3D CT-Inclusive Deep-Learning Model to Predict Mortality, ICU Admittance, and Intubation in COVID-19 Patients.

Chest CT is a useful initial exam in patients with coronavirus disease 2019 (COVID-19) for assessing lung damage. AI-powered predictive models could be useful to better allocate resources in the midst of the pandemic. Our aim was to build a deep-learning (DL) model for COVID-19 outcome prediction inclusive of 3D chest CT images acquired at hospital admission. This retrospective multicentric study included 1051 patients (mean age 69, SD = 15) who presented to the emergency department of three different institutions between 20th March 2020 and 20th January 2021 with COVID-19 confirmed by real-time reverse transcriptase polymerase chain reaction (RT-PCR). Chest CT at hospital admission were evaluated by a 3D residual neural network algorithm. Training, internal validation, and external validation groups included 608, 153, and 290 patients, respectively. Images, clinical, and laboratory data were fed into different customizations of a dense neural network to choose the best performing architecture for the prediction of mortality, intubation, and intensive care unit (ICU) admission. The AI model tested on CT and clinical features displayed accuracy, sensitivity, specificity, and ROC-AUC, respectively, of 91.7%, 90.5%, 92.4%, and 95% for the prediction of patient's mortality; 91.3%, 91.5%, 89.8%, and 95% for intubation; and 89.6%, 90.2%, 86.5%, and 94% for ICU admission (internal validation) in the testing cohort. The performance was lower in the validation cohort for mortality (71.7%, 55.6%, 74.8%, 72%), intubation (72.6%, 74.7%, 45.7%, 64%), and ICU admission (74.7%, 77%, 46%, 70%) prediction. The addition of the available laboratory data led to an increase in sensitivity for patient's mortality (66%) and specificity for intubation and ICU admission (50%, 52%, respectively), while the other metrics maintained similar performance results. We present a deep-learning model to predict mortality, ICU admittance, and intubation in COVID-19 patients. KEY POINTS: • 3D CT-based deep learning model predicted the internal validation set with high accuracy, sensibility and specificity (> 90%) mortality, ICU admittance, and intubation in COVID-19 patients. • The model slightly increased prediction results when laboratory data were added to the analysis, despite data imbalance. However, the model accuracy dropped when CT images were not considered in the analysis, implying an important role of CT in predicting outcomes.

Diffusion kurtosis imaging in head and neck cancer: A correlation study with dynamic contrast enhanced MRI.

PURPOSE: To investigate the biophysical meaning of Diffusion Kurtosis Imaging (DKI) parameters via correlations with the perfusion parameters obtained from a long Dynamic Contrast Enhanced MRI scan, in head and neck (HN) cancer. METHODS: Twenty two patients with newly diagnosed HN tumor were included in the present retrospective study. Some patients had multiple lesions, therefore a total of 26 lesions were analyzed. DKI was acquired using 5b values at 0, 500, 1000,1500 and 2000 s/mm2. DCE-MRI was obtained with 130 dynamic volumes, with a temporal resolution of 5 s, to achieve a long scan time (>10 min). The apparent diffusion coefficient Dapp and apparent diffusional kurtosis Kapp were calculated voxel-by-voxel, removing the point at b value = 0 to eliminate possible perfusion effects on the parameter estimations. The transfer constants Ktrans and Kep, ve, and the histogram-based entropy (En) and interquartile range (IQR) of each DCE-MRI parameter were quantified. Correlations between all variables were investigated by the Spearman's Rho correlation test. RESULTS: Moderate relationships emerged between Dapp and Kep (Rho =  - 0.510, p = 0.009), and between Dapp and ve (Rho = 0.418, p = 0.038). En(Kep) was significantly related to Kapp (Rho = 0.407, p = 0.043), while IQR(Kep) showed an inverse association with Dapp (Rho = -0.422, p = 0.035). CONCLUSIONS: A weak to intermediate correlation was found between DKI parameters and both Kep and ve. The kurtosis was associated to the intratumoral heterogeneity and complexity of the capillary permeability, expressed by En(Kep).

Prediction of survival in patients affected by glioblastoma: histogram analysis of perfusion MRI.

PURPOSE: The identification of prognostic biomarkers plays a pivotal role in the management of glioblastoma. The aim of this study was to assess the role of magnetic resonance dynamic susceptibility contrast imaging (DSC-MRI) with histogram analysis in the prognostic evaluation of patients suffering from glioblastoma. MATERIALS AND METHODS: Sixty-eight patients with newly diagnosed pathologically verified GBM were retrospectively evaluated. All patients underwent MRI investigations, including DSC-MRI, surgical procedure and received postoperative focal radiotherapy plus daily temozolomide (TMZ), followed by adjuvant TMZ therapy. Relative cerebral blood volume (rCBV) histograms were generated from a volume of interest covering the solid portions of the tumor and statistically evaluated for kurtosis, skewness, mean, median and maximum value of rCBV. To verify if histogram parameters could predict survival at 1 and 2 years, receiver operating characteristic (ROC) curves were obtained. Kaplan-Meier method was used to calculate patient's overall survival. RESULTS: rCBV kurtosis and rCBV skewness showed significant differences between subjects surviving > 1 and > 2 years, According to ROC analysis, the rCBV kurtosis showed the best statistic performance compared to the other parameters; respectively, values of 1 and 2.45 represented an optimised cut-off point to distinguish subjects surviving over 1 or 2 years. Kaplan-Meier curves showed a significant difference between subjects with rCBV kurtosis values higher or lower than 1 (respectively 1021 and 576 days; Log-rank test: p = 0.007), and between subjects with rCBV kurtosis values higher or lower than 2.45 (respectively 802 and 408 days; Log-rank test: p = 0.001). CONCLUSION: The histogram analysis of perfusion MRI proved to be a valid method to predict survival in patients affected by glioblastoma.

Pituitary apoplexy: an update on clinical and imaging features.

Pituitary apoplexy (PA) is a rare and potentially fatal clinical condition presenting acute headache, vomiting, visual impairment, ophthalmoplegia, altered mental state and possible panhypopituitarism. It mostly occurs in patients with haemorrhagic infarction of the pituitary gland due to a pre-existing macroadenoma. Although there are pathological and physiological conditions that may share similar imaging characteristics, both clinical and imaging features can guide the radiologist towards the correct diagnosis, especially using magnetic resonance imaging (MRI). In this review, we will describe the main clinical and epidemiological features of PA, illustrating CT and MRI findings and discussing the role of imaging in the differential diagnosis, prognosis and follow-up. Teaching points • Headache, ophtalmoplegia and visual impairment are frequent symptoms of pituitary apoplexy. • CT is often the first imaging tool in PA, showing areas of hyperdensity within the sellar region. • MRI could confirm haemorrhage within the pituitary gland and compression on the optic chiasm. • Frequent simulating conditions are aneurysms, Rathke cleft cysts, craniopharingioma and mucocele. • The role of imaging is still debated and needs more studies.

Dynamic intraoperative MRI in transsphenoidal resection of pituitary macroadenomas: a quantitative analysis.

PURPOSE: To compare intraoperative dynamic contrast-enhanced (dCE) sequences with conventional CE (cCE) in the evaluation of the surgical bed after transsphenoidal removal of pituitary macroadenomas. MATERIALS AND METHODS: Twenty-one patients with macroadenoma were selected. They all underwent intraoperative magnetic resonance imaging (iMRI) (1.5T) acquisitions during transsphenoidal resection of the tumor. For each patient, dCE and cCE images were acquired in the operating room after tumor removal. The mean values of surgical cavities volumes were measured and statistically compared through Student's t-test analysis. Informed consent to iMRI was obtained from the patients as a part of the surgical procedure. Institutional Review Board (IRB) approval was obtained. RESULTS: No patient showed recurrence within at least 1 year of follow-up. Two patients showed residual tumor in the iMRI. Intraoperative analysis of the remaining 19 demonstrated that the mean value of the surgical cavities was significantly bigger in dCE than in cCE images (2955 mm(3) vs. 1963 mm(3) , respectively, P = 0.022). CONCLUSION: This study demonstrated underestimation of surgical cavity by conventional iMRI, simulating residual tumor and potentially leading to unnecessary surgical revision.