Deep Learning Denoising of Low-Dose Computed Tomography Chest Images: A Quantitative and Qualitative Image Analysis.

PURPOSE: To assess deep learning denoised (DLD) computed tomography (CT) chest images at various low doses by both quantitative and qualitative perceptual image analysis. METHODS: Simulated noise was inserted into sinogram data from 32 chest CTs acquired at 100 mAs, generating anatomically registered images at 40, 20, 10, and 5 mAs. A DLD model was developed, with 23 scans selected for training, 5 for validation, and 4 for test.Quantitative analysis of perceptual image quality was assessed with Structural SIMilarity Index (SSIM) and Fréchet Inception Distance (FID). Four thoracic radiologists graded overall diagnostic image quality, image artifact, visibility of small structures, and lesion conspicuity. Noise-simulated and denoised image series were evaluated in comparison with one another, and in comparison with standard 100 mAs acquisition at the 4 mAs levels. Statistical tests were conducted at the 2-sided 5% significance level, with multiple comparison correction. RESULTS: At the same mAs levels, SSIM and FID between noise-simulated and reconstructed DLD images indicated that images were closer to a perfect match with increasing mAs (closer to 1 for SSIM, and 0 for FID).In comparing noise-simulated and DLD images to standard-dose 100-mAs images, DLD improved SSIM and FID. Deep learning denoising improved SSIM of 40-, 20-, 10-, and 5-mAs simulations in comparison with standard-dose 100-mAs images, with change in SSIM from 0.91 to 0.94, 0.87 to 0.93, 0.67 to 0.87, and 0.54 to 0.84, respectively. Deep learning denoising improved FID of 40-, 20-, 10-, and 5-mAs simulations in comparison with standard-dose 100-mAs images, with change in FID from 20 to 13, 46 to 21, 104 to 41, and 148 to 69, respectively.Qualitative image analysis showed no significant difference in lesion conspicuity between DLD images at any mAs in comparison with 100-mAs images. Deep learning denoising images at 10 and 5 mAs were rated lower for overall diagnostic image quality ( P < 0.001), and at 5 mAs lower for overall image artifact and visibility of small structures ( P = 0.002), in comparison with 100 mAs. CONCLUSIONS: Deep learning denoising resulted in quantitative improvements in image quality. Qualitative assessment demonstrated DLD images at or less than 10 mAs to be rated inferior to standard-dose images.

Increased Incidence of Barotrauma in Patients with COVID-19 on Invasive Mechanical Ventilation.

Background A high number of patients with coronavirus disease 2019 (COVID-19) pneumonia who had barotrauma related to invasive mechanical ventilation at the authors' institution were observed. Purpose To determine if the rate of barotrauma in patients with COVID-19 infection was greater than in other patients requiring invasive mechanical ventilation at the authors' institution. Materials and Methods In this retrospective study, clinical and imaging data of patients seen between March 1, 2020, and April 6, 2020, who tested positive for COVID-19 and experienced barotrauma associated with invasive mechanical ventilation, were compared with patients without COVID-19 infection during the same period. Historical comparison was made to barotrauma rates of patients with acute respiratory distress syndrome from February 1, 2016, to February 1, 2020, at the authors' institution. Comparison of patient groups was performed using categoric or continuous statistical testing as appropriate, with multivariable regression analysis. Patient survival was assessed using Kaplan-Meier curves analysis. Results A total of 601 patients with COVID-19 infection underwent invasive mechanical ventilation (mean age, 63 years ± 15 [standard deviation]; 71% men). Of the total, there were 89 (15%) patients with one or more barotrauma events for a total of 145 barotrauma events (24% overall events) (95% confidence interval [CI]: 21%, 28%). During the same period, 196 patients without COVID-19 infection (mean age, 64 years ± 19; 52% men) with invasive mechanical ventilation had one barotrauma event (0.5%; 95% CI: 0%, 3%; P < .001 vs the group with COVID-19 infection). Of 285 patients with acute respiratory distress syndrome on invasive mechanical ventilation during the previous 4 years (mean age, 68 years ± 17; 60% men), 28 patients (10%) had 31 barotrauma events, with an overall barotrauma rate of 11% (95% CI: 8%, 15%; P < .001 vs the group with COVID-19 infection). Barotrauma is an independent risk factor for death in COVID-19 (odds ratio = 2.2; P = .03) and is associated with a longer hospital stay (odds ratio = 0.92; P < .001). Conclusion Patients with coronavirus disease 2019 (COVID-19) infection and invasive mechanical ventilation had a higher rate of barotrauma than patients with acute respiratory distress syndrome and patients without COVID-19 infection. © RSNA, 2020 Online supplemental material is available for this article.

High-throughput antibody engineering in mammalian cells by CRISPR/Cas9-mediated homology-directed mutagenesis.

Antibody engineering is often performed to improve therapeutic properties by directed evolution, usually by high-throughput screening of phage or yeast display libraries. Engineering antibodies in mammalian cells offer advantages associated with expression in their final therapeutic format (full-length glycosylated IgG); however, the inability to express large and diverse libraries severely limits their potential throughput. To address this limitation, we have developed homology-directed mutagenesis (HDM), a novel method which extends the concept of CRISPR/Cas9-mediated homology-directed repair (HDR). HDM leverages oligonucleotides with degenerate codons to generate site-directed mutagenesis libraries in mammalian cells. By improving HDR to a robust efficiency of 15-35% and combining mammalian display screening with next-generation sequencing, we validated this approach can be used for key applications in antibody engineering at high-throughput: rational library construction, novel variant discovery, affinity maturation and deep mutational scanning (DMS). We anticipate that HDM will be a valuable tool for engineering and optimizing antibodies in mammalian cells, and eventually enable directed evolution of other complex proteins and cellular therapeutics.

SARS-CoV-2 reactive and neutralizing antibodies discovered by single-cell sequencing of plasma cells and mammalian display.

Characterization of COVID-19 antibodies has largely focused on memory B cells; however, it is the antibody-secreting plasma cells that are directly responsible for the production of serum antibodies, which play a critical role in resolving SARS-CoV-2 infection. Little is known about the specificity of plasma cells, largely because plasma cells lack surface antibody expression, thereby complicating their screening. Here, we describe a technology pipeline that integrates single-cell antibody repertoire sequencing and mammalian display to interrogate the specificity of plasma cells from 16 convalescent patients. Single-cell sequencing allows us to profile antibody repertoire features and identify expanded clonal lineages. Mammalian display screening is used to reveal that 43 antibodies (of 132 candidates) derived from expanded plasma cell lineages are specific to SARS-CoV-2 antigens, including antibodies with high affinity to the SARS-CoV-2 receptor-binding domain (RBD) that exhibit potent neutralization and broad binding to the RBD of SARS-CoV-2 variants (of concern/interest).

A Single-Cell Atlas of Lymphocyte Adaptive Immune Repertoires and Transcriptomes Reveals Age-Related Differences in Convalescent COVID-19 Patients.

COVID-19 disease outcome is highly dependent on adaptive immunity from T and B lymphocytes, which play a critical role in the control, clearance and long-term protection against SARS-CoV-2. To date, there is limited knowledge on the composition of the T and B cell immune receptor repertoires [T cell receptors (TCRs) and B cell receptors (BCRs)] and transcriptomes in convalescent COVID-19 patients of different age groups. Here, we utilize single-cell sequencing (scSeq) of lymphocyte immune repertoires and transcriptomes to quantitatively profile the adaptive immune response in COVID-19 patients of varying age. We discovered highly expanded T and B cells in multiple patients, with the most expanded clonotypes coming from the effector CD8+ T cell population. Highly expanded CD8+ and CD4+ T cell clones show elevated markers of cytotoxicity (CD8: PRF1, GZMH, GNLY; CD4: GZMA), whereas clonally expanded B cells show markers of transition into the plasma cell state and activation across patients. By comparing young and old convalescent COVID-19 patients (mean ages = 31 and 66.8 years, respectively), we found that clonally expanded B cells in young patients were predominantly of the IgA isotype and their BCRs had incurred higher levels of somatic hypermutation than elderly patients. In conclusion, our scSeq analysis defines the adaptive immune repertoire and transcriptome in convalescent COVID-19 patients and shows important age-related differences implicated in immunity against SARS-CoV-2.

Optimization of therapeutic antibodies by predicting antigen specificity from antibody sequence via deep learning.

The optimization of therapeutic antibodies is time-intensive and resource-demanding, largely because of the low-throughput screening of full-length antibodies (approximately 1 × 103 variants) expressed in mammalian cells, which typically results in few optimized leads. Here we show that optimized antibody variants can be identified by predicting antigen specificity via deep learning from a massively diverse space of antibody sequences. To produce data for training deep neural networks, we deep-sequenced libraries of the therapeutic antibody trastuzumab (about 1 × 104 variants), expressed in a mammalian cell line through site-directed mutagenesis via CRISPR-Cas9-mediated homology-directed repair, and screened the libraries for specificity to human epidermal growth factor receptor 2 (HER2). We then used the trained neural networks to screen a computational library of approximately 1 × 108 trastuzumab variants and predict the HER2-specific subset (approximately 1 × 106 variants), which can then be filtered for viscosity, clearance, solubility and immunogenicity to generate thousands of highly optimized lead candidates. Recombinant expression and experimental testing of 30 randomly selected variants from the unfiltered library showed that all 30 retained specificity for HER2. Deep learning may facilitate antibody engineering and optimization.

Antibody discovery and engineering by enhanced CRISPR-Cas9 integration of variable gene cassette libraries in mammalian cells.

Antibody engineering in mammalian cells offers the important advantage of expression and screening of libraries in their native conformation, increasing the likelihood of generating candidates with more favorable molecular properties. Major advances in cellular engineering enabled by CRISPR-Cas9 genome editing have made it possible to expand the use of mammalian cells in biotechnological applications. Here, we describe an antibody engineering and screening approach where complete variable light (VL) and heavy (VH) chain cassette libraries are stably integrated into the genome of hybridoma cells by enhanced Cas9-driven homology-directed repair (HDR), resulting in their surface display and secretion. By developing an improved HDR donor format that utilizes in situ linearization, we are able to achieve >15-fold improvement of genomic integration, resulting in a screening workflow that only requires a simple plasmid electroporation. This proved suitable for different applications in antibody discovery and engineering. By integrating and screening an immune library obtained from the variable gene repertoire of an immunized mouse, we could isolate a diverse panel of >40 unique antigen-binding variants. Additionally, we successfully performed affinity maturation by directed evolution screening of an antibody library based on random mutagenesis, leading to the isolation of several clones with affinities in the picomolar range.

Genome Engineering of Hybridomas to Generate Stable Cell Lines for Antibody Expression.

From the perspective of academic and small research laboratories, the most common and practical strategy for recombinant expression of full-length monoclonal antibodies is to perform transient plasmid transfection of mammalian cells, resulting in small-scale and limited protein production. The generation of stable antibody producing mammalian cell lines enables larger-scale and consistent expression, however this approach is rarely pursued due to the time-consuming and expensive process of single colony screening and characterization. In order to bridge the gap between the simplicity of transient transfection and consistent production by stable cell lines, we describe a method to stably integrate antibody genes into the endogenous immunogenomic loci of hybridoma cells using CRISPR/Cas9 genome editing. Initially, the antibody variable light (VL) chain is deleted by multiplexed Cas9 cleavage; subsequently, the variable heavy (VH) chain is replaced by a fluorescent reporter gene (mRuby) by Cas9-assisted homology-directed repair (HDR). This cell line is customized by replacing mRuby with a synthetic antibody (consisting of a VL, light constant region and VH) by once again using Cas9-assisted HDR. Due to hybridomas' inherent ability to surface display and secrete antibodies, they provide a suitable host for both the selection and the production process, substantially streamlining the process for stable cell line generation, and thus we refer to this platform as plug-and-(dis)play (PnP) hybridomas.