SVM to classify voxels in the brain regions. The left subgenual anterior cingulate cortex (ACC_sub_L) was clearly distinguished in both NCs and PD patients using SVM, and we obtained satisfactory diagnostic rates (accuracy = 92.31%, specificity = 96.97%, sensitivity = 84.21%, and AUCmax = 0.9585). For the right supramarginal gyrus (SupraMarginal_R), SVM distinguished the TD group from the other groups with satisfactory diagnostic rates (accuracy = 84.21%, sensitivity = 63.64%, specificity = 92.59%, and AUCmax = 0.9192). For the right intralaminar of thalamus (Thal_IL_R), SVM distinguished the PIGD group from the other groups with satisfactory diagnostic rates (accuracy = 89.47%, sensitivity = 70.00%, specificity = 6.43%, and AUCmax = 0.9464). These results are consistent with the changes in blood perfusion related to PD subtypes. In addition, the sensitive brain regions of the TD group and PIGD group involve the brain regions where the cerebellothalamocortical (CTC) and the striatal thalamocortical (STC) loops are located. Therefore, it is suggested that the blood perfusion patterns of the two loops may be different. These characteristic brain regions could become potential imaging markers of cerebral blood flow to distinguish TD from PIGD. Meanwhile, our findings provide an imaging basis for personalised treatment, thereby optimising clinical diagnostic and treatment approaches.

Habitat-based radiomics enhances the ability to predict lymphovascular space invasion in cervical cancer: a multi-center study.

Lymphovascular space invasion (LVSI) is associated with lymph node metastasis and poor prognosis in cervical cancer. In this study, we investigated the potential of radiomics, derived from magnetic resonance (MR) images using habitat analysis, as a non-invasive surrogate biomarker for predicting LVSI in cervical cancer. This retrospective study included 300 patients with cervical cancer who underwent surgical treatment at two centres (centre 1=198 and centre 2=102). Using the k-means clustering method, contrast-enhanced T1-weighted imaging (CE-T1WI) images were segmented based on voxel and entropy values, creating sub-regions within the volume ofinterest. Radiomics features were extracted from these sub-regions. Pearson correlation coefficient and least absolute shrinkage and selection operator LASSO) regression methods were used to select features associated with LVSI in cervical cancer. Support vector machine (SVM) model was developed based on the radiomics features extracted from each sub-region in the training cohort. The voxels and entropy values of the CE-T1WI images were clustered into three sub-regions. In the training cohort, the AUCs of the SVM models based on radiomics features derived from the whole tumour, habitat 1, habitat 2, and habitat 3 models were 0.805 (95% confidence interval [CI]: 0.745-0.864), 0.873(95% CI: 0.824-0.922), 0.869 (95% CI: 0.821-0.917), and 0.870 (95% CI: 0.821-0.920), respectively. Compared with whole tumour model, the predictive performances of habitat 3 model was the highest in the external test cohort (0.780 [95% CI: 0.692-0.869]). The radiomics model based on the tumour sub-regional habitat demonstrated superior predictive performance for an LVSI in cervical cancer than that of radiomics model derived from the whole tumour.

Predicting Prostate VMAT 3D Radiation Doses of Continuously Varying Organ Dose Trade-Offs Using a Conditional Variational Autoencoder

Predicting 3D radiation doses from planning structures is a promising method of knowledge-based treatment planning. However, most models are designed to predict only one 3D dose distribution per patient, based on historical organ dose trade-offs. To allow customizable plan generation, in this study, we aim to show the feasibility of dose prediction in which the degrees of organ dose trade-off could be explicitly specified. Specifically, the bladder vs. rectum dose trade-off in prostate cancer was investigated. In an IRB-approved study, we obtained imaging and structure contours for 167 patients with prostate cancer who received definitive radiotherapy. Training data was generated by automatically creating 3 different plans for each patient: while keeping target dose patterns constant, 1 base plan was generated with optimization objectives directly based on the output of a custom RapidPlan model prediction (S = 0), 1 plan with the goal to significantly lower bladder dose relative to the rectum (S = -1), and 1 plan with the goal to significantly lower rectum dose relative to the bladder (S = 1). This process was achieved by adjusting priority values during optimization. S is a scalar indicating the degree of bladder vs. rectum dose trade-off (higher S = higher dose to the bladder relative to the rectum). A conditional variational autoencoder (cVAE) was constructed as the generative model. Training, validation, and testing sets consist of 124, 10, and 33 patients, respectively. During training, the inputs to the model were 3D structure masks with voxel values modified based on S, and the output was the corresponding 3D dose. For model testing, we selected 7 equispaced values of S in the range [-1, +1] for each of the 33 test patients, generated the 3D doses for each S value (normalized to D2% = 110%), and calculated the differences of key dosimetric parameters (for S levels other than 0) compared to the predicted base plan (S = 0). The mean and standard deviations for these differences were reported. The cVAE model converged after training for 800 epochs. As the value of S increased from -1 to +1, the target coverage remained similar, while the doses to the bladder and rectum increased and decreased, respectively, as expected (Table 1). This pattern was also confirmed by qualitative examination of dose-volume histograms for additional S values. We demonstrated the feasibility of predicting 3D radiation dose distributions for prostate cancer where the degrees of organ dose trade-off could be explicitly specified.

Evaluation of Brain Atrophy after One Year of Whole Brain Irradiation Using Voxel Based Morphometry

Whole-brain irradiation has often been used for metastatic brain tumors and in prophylactic irradiation, but there are few reports on the frequency of cognitive decline and brain atrophy. We attempted to adapt an objective method of measuring brain atrophy to patients after whole-brain irradiation, Using Voxel based morphometry (VBM) measured from MRI to determine if brain atrophy has occurred one year after whole brain irradiation. VBM is a method to measure the change in signal intensity when an atrophic brain is transformed into a normal brain by Deformable Image Registration. We use VBM to measure brain atrophy before and one year after irradiation in patients who received whole brain irradiation at our institution. The MRI used for the analysis was a contrast-enhanced T1-weighted image, Sagittal cross-sectional images of 1 mm thickness were obtained from the whole brain using the 3D variable flip angle turbo spin echo sequence, which is clinically used at our hospital for the cancer patient's management. The subjects were 28 patients who had undergone whole-brain irradiation at our hospital and for whom MRIs were available before treatment and after one year. The cases analyzed included patients who received prophylactic Cranial irradiation (PCI) in cases of small cell lung cancer obtained complete remission. Z-scores obtained from the medial temporal lobe were compared before and after whole brain irradiation treatment to quantitatively assess brain atrophy. The Z-score is the normalized difference between the voxel values obtained from the normal group average and the individual case. There were 15 cases of lung cancer, 5 cases of malignant lymphoma, 5 cases of breast cancer, and 3 cases of various types of cancer, with an average age of 65 years. Cases with tumors or brain edema in the vicinity of the hippocampus were excluded. Various irradiation doses were used, but the commonly used radiation doses were 23.4 Gy/13Fr (lymphoma), 25 Gy/10 Fr (PCI), 35 Gy/14 Fr (multiple metastasis). A Paired-samples one sided t-test was performed to Mean z-score of positive values within the medial temporal lobe before and one year after whole brain irradiation. This number reflects the degree of atrophy within the region of interest and is determined as mild atrophy if it is greater than 1.0. The mean Z-score was found to increase at a risk of 0.01 after whole brain irradiation. This means that brain atrophy is progressing after whole brain irradiation. Hippocampal atrophy becomes apparent about a year after whole brain irradiation. It suggests that brain atrophy occurs after radiotherapy. The problem with this study is that the effects of chemotherapy-induced brain atrophy and aging cannot be ruled out.

Abnormal voxel-mirrored homotopic connectivity in first-episode major depressive disorder using fMRI: a machine learning approach.

To explore the interhemispheric information synergy ability of the brain in major depressive disorder (MDD) patients by applying the voxel-mirrored homotopic connectivity (VMHC) method and further explore the potential clinical diagnostic value of VMHC metric by a machine learning approach. 52 healthy controls and 48 first-episode MDD patients were recruited in the study. We performed neuropsychological tests and resting-state fMRI scanning on all subjects. The VMHC values of the symmetrical interhemispheric voxels in the whole brain were calculated. The VMHC alterations were compared between two groups, and the relationship between VMHC values and clinical variables was analyzed. Then, abnormal brain regions were selected as features to conduct the classification model by using the support vector machine (SVM) approach. Compared to the healthy controls, MDD patients exhibited decreased VMHC values in the bilateral middle frontal gyrus, fusiform gyrus, medial superior frontal gyrus and precentral gyrus. Furthermore, the VMHC value of the bilateral fusiform gyrus was positively correlated with the total Hamilton Depression Scale (HAMD). Moreover, SVM analysis displayed that a combination of all clusters demonstrated the highest area under the curve (AUC) of 0.87 with accuracy, sensitivity, and specificity values of 86.17%, 76.74%, and 94.12%, respectively. MDD patients had reduced functional connectivity in the bilateral middle frontal gyrus, fusiform gyrus, medial superior frontal gyrus and precentral gyrus, which may be related to depressive symptoms. The abnormality in these brain regions could represent potential imaging markers to distinguish MDD patients from healthy controls.

Effect of metal implants and metal artifacts on back-projected two-dimensional entrance fluence determined by EPID dosimetry.

To evaluate the errors caused by metal implants and metal artifacts in the two-dimensional entrance fluences reconstructed using the back-projection algorithm based on electronic portal imaging device (EPID) images. The EPID in the Varian VitalBeam accelerator was used to acquire portal dose images (PDIs), and then commercial EPID dosimetry software was employed to reconstruct the two-dimensional entrance fluences based on computed tomography (CT) images of the head phantoms containing interchangeable metal-free/titanium/aluminum round bars. The metal-induced errors in the two-dimensional entrance fluences were evaluated by comparing the γ results and the pixel value errors in the metal-affected regions. We obtained metal-artifact-free CT images by replacing the voxel values of non-metal inserts with those of metal inserts in metal-free CT images to evaluate the metal-artifact-induced errors. The γ passing rates (versus PDIs obtained without a phantom in the beam field (PDIair ), 2%/2mm) for the back-projected two-dimensional entrance fluences of phantoms containing titanium or aluminum (BPTi /BPAl ) were reduced from 92.4% to 90.5% and 90.6%, respectively, relative to the metal-free phantom (BPmetal-free ). Titanium causes more severe metal artifacts in CT images than aluminum, and its removal resulted in a 0.0022 CU (median) reduction in the pixel value of BPTi artifact-free relative to BPTi in the metal-affected region. Moreover, the mean absolute error (MAE) and root mean square error (RMSE) decreased from 0.0050 CU and 0.0063 CU to 0.0034 CU and 0.0040 CU, respectively (vs. BPmetal-free ). Metal implants increase the errors in back-projected two-dimensional entrance fluences, and metals with higher electron densities cause more errors. For high-electron-density metal implants that produce severe metal artifacts (e.g., titanium), removing metal artifacts from the CT images can improve the accuracy of the two-dimensional entrance fluences reconstructed by back-projection algorithms.

Tumor dosimetry using 177Lu: influence of background activity, measurement method and reconstruction algorithm

BackgroundImage-based tumor dosimetry after radionuclide therapy, using the isotope 177Lu, finds application e.g., for tumor-to-organ dose comparison and for dose response evaluation. When the tumor extent is not much larger than the image resolution, and when 177Lu is found in nearby organs or other tumors, an accurate determination of tumor dose is particularly challenging. Here a quantitative evaluation of three different methods for determining the 177Lu activity concentration in a phantom is performed, and the dependence on a variety of parameters is described. The phantom (NEMA IEC body phantom) has spheres of different size in a background volume, and sphere-to-background 177Lu activity concentration ratios of infinity, 9.5, 5.0 and 2.7 are applied. The methods are simple to implement and well-known from the literature. They are based on (1) a large VOI encompassing the whole sphere, without background activity and with volume information from other sources, (2) a small VOI located in the sphere center, and (3) a VOI consisting of voxels with voxel value above a certain percentage of the maximum voxel value.ResultsThe determined activity concentration varies significantly with sphere size, sphere-to-background ratio, SPECT reconstruction method and method for determining the concentration. Based on the phantom study, criteria are identified under which the activity concentration can be determined with a maximal error of 40% even in the presence of background activity.ConclusionsTumor dosimetry is feasible in the presence of background activity using the above-mentioned methods, provided appropriate SPECT reconstructions are applied and tumors are selected for dosimetry analysis according to the following criteria for the three methods: (1) solitary tumor with diameter > 15 mm, (2) tumor diameter > 30 mm and tumor-to-background ratio > 2, and (3) tumor diameter > 30 mm and tumor-to-background ratio > 3.

Multi-head attention-based masked sequence model for mapping functional brain networks.

The investigation of functional brain networks (FBNs) using task-based functional magnetic resonance imaging (tfMRI) has gained significant attention in the field of neuroimaging. Despite the availability of several methods for constructing FBNs, including traditional methods like GLM and deep learning methods such as spatiotemporal self-attention mechanism (STAAE), these methods have design and training limitations. Specifically, they do not consider the intrinsic characteristics of fMRI data, such as the possibility that the same signal value at different time points could represent different brain states and meanings. Furthermore, they overlook prior knowledge, such as task designs, during training. This study aims to overcome these limitations and develop a more efficient model by drawing inspiration from techniques in the field of natural language processing (NLP). The proposed model, called the Multi-head Attention-based Masked Sequence Model (MAMSM), uses a multi-headed attention mechanism and mask training approach to learn different states corresponding to the same voxel values. Additionally, it combines cosine similarity and task design curves to construct a novel loss function. The MAMSM was applied to seven task state datasets from the Human Connectome Project (HCP) tfMRI dataset. Experimental results showed that the features acquired by the MAMSM model exhibit a Pearson correlation coefficient with the task design curves above 0.95 on average. Moreover, the model can extract more meaningful networks beyond the known task-related brain networks. The experimental results demonstrated that MAMSM has great potential in advancing the understanding of functional brain networks.

Improving synthetic CT accuracy by combining the benefits of multiple normalized preprocesses.

To investigate the effect of different normalization preprocesses in deep learning on the accuracy of different tissues in synthetic computed tomography (sCT) and to combine their advantages to improve the accuracy of all tissues. The cycle-consistent adversarial network (CycleGAN) model was used to generate sCT images from megavolt cone-beam CT (MVCBCT) images. In this study, 2639 head MVCBCT and CT image pairs from 203 patients were collected as a training set, and 249 image pairs from 29 patients were collected as a test set. We normalized the voxel values in images to 0 to 1 or -1 to 1, using two linear and five nonlinear normalization preprocessing methods to obtain seven data sets and compared the accuracy of different tissues in different sCT obtained from training these data. Finally, to combine the advantages of different normalization preprocessing methods, we obtained sCT_Blur by cropping, stitching, and smoothing (OpenCV's cv2.medianBlur, kernel size 5) each group of sCTs and evaluated its image quality and accuracy of OARs. Different normalization preprocesses made sCT more accurate in different tissues. The proposed sCT_Blur took advantage of multiple normalization preprocessing methods, and all tissues are more accurate than the sCT obtained using a single conventional normalization method. Compared with other sCT images, the structural similarity of sCT_Blur versus CT was improved to 0.906±0.019. The mean absolute errors of the CT numbers were reduced to 15.7±4.1 HU, 23.2±7.1 HU, 11.5±4.1 HU, 212.8±104.6 HU, 219.4±35.1 HU, and 268.8±88.8 HU for the oral cavity, parotid, spinal cord, cavity, mandible, and teeth, respectively. The proposed approach combined the advantages of several normalization preprocessing methods to improve the accuracy of all tissues in sCT images, which is promising for improving the accuracy of dose calculations based on CBCT images in adaptive radiotherapy.

68Ga-Prostate-Specific Membrane Antigen PET Radiomics For the Prediction of PostSurgical International Society of Urological Pathology Grade in Patients with Primary Prostate Cancer

INTRODUCTION Radiomics has been proven effective for the characterisation of primary prostate cancer (PCa).1,2 However, the limited interpretability of the proposed models represents one of the major limitations in this field.3,4 This study investigated 68Ga-prostate-specific membrane antigen (PSMA) PET radiomics for the prediction of post-surgical International Society of Urological Pathology (ISUP) grade in patients with primary PCa, ensuring model interpretability. MATERIALS AND METHODS Forty-seven patients with PCa were examined with 68Ga-PSMA PET at the authors’ institution. Those patients were enrolled in this study prior to radical prostatectomy. Images were acquired using either PET/MRI or PET/CT. ISUP grade was available at both biopsy and radical prostatectomy for all patients. A radiologist manually segmented the whole prostate on PET images using the co-registered CT or MRI for anatomical localisation on 3D Slicer software (Brigham and Women’s Hospital, Boston, Massachusetts, USA).5 The whole prostate was used as volume of interest (VOI) to avoid the limitations of radiomics for small volumes.6 VOIs were normalised, resampled, and discretised. A total of 103 image biomarker standardisation initiative-compliant, radiomic features (RF) were extracted using PyRadiomics (Python Software Foundation, Beaverton, Oregon, USA).7 RFs were harmonised with the ComBat method8 to control for the scanner effect, and selected using the minimum redundancy maximum relevance algorithm. Combinations of the four most relevant RFs were used to train 12 radiomics machine learning models for the prediction of post-surgical ISUP ≥4 versus ISUP &lt;4 that were validated by five-fold repeated stratified cross-validation. To ensure that results were not driven by spurious associations, two ad hoc control models were generated. The first one Creative Commons Attribution-Non Commercial 4.0 ● April 2023 ● Urology 37 EAU 2023 • Abstract had SUVmax and VOI volume as input (radiomics baseline), while the other was made by setting to zero all voxel values prior features extraction (PET zeros). Balanced accuracy, sensitivity, specificity, and positive and negative predictive values were collected. The performance of the best developed model was compared with that of ISUP grade biopsy. RESULTS ISUP grade at biopsy was upgraded in 9 out of 47 patients after prostatectomy, resulting in a balanced accuracy of 85.9%; sensitivity of 71.9%; specificity of 100.0%; positive predicted value of 100.0%; and negative predictive value of 62.5%. The best performing radiomic model yielded a balanced accuracy of 87.6%; sensitivity of 88.6%; specificity of 86.7%; positive predicted value of 94.0%; and negative predicted value of 82.5%. All radiomic models trained with at least two RFs (grey level size zone matrix; zone entropy and shape; least axis length) outperformed the control models. Conversely, no significant differences were found for radiomic models trained with two or more RFs (Mann–Whitney U test; p&gt;0.05). See Table 1 for a detailed report of all the generated models’ performance. CONCLUSION These findings support the role of 68Ga-PSMA PET radiomics for the accurate and non-invasive prediction of post-surgical ISUP grade. Future multicentre studies will be needed to establish with certainty the accuracy and reproducibility of the radiomic signature proposed here.

Introducing the THz time domain CT system for evaluating kernel weight and plumpness of sunflower seed

The Terahertz (THz) time domain computed tomography (CT) system was used for evaluating kernel weight and plumpness of sunflower seed in present study. Specifically, a series of THz projections of the sunflower seed on a rotation stage were captured by a customized transmission mode THz time domain imaging system from different angles. These projections were preprocessed and then modelled by the inverse Radon transform to reconstruct the three-dimensional (3D) sunflower seed. The shapes and structures of sunflower seeds in reconstructed volumetric images were well presented. The volume ratio (obtained by THz CT imaging) and the area ratio (obtained by RGB imaging) of the whole seed to the kernel were calculated as 3D and 2D plumpness, respectively. The comparison between these two different plumpness indicating that our 3D plumpness can better describe the sunflower seed structure. Correlation coefficient (R) between sum intensity values of kernel voxels and kernel weights is 0.89 at 1.2 THz. This is the first application of applying THz CT systems to agri-food research, and the results indicate that THz time domain CT can be a useful nondestructive tool for evaluating weight and plumpness of intact sunflower seed. In addition, there is potential to extend the THz time domain CT system to the analysis of other agri-food sample types.