In women younger than 50 years, benefits of early diagnosis with tomosynthesis plus mammography screening seemed appreciable and could increase diagnosis of slow-growing, invasive cancers for women... Background Adding digital breast tomosynthesis (DBT) to digital mammography (DM) improves breast cancer screening sensitivity, but how this impacts mortality and other end points is unknown. Purpos...

Aim: The aim of this study was to develop robust prognostic models for mortality prediction of COVID-19 patients, applicable to different sets of real scenarios, using radiomic and neural network features extracted from chest X-rays (CXRs) with a certified and commercially available software. Methods: 1816 patients from 5 different hospitals in the Province of Reggio Emilia were included in the study. Overall, 201 radiomic features and 16 neural network features were extracted from each COVID-19 patient’s radiography. The initial dataset was balanced to train the classifiers with the same number of dead and survived patients, randomly selected. The pipeline had three main parts: balancing procedure; three-step feature selection; and mortality prediction with radiomic features through three machine learning (ML) classification models: AdaBoost (ADA), Quadratic Discriminant Analysis (QDA) and Random Forest (RF). Five evaluation metrics were computed on the test samples. The performance for death prediction was validated on both a balanced dataset (Case 1) and an imbalanced dataset (Case 2). Results: accuracy (ACC), area under the ROC-curve (AUC) and sensitivity (SENS) for the best classifier were, respectively, 0.72 ± 0.01, 0.82 ± 0.02 and 0.84 ± 0.04 for Case 1 and 0.70 ± 0.04, 0.79 ± 0.03 and 0.76 ± 0.06 for Case 2. These results show that the prediction of COVID-19 mortality is robust in a different set of scenarios. Conclusions: Our large and varied dataset made it possible to train ML algorithms to predict COVID-19 mortality using radiomic and neural network features of CXRs.

The aim of this article was to characterise the performance of four different digital breast tomosynthesis (DBT) systems in terms of dose and image quality parameters. One of them, GE Pristina, has never been tested before. Average glandular doses were measured both in DBT and 2D full field digital mammography mode. Several phantoms were employed to perform signal difference to noise ratio, slice sensitivity profile, slice to slice incrementation, chest wall offset, z-axis geometry, artefact spread function, low contrast detectability, contrast detail evaluations, image uniformity and in-plane MTF in chest wall-nipple and in tube-travel directions. There are many differences in DBT systems explored: the angular range, detector type, reconstruction algorithms, and the presence or not of the grid. Even if it is not simple to calculate a global figure of merit, the analysis of all the collected data can be useful in a contest of a quality assurance program to define a set of values that could be used as benchmarks.

Aim: Machine learning (ML) and deep learning (DL) predictive models have been employed widely in clinical settings. Their potential support and aid to the clinician of providing an objective measure that can be shared among different centers enables the possibility of building more robust multicentric studies. This study aimed to propose a user-friendly and low-cost tool for COVID-19 mortality prediction using both an ML and a DL approach. Method: We enrolled 2348 patients from several hospitals in the Province of Reggio Emilia. Overall, 19 clinical features were provided by the Radiology Units of Azienda USL-IRCCS of Reggio Emilia, and 5892 radiomic features were extracted from each COVID-19 patient’s high-resolution computed tomography. We built and trained two classifiers to predict COVID-19 mortality: a machine learning algorithm, or support vector machine (SVM), and a deep learning model, or feedforward neural network (FNN). In order to evaluate the impact of the different feature sets on the final performance of the classifiers, we repeated the training session three times, first using only clinical features, then employing only radiomic features, and finally combining both information. Results: We obtained similar performances for both the machine learning and deep learning algorithms, with the best area under the receiver operating characteristic (ROC) curve, or AUC, obtained exploiting both clinical and radiomic information: 0.803 for the machine learning model and 0.864 for the deep learning model. Conclusions: Our work, performed on large and heterogeneous datasets (i.e., data from different CT scanners), confirms the results obtained in the recent literature. Such algorithms have the potential to be included in a clinical practice framework since they can not only be applied to COVID-19 mortality prediction but also to other classification problems such as diabetic prediction, asthma prediction, and cancer metastases prediction. Our study proves that the lesion’s inhomogeneity depicted by radiomic features combined with clinical information is relevant for COVID-19 mortality prediction.

The aim of the study is to use the well-known Channelized Hotelling Observer model (CHO) to characterize a recently installed angiography system (GE Discovery IGS 740) using sets of images of a contrast-detail phantom acquired with clinical protocols. A Leeds TO10 phantom was used. The phantom has 108 details: 12 diameters (size range: 0.25 mm – 11 mm), each with 9 contrasts (declared range: 0.012 – 0.930 at 70 kVp 1.00 with 1 mm Cu filtration). TO10 has been imaged between two 10 cm thick homogeneous solid water slabs. Two FOVs (32 cm and 20 cm) were used. Fluoroscopy images were taken using an abdominal protocol at 2 different frame rates (15 fps and 7.5 fps) and at 2 dose levels (low and normal); cineangiography images were acquired using an abdominal protocol at 15 fps at 2 dose levels (low and normal). A 40 Gabor channels CHO with internal noise was used. Human observers' studies were carried out to tune the internal noise parameter and to validate the model observer. Contrast-detail curves were obtained from the CHO output using a visibility threshold of 75% and fitted with Rose's model theory in order to characterize the angiography system. Wilcoxon rank-sum tests were performed to investigate possible differences among the different sets of images. The CHO can distinguish between the two dose levels (p-values<0.002), while FOV and frame rate do not affect the contrast-detail curves significantly. It is important to note that the CHO does not find statistically significant differences between a fluoroscopy with FOV=20 cm at normal dose level (17.6 mGy/min) and a cineangiography with FOV=32 cm low dose level (42.1 mGy/min). This result can lead to a dose reduction of about 70% for our specific task (i.e., a static, disc shaped object at known location in homogeneous field). Given their stability in comparison to human observers, model observers provide an effective tool for image quality evaluation.

This work aims to characterize the novel DRX Plus 3543C detector in terms of detective quantum efficiency (DQE) using both a mobile x-ray system called Carestream DRX Revolution Nano and a traditional x-ray system (Carestream DRX Evolution). We used the commercial system DRX Revolution Nano, equipped with a new x-ray source based on CNT technology and field emission (FE) as the electron emitter (cathode). An innovative aspect of this device is its intrinsic selection of the focal spot size. We tested the system using three IEC-specified beam qualities (RQA3, 5 and 7) in terms of modulation transfer function (MTF), normalized noise power spectra (NNPS) and DQE as defined in the IEC 62220-1-1:2015. We compared the results obtained using DRX Revolution Nano and DRX Evolution with correlation and with Bland–Altman plots to study their agreement. RQA3 MTF is slightly lower than the RQA5 and 7 curves between 0.5 and 2.5 cycles mm ⁻¹ . We measured MTF values of about 0.6 at 1 lp mm ⁻¹ and about 0.28 lp mm ⁻¹ at 2 lp mm ⁻¹ . The NNPS curves show a decreasing trend with the energy regarding the DRX Revolution Nano. On the other hand, the DRX Evolution NNPS curve at RQA3 is greater than the one at RQA5, but the one at RQA5 is less than the one at RQA7. The DQE(0) ranged between about 0.82 (DRX Evolution at RQA3) and 0.54 (DRX Evolution at RQA7). As expected, the squared Pearson’s correlation coefficients between the two x-ray tubes were always in an optimal agreement, and Bland–Altman plots confirmed a substantial equivalence between the two physical characterizations of the wireless detector. In conclusion, we can show that the dynamic focal selection of the system equipped with CNT does not play a substantial role in image quality compared to a traditional system in terms of physical characterisation of the detector in our measurement conditions.