ProgNet: COVID-19 Prognosis Using Recurrent and Convolutional Neural Networks

Aims: Prognosis of lung mathology severity after Covid-19 infection using chest X-ray time series Background: We have been inspired by methods analysing time series of images in remote sensing for change detection. During the current Covid-19 pandemic, our motivation is to provide an automatic tool to predict severity of lung pathologies due to Covid-19. This can be done by analysing images of the same patient acquired at different dates. Since no analytical model is available, and also no accurate quantification tools can be used due to many unknowns about the pathology, feature-free methods are good candidates to analyse such temporal images. Objective: This contribution helps improving performances of medical structures facing the Covid-19 pandemic. The first impact is medical and social since more lives could be saved with a 92% rate of good prognosis. In addition to that, patients in intensive care units (up to 15%) could a posteriori suffer from less sequels due to an early and accurate prognosis of their PP. Moreover, accurate prognosis can lead to a better planning of patient’s transfer between units and hospitals, which is linked to the second claimed economical impact. Indeed, prognosis is linked to lower treatment costs due to an optimized predictive protocol using ragiological prognosis. Methods: Using Convolutional Neural Networks (CNN) in combination with Recurrent Neural Networks (RNN). Spatial and temporal features are combines to analyse image time series. A prognosis score is delivered indicating the severity of the pathology. Learning is made on a publicly available database. Results: When applied to radiological time-series, promising results are obtained with an accuracy rates higher than 92%. Sensitivity and specificity rates are also very interesting. Conclusion: Our method is segmentation-free, which makes it competitive with respect to other assessment methods relying on time-consuming lung segmentation algorithms. When applied on radiographic data, the proposed ProgNet architecture showed promising results with good classification performances, especially for ambiguous cases. Specifically, the reported low false positive rates are interesting for an accurate and personalised care workflow.

Development of Physical Breast Phantoms for X-ray Imaging Employing 3D Printing Techniques

Objective: This study aims to investigate the use of 3D printing techniques for the fabrication of physical breast phantoms, suitable for conventional and phase contrast breast imaging. Such phantoms could provide essential information for the design, development and optimization of emerging X-ray imaging modalities. Materials and Methods: Physical phantoms were constructed using two 3D printing techniques: Fused Deposition Modeling and Stereolithography. Eight materials suitable for 3D printing, including thermoplastic filaments and photopolymer resins, were investigated for the optimal representation of breast tissues, based on their attenuation and refractive characteristics. The phantoms consisted of a 3D-printed mold, which was then manually filled with paraffin wax. Additionally, a 3D complex-patterned layer and details representing abnormalities were embedded in different depths. Images of the phantoms were obtained in attenuation and phase contrast mode. Experiments were conducted using an X-ray microfocus tube with Tungsten anode set to 55kVp, combined with a photon-counting detector. The distance between source and detector was 56.5cm. The images were acquired at different object-to-detector distances starting from 5cm up to 40cm in a free space propagation set-up. Results and Conclusion: Results show that among all combinations with paraffin used as an adipose substitute, phantoms created with the Stereolithography technique and resins (especially Flex) as glandular equivalent, were found to be more appropriate for both attenuation and phase contrast imaging. The edge enhancement effect was well observed in the experimental images acquired at 35cm object-to-detector distance, indicating the potential for improved feature visualization using this set-up in phase contrast compared to attenuation mode.

Ripplet-Transform-based Cycle Spinning Denoising and Fuzzy-CLA Segmentation of Retinal Images for Accurate Hard Exudates and Lesion Detection

Aims: Digital retinal images are commonly used for hard exudates and lesion detection. These images are rarely noiseless and therefore before any further processing they should be underwent noise removal. Background: An efficient segmentation method is then needed to detect and discern the lesions from the retinal area. Objective: In this paper, a hybrid method is presented for digital retinal image processing for diagnosis and screening purposes. The aim of this study is to present a supervised/semi-supervised approach for exudate detection in fundus images and also to analyze the method to find the optimum structure. Methods: Ripplet transform and cycle spinning method is first used to remove the noises and artifacts. Results: The noises may be normal or any other commonly occurring forms such as salt and pepper. The image is transformed into fuzzy domain after it is denoised. Conclusion: A cellular learning automata model is used to detect any abnormality on the image which is related to a lesion. The automaton is created with an extra term as the rule updating term to improve the adaptability and efficiency of the cellular automata.Three main statistical criteria are introduced as the sensitivity, specificity and accuracy. A number of 50 retinal images with visually detection hard exudates and lesions are the experimental dataset for evaluation and validation of the method.

Comparison Between Major Types of Arthritis Based on Diagnostic Ultrasonography

Background: Ultrasound has been widely used in clinical settings for the assessment of different types of Arthritis as well as in their management. This Review study assessed the diagnostic value of Ultrasonography in comparison with major types of Arthritis namely Osteoarthritis (OA), Rheumatoid Arthritis (RA), Gouty Arthritis (GA), Pseudogout (Calcium Pyrophosphate Deposition Disease - CPPD), Psoriatic Arthritis (PA), Infectious Arthritis (IA) and Spondyloarthritis (SA). Methodology: Computerized literature search of PubMed was conducted from 1990 to present, for publications in English on diagnostic ultrasonography and major types of arthritis. A total of 206 publications were identified. Experimental and clinical studies that focused on the ultrasound features of the major types of Arthritis were accepted. A total of 52 out of the 206 publications, met our search criteria. Among these, 12 studies focused on OA, 7 on RA, 7 on GA, 5 on CPPD, 10 on PA, 4 on IA, and 7 on SA. From all the studies, some distinctive US features are reviewed for each of the major arthritis. Some of the features were unique and some overlapped. Conclusion: Ultrasound may demonstrate the ability to differentiate between the major types of Arthritis on a basic level when combined with history and physical examination. This can prove to be beneficial in the early diagnosis of the major types of arthritis, but with few limitations. This review literature shows that Ultrasound can be very helpful in bed side analysis of the major types of arthritis as well as in differentiating between them, because this modality besides being non-invasive is also very cheap.

Importance and Significance of Knee-Chest Decubitus Position of the Pregnant Women, in the Delineation of Fetal Facial Imaging with Four-Dimensional Color Doppler Ultrasound

Objective: To investigate and validate the role of the Knee-chest Decubitus position in fetal facial feature delineation by 4D Ultrasonography. Methods: Pregnant women were randomly divided into two groups: One group underwent knee-chest decubitus position prior to re-examination, and the second group underwent free activities like walking for 5, 15 or 30 minutes followed by a re-examination of the fetus. The acceptability of the fetal facial images following the two above mentioned activities was compared. Results: The Knee-chest Decubitus position was identified to be a more successful procedure for obtaining acceptable images. Additionally, it improved the fetal position and the resulting images were achieved significantly rapidly by this maneuver compared to free movement. Conclusion: The knee-chest decubitus position is simple, easy, safe and fast and thus of great convenience and promising for pregnant women.

Modifications of Midfacial Soft-Tissue Thickness Among Different Skeletal Classes in Italian Children

Background: The anatomical assessment of the arrangement of facial soft tissues has important applications in different fields from orthodontics to plastic surgery. One of the issues concerns the relationship between facial soft tissue thickness and skeletal class. Literature mainly deals with adult populations, whereas very few studies have been focused on children. Objective: This study aims at investigating the relationship between midline facial soft tissue thickness and skeletal classes in Italian pre-treatment orthodontic child patients. Methods: Lateral cephalometric X-ray films were obtained from 220 healthy Caucasoid children (91 males and 129 females), aged between 6 and 18 years (Class I: 41 males and 70 females; Class II: 18 males and 25 females; Class III: 32 males and 34 females). All the films were digitized and 14 soft tissue thicknesses were measured on the midface; in addition, the skeletal class was assessed according to the corrected ANB angle (ANBc). Differences in facial soft tissue thickness according to sex and skeletal class were assessed through two-way ANOVA test (p<0.01). Results: Statistically significant differences according to sex were found for labrale superius, stomion and labrale inferius, with thicker soft tissues in males than in females (p<0.01). Only measurements at labrale superius and gnathion showed statistically significant differences according to skeletal class, with thicker soft tissues in Class III children and thinner ones in Class II children (p<0.01). Conclusion: The limited number of investigations, as well as the differences in protocols, renders the comparison of results from different studies difficult, suggesting further investigations to enlighten this complex and debated anatomical issue.

Aortic Pulse Wave Velocities Using Real-Time Phase-Contrast MRI

Purpose: To evaluate aortic pulse wave velocities obtained by real-time phase-contrast (PC) MRI in comparison to cine PC MRI. Methods: Real-time PC MRI of eight healthy volunteers employed highly undersampled radial FLASH sequences and phase-sensitive image reconstructions by regularized nonlinear inversion (NLINV) at 40 ms temporal resolution and 1.3 mm in-plane resolution. Pulse wave velocities were analyzed for combinations of 2, 3 and 4 locations of aortic flow using time-to-upslope and cross-correlation methods. Results: For the time-to-upslope analysis mean pulse wave velocities ranged from 3.5 to 3.9 m s-1 for real-time PC MRI and from 3.5 to 3.8 m s-1 for cine PC MRI. A cross-correlation analysis of the same data resulted in 2.9 to 3.3 m s-1 and 3.3 to 3.7 m s-1, respectively. Conclusion: Real-time PC MRI determined aortic pulse wave velocities from single cardiac cycles in close correspondence to values obtained by cine PC MRI.

Single-shot Multi-slice T1 Mapping at High Spatial Resolution – Inversion-Recovery FLASH with Radial Undersampling and Iterative Reconstruction

Purpose: To develop a method for T1 mapping at high spatial resolution and for multiple slices. Methods: The proposed method emerges as a single-shot inversion-recovery experiment which covers the entire spin-lattice relaxation process by serial acquisitions of highly undersampled radial FLASH images, either in single-slice or multi-slice mode. Serial image reconstructions are performed in time-reversed order and first involve regularized nonlinear inversion (NLINV) to estimate optimum coil sensitivity profiles. Subsequently, the coil profiles are fixed for the calculation of differently T1-weighted frames and the resulting linear inverse problem is solved by a conjugate gradient (CG) technique. T1 values are obtained by pixelwise fitting with a Deichmann correction modified for multi-slice applications. Results: T1 accuracy was validated for a reference phantom. For human brain, T1 maps were obtained at 0.5 mm resolution for single-slice acquisitions and at 0.75 mm resolution for up to 5 simultaneous slices (5 mm thickness). Corresponding T1 maps of the liver were acquired at 1 mm and 1.5 mm resolution, respectively. All T1 values were in agreement with literature data. Conclusion: Inversion-recovery sequences with highly undersampled radial FLASH images and NLINV/CG reconstruction allow for fast, robust and accurate T1 mapping at high spatial resolution and for multiple slices.

Efficacy and Safety of Percutaneous Transhepatic Portal Embolization with Dehydrated Ethanol

Purpose: The efficacy and safety of percutaneous transhepatic portal embolization (PTPE) with dehydrated ethanol was determined by measuring the liver lobe volume before and after the procedure. Materials and Methods: A total of 38 patients (25 men, 13 women; mean age: 62.0 ± 10.8 years) who underwent PTPE with dehydrated ethanol between April 2005 and March 2011 participated in this study. Dehydrated ethanol containing 17% lipiodol was injected into the target portal vein branch under balloon occlusion, and the portal vein was subsequently embolized. The liver lobe volume was measured via contrast-enhanced computed tomography, and the percent increase in the unembolized lobe volume was then calculated. In addition, PTPE-related complications were surveyed, and the procedural safety was evaluated. Results: The mean percent increase in the unembolized lobe volume after PTPE was 33.8% ± 20.2%. The procedure could not be completed in one patient because of an insufficient increase in the unembolized lobe volume. No serious post- PTPE complications were observed. Conclusion: These data suggest that PTPE with dehydrated ethanol is a safe and effective method for enlarging the planned residual liver volume before extensive liver resection.

Evaluation of Sternal Bone Healing with Computed Tomography and a Quantitative Scoring Algorithm

Objective: The exquisite bone detail offered by computed tomography makes it the ideal modality for evaluation of bone healing. However, few studies have investigated the normal computed tomographic appearance of the sternum after median sternotomy and, to the best of our knowledge, no computed tomographic classification of sternal healing has been proposed. Given the potential benefit of objective criteria, we propose a validated scoring classification of sternal healing using computed tomography for both clinical and investigational purposes. Methods: Computed tomography scans from 20 patients who underwent a median sternotomy were evaluated for sternal healing at either 3 or 6 months postoperatively. Five anatomic locations along the sternum were selected using defined criteria, and a 6-point quantitative scale was developed to evaluate sternal healing. Independent radiologists read and scored each of the 5 locations on the sternum. Inter- and intra-observer variability was assessed by calculating the kappa statistics to measure the reliability of the scoring algorithm. Results: Calculation of the kappa statistics indicated substantial agreement for intra-observer variability and substantial to almost perfect agreement for inter-observer variability. For intra-observer variability, the kappa statistics ranged from 0.591 to 0.802, and for inter-observer variability, the kappa statistics ranged from 0.590 to 0.969. When the two radiologists differed, the magnitude of the difference was no more than 1 or 2 points. Conclusion: This simple system of evaluating sternal healing had high inter- and intra-observer reliability. Therefore, it may be considered a valid method for assessing sternal osteosynthesis for both clinical and investigative purposes. Ultramini abstract: (49 words): Few studies have investigated the normal computed tomography appearance of the sternum after median sternotomy, and we knew of no computed tomography-based classification of sternal healing. Given the potential benefit of objective criteria, we designed and validated a scoring classification of sternal osteosynthesis for both clinical and investigational purposes.