A 2-step prediction model for diagnosis of germinomas in the pineal region.

Background: Germinomas are sensitive to radiation and chemotherapy, and their management distinctly differs from other kinds of pineal region tumors. The aim of this study was to construct a prediction model based on clinical features and preoperative magnetic resonance (MR) manifestations to achieve noninvasive diagnosis of germinomas in pineal region. Methods: A total of 126 patients with pineal region tumors were enrolled, including 36 germinomas, 53 nongerminomatous germ cell tumors (NGGCTs), and 37 pineal parenchymal tumors (PPTs). They were divided into a training cohort (n = 90) and a validation cohort (n = 36). Features were extracted from clinical records and conventional MR images. Multivariate analysis was performed to screen for independent predictors to differentiate germ cell tumors (GCTs) and PPTs, germinomas, and NGGCTs, respectively. From this, a 2-step nomogram model was established, with model 1 for discriminating GCTs from PPTs and model 2 for identifying germinomas in GCTs. The model was tested in a validation cohort. Results: Both model 1 and model 2 yielded good predictive efficacy, with c-indexes of 0.967 and 0.896 for the diagnosis of GCT and germinoma, respectively. Calibration curve, decision curve, and clinical impact curve analysis further confirmed their predictive accuracy and clinical usefulness. The validation cohort achieved areas under the receiver operating curves of 0.885 and 0.926, respectively. Conclusions: The 2-step model in this study can noninvasively differentiate GCTs from PPTs and further identify germinomas, thus holding potential to facilitate treatment decision-making for pineal region tumors.

Deep Learning-Assisted Quantitative Susceptibility Mapping as a Tool for Grading and Molecular Subtyping of Gliomas.

This study aimed to explore the value of deep learning (DL)-assisted quantitative susceptibility mapping (QSM) in glioma grading and molecular subtyping. Forty-two patients with gliomas, who underwent preoperative T2 fluid-attenuated inversion recovery (T2 FLAIR), contrast-enhanced T1-weighted imaging (T1WI + C), and QSM scanning at 3.0T magnetic resonance imaging (MRI) were included in this study. Histopathology and immunohistochemistry staining were used to determine glioma grades, and isocitrate dehydrogenase (IDH) 1 and alpha thalassemia/mental retardation syndrome X-linked gene (ATRX) subtypes. Tumor segmentation was performed manually using Insight Toolkit-SNAP program (www.itksnap.org). An inception convolutional neural network (CNN) with a subsequent linear layer was employed as the training encoder to capture multi-scale features from MRI slices. Fivefold cross-validation was utilized as the training strategy (seven samples for each fold), and the ratio of sample size of the training, validation, and test dataset was 4:1:1. The performance was evaluated by the accuracy and area under the curve (AUC). With the inception CNN, single modal of QSM showed better performance in differentiating glioblastomas (GBM) and other grade gliomas (OGG, grade II-III), and predicting IDH1 mutation and ATRX loss (accuracy: 0.80, 0.77, 0.60) than either T2 FLAIR (0.69, 0.57, 0.54) or T1WI + C (0.74, 0.57, 0.46). When combining three modalities, compared with any single modality, the best AUC/accuracy/F1-scores were reached in grading gliomas (OGG and GBM: 0.91/0.89/0.87, low-grade and high-grade gliomas: 0.83/0.86/0.81), predicting IDH1 mutation (0.88/0.89/0.85), and predicting ATRX loss (0.78/0.71/0.67). As a supplement to conventional MRI, DL-assisted QSM is a promising molecular imaging method to evaluate glioma grades, IDH1 mutation, and ATRX loss. Supplementary Information: The online version contains supplementary material available at 10.1007/s43657-022-00087-6.

Iterative dynamic dual-energy CT algorithm in reducing statistical noise in multi-energy CT imaging.

Multi-energy spectral CT has a broader range of applications with the recent development of photon-counting detectors. However, the photons counted in each energy bin decrease when the number of energy bins increases, which causes a higher statistical noise level of the CT image. In this work, we propose a novel iterative dynamic dual-energy CT algorithm to reduce the statistical noise. In the proposed algorithm, the multi-energy projections are estimated from the dynamic dual-energy CT data during the iterative process. The proposed algorithm is verified on sufficient numerical simulations and a laboratory two-energy-threshold PCD system. By applying the same reconstruction algorithm, the dynamic dual-energy CT's final reconstruction results have a much lower statistical noise level than the conventional multi-energy CT. Moreover, based on the analysis of the simulation results, we explain why the dynamic dual-energy CT has a lower statistical noise level than the conventional multi-energy CT. The underlying idea is to sample sparse in the energy dimension, which can be done because there is a high correlation between projection data of different energy bins.

Dynamic-dual-energy spectral CT for improving multi-material decomposition in image-domain.

Dual-energy CT, as well as spectral CT, has a great potential in material decomposition. However, dual-energy CT is difficult to apply to multi-material decomposition because the number of energy bins is limited to two. Current spectral CT systems have more energy bins, but the statistical noise in each energy bin is high because of the decreased photon number, which causes errors in the material decomposition results. In this paper, we propose a dynamic-dual-energy spectral CT for accurate multi-material decomposition. In the course of scanning, the energy threshold of the dynamic-dual-energy detector randomly changes to obtain the spectral information of photons. With the proposed statistical noise-weighted tPRISM algorithm, the multi-energy image reconstruction using dynamic-dual-energy CT data was implemented, followed by multi-material decomposition. Both simulation and experiment results show that the multi-energy reconstruction and multi-material decomposition using the dynamic-dual-energy method are more accurate and have less noise compared with that of the conventional static-multi-energy method with the same number of energy bins. The ring artifacts which are severe in the experimental data simulation and experiment results using the conventional spectral CT method are reduced in great extent when using our proposed method. In conclusion, our proposed dynamic-dual-energy spectral CT method is highly feasible and has a great potential in high-quality multi-material decomposition.