Nature Biomedical Engineering最新文献

In structural magnetic resonance (MR) imaging, motion artefacts, low resolution, imaging noise and variability in acquisition protocols frequently degrade image quality and confound downstream analyses. Here we report a foundation model for the motion correction, resolution enhancement, denoising and harmonization of MR images. Specifically, we trained a tissue-classification neural network to predict tissue labels, which are then leveraged by a ‘tissue-aware’ enhancement network to generate high-quality MR images. We validated the model’s effectiveness on a large and diverse dataset comprising 2,448 deliberately corrupted images and 10,963 images spanning a wide age range (from foetuses to elderly individuals) acquired using a variety of clinical scanners across 19 public datasets. The model consistently outperformed state-of-the-art algorithms in improving the quality of MR images, handling pathological brains with multiple sclerosis or gliomas, generating 7-T-like images from 3 T scans and harmonizing images acquired from different scanners. The high-quality, high-resolution and harmonized images generated by the model can be used to enhance the performance of models for tissue segmentation, registration, diagnosis and other downstream tasks.

Machine learning models for the diagnosis of breast cancer can facilitate the prediction of cancer risk and subsequent patient management among other clinical tasks. For the models to impact clinical practice, they ought to follow standard workflows, help interpret mammography and ultrasound data, evaluate clinical contextual information, handle incomplete data and be validated in prospective settings. Here we report the development and testing of a multimodal model leveraging mammography and ultrasound modules for the stratification of breast cancer risk based on clinical metadata, mammography and trimodal ultrasound (19,360 images of 5,216 breasts) from 5,025 patients with surgically confirmed pathology across medical centres and scanner manufacturers. Compared with the performance of experienced radiologists, the model performed similarly at classifying tumours as benign or malignant and was superior at pathology-level differential diagnosis. With a prospectively collected dataset of 191 breasts from 187 patients, the overall accuracies of the multimodal model and of preliminary pathologist-level assessments of biopsied breast specimens were similar (90.1% vs 92.7%, respectively). Multimodal models may assist diagnosis in oncology.

Organoids for modelling the physiology and pathology of gastrointestinal tissues are constrained by a poorly accessible lumen. Here we report the development and applicability of bilaterally accessible organoid-derived patterned epithelial monolayers that allow the independent manipulation of their apical and basal sides. We constructed gastric, small-intestinal, caecal and colonic epithelial models that faithfully reproduced their respective tissue geometries and that exhibited stem cell regionalization and transcriptional resemblance to in vivo epithelia. The models’ enhanced observability allowed single-cell tracking and studies of the motility of cells in immersion culture and at the air–liquid interface. Models mimicking infection of the caecal epithelium by the parasite Trichuris muris allowed us to live image syncytial tunnel formation. The enhanced observability of bilaterally accessible organoid-derived gastrointestinal tissue will facilitate the study of the dynamics of epithelial cells and their interactions with pathogens.

Potent agonists of the inducible co-stimulatory receptor 4-1BB are too toxic for patients with advanced cancer. Here, on the basis of observations of a weak agonist of 4-1BB depleting regulatory T (T_reg) cells within the tumour microenvironment without leading to substantial restoration of dysfunctional cytotoxic T cells (CTLs), we show that effective tumour control can be achieved via concurrent T_reg cell depletion and CTL expansion through an anti-4-1BB antibody fused to interleukin-15 (IL-15) via a peptide sensitive to tumour proteases. In mouse models of advanced cancers, intraperitoneal injection of the bifunctional protein attenuated the activity of the interleukin mostly in the periphery of the primary tumour while allowing for the expansion of CTLs within the tumour microenvironment, led to more effective tumour inhibition and to lower systemic toxicity than treating the cancers with combinatorial treatment with unlinked anti-4-1BB antibody and IL-15, and reduced the resistance of tumours to checkpoint blockade. Concurrent eradication of T_reg cells and activation of tumour-infiltrating lymphocytes may represent a general strategy for the effective control of advanced metastatic tumours.

In magnetic resonance imaging (MRI), quantitative measurements of analytes are hindered by difficulties in distinguishing the MRI signals of activation of the probe by the analyte from those of the accumulation of the intact probe. Here we show that imaging sensitivity and quantitation can be enhanced by ratiometric MRI probes with a high relaxivity-ratio change (more than 2.5-fold at 7 T) via magnetic-susceptibility-dependent magnetic resonance tuning. Specifically, polymeric probes that incorporate paramagnetic Mn-porphyrin and superparamagnetic iron oxide nanoparticles inducing opposite changes in the longitudinal and transverse magnetic relaxivities responded to analyte concentration independently of probe concentration. In mice, the probes allowed for quantitative real-time dynamic imaging of H₂O₂, H₂S or pH in subcutaneous tumours, in livers with drug-induced injury and in orthotropic gliomas. The ratiometric MRI probes may be advantageously used to obtain molecular insight into pathological processes and to circumvent interference from dynamic changes in probe concentration within the body while providing anatomical information.

Optimization of oncolytic viruses for therapeutic applications requires the strategic removal or mutagenesis of virulence genes alongside the insertion of transgenes that enhance viral replication, spread and immunogenicity. However, the complexity of many viral genomes and the labour-intensive nature of methods for the generation and isolation of recombinant viruses have hindered the development of therapeutic oncolytic viruses. Here we report an iterative strategy that exploits the preferential susceptibility of viruses to certain antibiotics to accelerate the engineering of the genomes of oncolytic viruses for the insertion of immunomodulatory cytokine transgenes, and the identification of dispensable genes with regard to replication of the recombinant oncolytic viruses in tumour cells. We applied the strategy by leveraging insertional mutagenesis via the Sleeping Beauty transposon system, combined with long-read nanopore sequencing, to generate libraries of herpes simplex virus type 1 and vaccinia virus, identifying stable transgene insertion sites and gene deletions that enhance the safety and efficacy of the viruses.