Towards nationally curated data archives for clinical radiology image analysis at scale: Learnings from national data collection in response to a pandemic

The prevalence of the coronavirus SARS-CoV-2 disease has resulted in the unprecedented collection of health data to support research. Historically, coordinating the collation of such datasets on a national scale has been challenging to execute for several reasons, including issues with data privacy, the lack of data reporting standards, interoperable technologies, and distribution methods. The coronavirus SARS-CoV-2 disease pandemic has highlighted the importance of collaboration between government bodies, healthcare institutions, academic researchers and commercial companies in overcoming these issues during times of urgency. The National COVID-19 Chest Imaging Database, led by NHSX, British Society of Thoracic Imaging, Royal Surrey NHS Foundation Trust and Faculty, is an example of such a national initiative. Here, we summarise the experiences and challenges of setting up the National COVID-19 Chest Imaging Database, and the implications for future ambitions of national data curation in medical imaging to advance the safe adoption of artificial intelligence in healthcare.

An overview of the National COVID-19 Chest Imaging Database: data quality and cohort analysis

BACKGROUND

The National COVID-19 Chest Imaging Database (NCCID) is a centralized database containing mainly chest X-rays and computed tomography scans from patients across the UK. The objective of the initiative is to support a better understanding of the coronavirus SARS-CoV-2 disease (COVID-19) and the development of machine learning technologies that will improve care for patients hospitalized with a severe COVID-19 infection. This article introduces the training dataset, including a snapshot analysis covering the completeness of clinical data, and availability of image data for the various use-cases (diagnosis, prognosis, longitudinal risk). An additional cohort analysis measures how well the NCCID represents the wider COVID-19-affected UK population in terms of geographic, demographic, and temporal coverage.

FINDINGS

The NCCID offers high-quality DICOM images acquired across a variety of imaging machinery; multiple time points including historical images are available for a subset of patients. This volume and variety make the database well suited to development of diagnostic/prognostic models for COVID-associated respiratory conditions. Historical images and clinical data may aid long-term risk stratification, particularly as availability of comorbidity data increases through linkage to other resources. The cohort analysis revealed good alignment to general UK COVID-19 statistics for some categories, e.g., sex, whilst identifying areas for improvements to data collection methods, particularly geographic coverage.

CONCLUSION

The NCCID is a growing resource that provides researchers with a large, high-quality database that can be leveraged both to support the response to the COVID-19 pandemic and as a test bed for building clinically viable medical imaging models.

Experimental recovery of a qubit from partial collapse

We describe and implement a method to restore the state of a single qubit, in principle perfectly, after it has partially collapsed. The method resembles the classical Hahn spin echo but works on a wider class of relaxation processes, in which the quantum state partially leaves the computational Hilbert space. It is not guaranteed to work every time, but successful outcomes are heralded. We demonstrate, using a single trapped ion, a better performance from this recovery method than can be obtained employing projection and postselection alone. The demonstration features a novel qubit implementation that permits both partial collapse and coherent manipulations with high fidelity.

High-fidelity readout of trapped-ion qubits

We demonstrate single-shot qubit readout with a fidelity sufficient for fault-tolerant quantum computation. For an optical qubit stored in 40Ca+ we achieve 99.991(1)% average readout fidelity in 10(6) trials, using time-resolved photon counting. An adaptive measurement technique allows 99.99% fidelity to be reached in 145 micros average detection time. For 43Ca+, we propose and implement an optical pumping scheme to transfer a long-lived hyperfine qubit to the optical qubit, capable of a theoretical fidelity of 99.95% in 10 micros. We achieve 99.87(4)% transfer fidelity and 99.77(3)% net readout fidelity.

Long-lived mesoscopic entanglement outside the Lamb-Dicke regime

We create entangled states of the spin and motion of a single 40Ca+ ion in a linear ion trap. We theoretically study and experimentally observe the behavior outside the Lamb-Dicke regime, where the trajectory in phase space is modified and the motional coherent states become squeezed. We directly observe the modification of the return time of the trajectory, and infer the squeezing. The mesoscopic entanglement is observed up to Deltaalpha=5.1 with coherence time 170 micros and mean phonon excitation n = 16.

ER retention may play a role in sorting of the nuclear pore membrane protein POM121

Integral membrane proteins of the nuclear envelope (NE) are synthesized on the rough endoplasmic reticulum (ER) and following free diffusion in the continuous ER/NE membrane system are targeted to their proper destinations due to interactions of specific domains with other components of the NE. By studying the intracellular distribution and dynamics of a deletion mutant of an integral membrane protein of the nuclear pores, POM121, which lacks the pore-targeting domain, we investigated if ER retention plays a role in sorting of integral membrane proteins to the nuclear envelope. A nascent membrane protein lacking sorting determinants is believed to diffuse laterally in the continuous ER/NE lipid bilayer and expected to follow vesicular traffic to the plasma membrane. The GFP-tagged deletion mutant, POM121(1-129)-GFP, specifically distributed within the ER membrane, but was completely absent from the Golgi compartment and the plasma membrane. Experiments using fluorescence recovery after photobleaching (FRAP) and fluorescence loss in photobleaching (FLIP) demonstrated that despite having very high mobility within the whole ER network (D = 0.41 +/- 0.11 micro m(2)/s) POM121(1-129)-GFP was unable to exit the ER. It was also not detected in post-ER compartments of cells incubated at 15 degrees C. Taken together, these experiments show that amino acids 1-129 of POM121 are able to retain GFP in the ER membrane and suggest that this retention occurs by a direct mechanism rather than by a retrieval mechanism. Our data suggest that ER retention might be important for sorting of POM121 to the nuclear pores.

Sequential degradation of proteins from the nuclear envelope during apoptosis

We have produced new antibodies specific for the integral pore membrane protein POM121. Using these antibodies we show that during apoptosis POM121 becomes proteolytically degraded in a caspase-dependent manner. The POM121 antibodies and antibodies specific for other proteins of the nuclear envelope were used in a comparative study of nuclear apoptosis in staurosporine-treated buffalo rat liver cells. Nuclei from these cells were classified in three different stages of apoptotic progression: stage I, moderately condensed chromatin surrounded by a smooth nuclear periphery; stage II, compact patches of condensed chromatin collapsing against a smooth nuclear periphery; stage III, round compact chromatin bodies surrounded by grape-shaped nuclear periphery. We have performed double labeling immunofluorescence microscopy of individual apoptotic cells and quantitative immunoblotting analysis of total proteins from apoptotic cell cultures. The results showed that degradation of nuclear envelope marker proteins occurred in a specific order. POM121 degradation occurred surprisingly early and was initiated before nucleosomal DNA degradation could be detected using TUNEL assay and completed before clustering of the nuclear pores. POM121 was eliminated significantly more rapid compared with NUP153 (a peripheral protein located in the nucleoplasmic basket of the nuclear pore complex) and lamin B (a component of the nuclear lamina). Disappearance of NUP153 and lamin B was coincident with onset of DNA fragmentation and clustering of nuclear pores. By contrast, the peripheral NPC protein p62 was degraded much later. The results suggest that degradation of POM121 may be an important early step in propagation of nuclear apoptosis.

Nuclear pore complexes form immobile networks and have a very low turnover in live mammalian cells

The nuclear pore complex (NPC) and its relationship to the nuclear envelope (NE) was characterized in living cells using POM121-green fluorescent protein (GFP) and GFP-Nup153, and GFP-lamin B1. No independent movement of single pore complexes was found within the plane of the NE in interphase. Only large arrays of NPCs moved slowly and synchronously during global changes in nuclear shape, strongly suggesting mechanical connections which form an NPC network. The nuclear lamina exhibited identical movements. NPC turnover measured by fluorescence recovery after photobleaching of POM121 was less than once per cell cycle. Nup153 association with NPCs was dynamic and turnover of this nucleoporin was three orders of magnitude faster. Overexpression of both nucleoporins induced the formation of annulate lamellae (AL) in the endoplasmic reticulum (ER). Turnover of AL pore complexes was much higher than in the NE (once every 2.5 min). During mitosis, POM121 and Nup153 were completely dispersed and mobile in the ER (POM121) or cytosol (Nup153) in metaphase, and rapidly redistributed to an immobilized pool around chromatin in late anaphase. Assembly and immobilization of both nucleoporins occurred before detectable recruitment of lamin B1, which is thus unlikely to mediate initiation of NPC assembly at the end of mitosis.

An integral membrane protein from the nuclear pore complex is also present in the annulate lamellae: implications for annulate lamella formation

Annulate lamellae (AL) are cytoplasmic arrays of stacked membrane cisternae containing densely packed pore complexes which are similar in structure to the nuclear pore complexes (NPCs) and thus referred to as annulate lamella pore complexes (ALPCs). We have recently shown that the integral nuclear pore membrane protein POM121 tagged with green fluorescent protein was correctly targeted to the nuclear pores (H. Söderqvist et al., 1997, Eur. J. Biochem. 250, 808-813). Here we have investigated if POM121 fused to three tandem molecules of yellow fluorescent protein (YFP) (POM121-YFP(3)) also was able to distribute in the extensive and well-characterized AL of RC37 and BMGE cells. Transfected RC37 or BMGE cells displayed YFP fluorescence around the nuclear envelope, as well as in the cytoplasmic AL structures. The YFP fluorescence colocalized perfectly with immunostaining using antibodies specific for different NPC proteins. The AL of both transfected and untransfected BMGE cells resisted extractions with Tx-100 and 250 mM NaCl, but were completely solubilized at 450 mM NaCl. Loss of YFP fluorescence and immunostaining for other NPC proteins correlated under all extraction conditions tested, suggesting that overexpressed POM121-YFP(3) had become an integrated part both of the NPCs and of the ALPCs. Furthermore, we have generated a stable BHK cell line expressing POM121-YFP(3) located exclusively at the nuclear pores. Treatment with vinblastine sulfate, which induces formation of AL in a variety of cells, resulted in distribution of POM121-YFP(3) into cytoplasmic foci colocalizing with immunostaining for peripheral NPC proteins. Taken together, the results show that YFP-tagged POM121 is able to distribute in drug-induced or naturally occurring AL, suggesting that POM121 is a natural constituent of ALPCs. In COS cells, which normally lack or have very little AL, YFP-tagged POM121 distributed in the nuclear pores when expressed at low levels. However, at high expression levels the YFP fluorescence also distributed in a number of brightly fluorescing cytoplasmic dots or foci, which were not present in untransfected cells. This was also true for untagged POM121. The cytoplasmic foci varied in size from 0. 1 to 2 microm and were distinctly located in the immediate vicinity of ER cisternae (without colocalizing) and also contained other nuclear pore proteins, indicating that they may represent cytoplasmic AL. This idea is supported by time-lapse studies of postmitotic assembly of these structures. This raises the question of the role of POM121 in ALPC and NPC biogenesis.

Noninvasive monitoring of apoptosis versus necrosis in a neuroblastoma cell line expressing a nuclear pore protein tagged with the green fluorescent protein

A fusion chimera between the integral nuclear pore membrane protein POM121 and GFP (green fluorescent protein) has been shown to correctly target to the nuclear pores when transiently expressed in a number of mammalian cell types. POM121-GFP is therefore an excellent marker for the noninvasive studies of the nuclear pores in living cells using fluorescence microscopy. We have established a line of neuroblastoma cells stably expressing the POM121-GFP fusion protein. We also monitored the nuclear envelope in living cells after induction of apoptosis or necrosis using 1 microM staurosporine or 100 microM p-benzoquinone, respectively. Interestingly, the POM121-GFP fluorescence was weaker or missing in the apoptotic cells. The disappearance of the nuclear pore marker accompanied apoptotic progression as judged by the degree of chromatin condensation and DNA fragmentation as analyzed by DNA staining and TUNEL assay, respectively. In contrast, the intensity of the nuclear rim fluorescence was unaffected in necrotic cells displaying an abnormal morphology with tilted nuclei. Thus, it was possible to distinguish between apoptotic and necrotic development in living cells using fluorescence microscopy. This cell line provides a fast and convenient model for screening suspected toxic xenobiotics.

Intracellular distribution of an integral nuclear pore membrane protein fused to green fluorescent protein--localization of a targeting domain

The 121-kDa pore membrane protein (POM121) is a bitopic integral membrane protein specifically located in the pore membrane domain of the nuclear envelope with its short N-terminal tail exposed on the luminal side and its major C-terminal portion adjoining the nuclear pore complex. In order to locate a signal for targeting of POM121 to the nuclear pores, we overexpressed selected regions of POM121 alone or fused to the green fluorescent protein (GFP) in transiently transfected COS-1 cells or in a stably transfected neuroblastoma cell line. Microscopic analysis of the GFP fluorescence or immunostaining was used to determine the intracellular distribution of the overexpressed proteins. The endofluorescent GFP tag had no effect on the distribution of POM121, since the chimerical POM121-GFP fusion protein was correctly targeted to the nuclear pores of both COS-1 cells and neuroblastoma cells. Based on the differentiated intracellular sorting of the POM121 variants, we conclude that the first 128 amino acids of POM121 contains signals for targeting to the continuous endoplasmic reticulum/nuclear envelope membrane system but not specifically to the nuclear pores and that a specific nuclear pore targeting signal is located between amino acids 129 and 618 in the endoplasmically exposed portion of POM121.