Artificial intelligence computed tomography helps evaluate the severity of COVID-19 patients:A retrospective study

BACKGROUND:Computed tomography(CT)is a noninvasive imaging approach to assist the early diagnosis of pneumonia.However,coronavirus disease 2019(COVID-19)shares similar imaging features with other types of pneumonia,which makes differential diagnosis problematic.Artificial intelligence(AI)has been proven successful in the medical imaging field,which has helped disease identification.However,whether AI can be used to identify the severity of COVID-19 is still underdetermined.METHODS:Data were extracted from 140 patients with confirmed COVID-19.The severity of COVID-19 patients(severe vs.non-severe)was defined at admission,according to American Thoracic Society(ATS)guidelines for community-acquired pneumonia(CAP).The AI-CT rating system constructed by Hangzhou YITU Healthcare Technology Co.,Ltd.was used as the analysis tool to analyze chest CT images.RESULTS:A total of 117 diagnosed cases were enrolled,with 40 severe cases and 77 non-severe cases.Severe patients had more dyspnea symptoms on admission(12 vs.3),higher acute physiology and chronic health evaluation(APACHE)II(9 vs.4)and sequential organ failure assessment(SOFA)(3 vs.1)scores,as well as higher CT semiquantitative rating scores(4 vs.1)and AI-CT rating scores than non-severe patients(P<0.001).The AI-CT score was more predictive of the severity of COVID-19(AUC=0.929),and ground-glass opacity(GGO)was more predictive of further intubation and mechanical ventilation(AUC=0.836).Furthermore,the CT semiquantitative score was linearly associated with the AI-CT rating system(Adj R2=75.5%,P<0.001).CONCLUSIONS:AI technology could be used to evaluate disease severity in COVID-19 patients.Although it could not be considered an independent factor,there was no doubt that GGOs displayed more predictive value for further mechanical ventilation.

High-performance eight-channel system with fractal superconducting nanowire single-photon detectors

Superconducting nanowire single-photon detectors(SNSPDs)have become a mainstream photon-counting technology that has been widely applied in various scenarios.So far,most multi-channel SNSPD systems,either reported in literature or commercially available,are polarization sensitive,that is,the system detection efficiency(SDE)of each channel is dependent on the state of polarization of the to-be-detected photons.Here,we reported an eight-channel system with fractal SNSPDs working in the wavelength range of 930 to 940 nm,which are all featured with low polarization sensitivity.In a close-cycled Gifford-McMahon cryocooler system with the base temperature of 2.2 K,we installed and compared the performance of two types of devices:(1)SNSPD,composed of a single,continuous nanowire and(2)superconducting nanowire avalanche photodetector(SNAP),composed of 16 cascaded units of two nanowires electrically connected in parallel.The highest SDE among the eight channels reaches 96+^(4)_(-5%),with the polarization sensitivity of 1.02 and a dark-count rate of 13 counts per second.The average SDE for eight channels for all states of polarization is estimated to be 90±5%.It is concluded that both the SNSPDs and the SNAPs can reach saturated,high SDE at the wavelength of interest,and the SNSPDs show lower dark-count(false-count)rates,whereas the SNAPs show better properties in the time domain.With the adoption of this system,we showcased the measurements of the second-order photon-correlation functions of light emission from a singlephoton source based on a semiconductor quantum dot and from a pulsed laser.It is believed that this work will provide new choices of systems with single-photon detectors combining the merits of high SDE,low polarization sensitivity,and low noise that can be tailored for different applications.