Development and validation of a clinical prediction model for prognostic factors in patients with primary pontine hemorrhage.

We aimed to develop a prognostic model for primary pontine hemorrhage (PPH) patients and validate the predictive value of the model for a good prognosis at 90 days. A total of 254 PPH patients were included for screening of the independent predictors of prognosis, and data were analyzed by univariate and multivariable logistic regression tests. The cases were then divided into training cohort (n=219) and validation cohort (n=35) based on the two centers. A nomogram was developed using independent predictors from the training cohort to predict the 90-day good outcome and was validated from the validation cohort. Glasgow Coma Scale score, normalized pixels (used to describe bleeding volume), and mechanical ventilation were significant predictors of a good outcome of PPH at 90 days in the training cohort (all P<0.05). The U test showed no statistical difference (P=0.892) between the training cohort and the validation cohort, suggesting the model fitted well. The new model showed good discrimination (area under the curve=0.833). The decision curve analysis of the nomogram of the training cohort indicated a great net benefit. The PPH nomogram comprising the Glasgow Coma Scale score, normalized pixels, and mechanical ventilation may facilitate predicting a 90-day good outcome.

A comparison study of dynamic [18F]Alfatide II imaging and [11C]MET in orthotopic rat models of glioblastoma.

PURPOSE: To investigate and compare the dynamic positron emission tomography (PET) imaging with [18F]Alfatide II Imaging and [11C]Methionine ([11C]MET) in orthotopic rat models of glioblastoma multiforme (GBM), and to assess the utility of [18F]Alfatide II in detecting and evaluating neoangiogenesis in GBM. METHODS: [18F]Alfatide II and [11C]MET were injected into the orthotopic GBM rat models (n = 20, C6 glioma cells), followed by dynamic PET/MR scans 21 days after surgery of tumor implantation. On the PET image with both radiotracers, the MRI-based volume-of-interest (VOI) was manually delineated encompassing glioblastoma. Time-activity curves were expressed as tumor-to-normal brain ratio (TNR) parameters and PET pharmacokinetic modeling (PKM) performed using 2-tissue-compartment models (2TCM). Immunofluorescent staining (IFS), western blotting and blocking experiment of tumor tissue were performed for the validation. RESULTS: Compared to 11C-MET, [18F]Alfatide II presented a persistent accumulation in the tumor, albeit with a slightly lower SUVmean of 0.79 ± 0.25, and a reduced uptake in the contralateral normal brain tissue, respectively. This resulted in a markedly higher tumor-to-normal brain ratio (TNR) of 18.22 ± 1.91. The time-activity curve (TACs) showed a significant increase in radioactive uptake in tumor tissue, followed by a plateau phase up to 60 min for [18F]Alfatide II (time to peak:255 s) and 40 min for [11C]MET (time to peak:135 s) post injection. PKM confirmed significantly higher K1 (0.23/0.07) and K3 (0.26/0.09) in the tumor region compared to the normal brain with [18F]Alfatide II. Compared to [11C]MET imaging, PKM confirmed both significantly higher K1/K2 (1.24 ± 0.79/1.05 ± 0.39) and K3/K4 (11.93 ± 4.28/3.89 ± 1.29) in the tumor region with [18F]Alfatide II. IFS confirmed significant expression of integrin and tumor vascularization in tumor region. CONCLUSION: [18F]Alfatide II demonstrates potential in imaging tumor-associated neovascularization in the context of glioblastoma multiforme (GBM), suggesting its utility as a tool for further exploration in neovascular characterization.

Development and validation of a clinical prediction model for prognostic factors in patients with primary pontine hemorrhage

Abstract We aimed to develop a prognostic model for primary pontine hemorrhage (PPH) patients and validate the predictive value of the model for a good prognosis at 90 days. A total of 254 PPH patients were included for screening of the independent predictors of prognosis, and data were analyzed by univariate and multivariable logistic regression tests. The cases were then divided into training cohort (n=219) and validation cohort (n=35) based on the two centers. A nomogram was developed using independent predictors from the training cohort to predict the 90-day good outcome and was validated from the validation cohort. Glasgow Coma Scale score, normalized pixels (used to describe bleeding volume), and mechanical ventilation were significant predictors of a good outcome of PPH at 90 days in the training cohort (all P<0.05). The U test showed no statistical difference (P=0.892) between the training cohort and the validation cohort, suggesting the model fitted well. The new model showed good discrimination (area under the curve=0.833). The decision curve analysis of the nomogram of the training cohort indicated a great net benefit. The PPH nomogram comprising the Glasgow Coma Scale score, normalized pixels, and mechanical ventilation may facilitate predicting a 90-day good outcome.

Predicting prognosis of primary pontine hemorrhage using CT image and deep learning.

Prognosis of primary pontine hemorrhage (PPH) is important for treatment planning and patient management. However, only few clinical factors were reported to have prognostic value to PPH. Here, we propose a deep learning (DL) model that mines high-dimensional prognostic information from computed tomography (CT) images and combines clinical factors for predicting individualized prognosis of PPH. We proposed a multi-task DL model to learn high-dimensional CT features of hematoma and perihematomal areas for predicting the risk of 30-day mortality, 90-day mortality and 90-day functional outcome of PPH simultaneously. We further explored the combination of the DL model and clinical factors by building a combined model. All the models were trained in a training cohort (n = 219) and tested in an independent testing cohort (n = 35). The DL model achieved area under the curve (AUC) of 0.886, 0.886, and 0.759 in predicting 30-day mortality, 90-day mortality and 90-day functional outcome of PPH in the independent testing cohort, which improved over the previously reported new PPH score and the clinical model. When combining the DL model and clinical factors, the combined model achieved improved performance (AUC = 0.920, 0.941, and 0.894), indicating that DL model mines CT information that complements clinical factors. Through DL visualization technique, we found that the internal structure of hematoma and its expansion to perihematomal regions are important for predicting the prognosis of PPH. This DL model provides an easy-to-use way for predicting individualized prognosis of PPH by mining high-dimensional information from CT images, and showed improvement over clinical factors and present methods.

Deep learning-based image reconstruction for TOF PET with DIRECT data partitioning format.

Conventional positron emission tomography (PET) image reconstruction is achieved by the statistical iterative method. Deep learning provides another opportunity for speeding up the image reconstruction process. However, conventional deep learning-based image reconstruction requires a fully connected network for learning the Radon transform. The use of fully connected networks greatly complicated the network and increased hardware cost. In this study, we proposed a novel deep learning-based image reconstruction method by utilizing the DIRECT data partitioning method. The U-net structure with only convolutional layers was used in our approach. Patch-based model training and testing were used to achieve 3D reconstructions within current hardware limitations. Time-of-flight (TOF)-histoimages were first generated from the listmode data to replace conventional sinograms. Different projection angles were used as different channels in the input. A total of 15 patient data were used in this study. For each patient, the dynamic whole-body scanning protocol was used to expand the training dataset and a total of 372 separate scans were included. The leave-one-patient-out validation method was used. Two separate studies were carried out. In the first study, the measured TOF-histoimages were directly used for model training and testing, to study the performance of the method in real-world applications. In the second study, TOF-histoimages were simulated from already reconstructed images to exclude the scatters, randoms, attenuation-activity mismatch effects. This study was used to evaluate the optimal performance when all other corrections are ideal. Volumes of interests were placed in the liver and lesion region to study image noise and lesion quantitations. The reconstructed images using the proposed deep learning method showed similar image quality when compared with the conventional expectation-maximization approach. A minimal difference was observed when the simulated TOF-histoimages were used as model input and testing, suggesting the deep learning model can indeed learn the reconstruction process. Some quantitative difference was observed when the measured TOF-histoimages were used. The two studies suggested that the major difference is caused by inaccurate corrections performed by the network itself, which indicated that physics-based corrections are still required for better quantitative performance. In conclusion, we have proposed a novel deep learning-based image reconstruction method for TOF PET. With the help of the DIRECT data partitioning method, no fully connected layers were used and 3D image reconstruction can be directly achieved within the limits of the current hardware.

Simplified protocol for whole-body Patlak parametric imaging with 18 F-FDG PET/CT: Feasibility and error analysis.

PURPOSE: Parametric imaging using the Patlak model has been shown to provide improved lesion detectability and specificity. The Patlak model requires both tissue time-activity curves (TACs) after equilibrium and knowledge of the input function from the start of injection. Therefore, the conventional dynamic scanning protocol typically starts from the radiotracer injection all the way to equilibrium. In this paper, we propose the use of hybrid population-based and model-based input function estimation and evaluate its use for whole-body Patlak analysis, in order to reduce the total scan time and simplify clinical Patlak parametric imaging protocols. Possible quantitative errors caused by the simplified scanning protocol were also analyzed both theoretically and with the use of clinical data. MATERIALS AND METHODS: Clinical data from 24 patients referred for tumor staging were included in this study. The patients underwent a whole-body dynamic PET study, 20 min after FDG injection (0.13 mCi/kg). The proposed whole-body scanning protocol includes 6 passes with 4-5 bed positions, depending on the size of the patient, with 2 min for each bed position. An input function from the literature was selected as the shape of the population-based input function. The descending aorta from the corresponding CT image was segmented and applied on the reconstructed dynamic PET images to acquire an image-based input function, which was later fitted using an exponential model. Due to the late scan time, only the later portion of the input function was available, which was used to scale the population-based input function. The hybrid input function was used to derive the whole-body Patlak images. Assuming a given error in the population-based input function, its influence on the final Patlak images were also derived theoretically and verified using the clinical data sets. Finally, the image quality of the reconstructed Patlak slope image was evaluated by an experienced radiologist in four different aspects: image artifacts, image noise, lesion sharpness, and lesion detectability. RESULTS: It was found that errors in the population-based input function only affect the absolute scale of the Patlak slope image. The induced error is proportional to the percentage area-under-curve (AUC) error in the input function. These findings were also confirmed by numerical analysis. The predicted global scale was in good agreement with results from both image-based Patlak and direct Patlak approach. The fractions of the AUC from the early portion population-based input function were also found to be around 18% of the total AUC of the input function, further limiting the propagation of quantitation error from population-based input function to the final Patlak slope image. The reconstructed Patlak images were also found by the radiologist to provide excellent confidence in lesion detection tasks. CONCLUSIONS: We have proposed a simplified whole-body scanning protocol that utilizes both population-based input function and model-based input function. The error from the population-based function was found to only affect the global scale and the overall quantitative impact can be predicted using our proposed formulas.