Reducing scan time of paediatric 99mTc-DMSA SPECT via deep learning

AIM

To investigate the feasibility of reducing the scan time of paediatric technetium 99m (^99mTc) dimercaptosuccinic acid (DMSA) single-photon-emission computed tomographic (SPECT) using a deep learning (DL) method.

MATERIAL AND METHODS

A total of 112 paediatric ^99mTc-DMSA renal SPECT scans were analysed retrospectively. Of the 112 examinations, 88 (84 for training and four for validation) were used to train a DL-based model that could generate full-acquisition-time reconstructed SPECT images from half-time acquisition. The remaining 24 examinations were used to evaluate the performance of the trained model.

RESULTS

DL-based SPECT images obtained from half-time acquisition have image quality similar to the standard clinical SPECT images obtained from full-acquisition-time acquisition. Moreover, the accuracy, sensitivity and specificity of the DL-based SPECT images for detection of affected kidneys were 91.7%, 83.3%, and 100%, respectively.

CONCLUSION

These preliminary results suggest that DL has the potential to reduce the scan time of paediatric ^99mTc-DMSA SPECT imaging while maintaining diagnostic accuracy.

Toward High Precision XCO2 Retrievals From TanSat Observations: Retrieval Improvement and Validation Against TCCON Measurements

TanSat is the 1st Chinese carbon dioxide (CO2) measurement satellite, launched in 2016. In this study, the University of Leicester Full Physics (UoL-FP) algorithm is implemented for TanSat nadir mode XCO2 retrievals. We develop a spectrum correction method to reduce the retrieval errors by the online fitting of an 8th order Fourier series. The spectrum-correction model and its a priori parameters are developed by analyzing the solar calibration measurement. This correction provides a significant improvement to the O2 A band retrieval. Accordingly, we extend the previous TanSat single CO2 weak band retrieval to a combined O2 A and CO2 weak band retrieval. A Genetic Algorithm (GA) has been applied to determine the threshold values of post-screening filters. In total, 18.3% of the retrieved data is identified as high quality compared to the original measurements. The same quality control parameters have been used in a footprint independent multiple linear regression bias correction due to the strong correlation with the XCO2 retrieval error. Twenty sites of the Total Column Carbon Observing Network (TCCON) have been selected to validate our new approach for the TanSat XCO2 retrieval. We show that our new approach produces a significant improvement on the XCO2 retrieval accuracy and precision when compared to TCCON with an average bias and RMSE of -0.08 ppm and 1.47 ppm, respectively. The methods used in this study can help to improve the XCO2 retrieval from TanSat and subsequently the Level-2 data production, and hence will be applied in the TanSat operational XCO2 processing.

Phase states, microstructure and dielectric characteristics of solid solutions (1 - x)NaNbO3 - xCa2Nb2O7 and (1 - x)NaNbO3 - xSr2Nb2O7

Ceramics of binary systems solid solutions (1 – x)NaNbO₃ – xCa₂Nb₂O₇ and (1 – x)NaNbO₃ – xSr₂Nb₂O₇ with non-isostructural extreme components were prepared by the solid-phase reactions technique with the following sintering using conventional ceramic technology. It was found that ceramics with x ≤ 0.2 have a perovskite structure. Layered type of structure predominates in the concentration range 0.2 < x ≤ 1. Phase diagrams of both systems at room temperature have been determined in the perovskite area. It was shown that this area contains two concentration regions with the different crystal structures and the morphotropic phase boundary between them. Microstructure and dielectric characteristics of selected solid solutions were investigated. The influence of technological regulations, such as mechanical activation and variation of sintering temperatures, on the formation of the microstructure and dielectric characteristics was studied for the individually selected concentrations (x = 0.1 and x = 0.25). Dielectric characteristics of ceramics revealed the presence of the Maxwell-Wagner polarization and its corresponding relaxation in the solid solutions (1 – x)NaNbO₃ – xCa₂Nb₂O₇ at x > 0.20.

Research on the invulnerability and optimization of the technical cooperation innovation network based on the patent perspective-A case study of new energy vehicles

This paper takes new energy vehicles as the research object, building the technical cooperation innovation network of new energy vehicles based on the patent perspective by establishing the related technology patent search expression, and analyzing the processes of the invulnerability and optimization in the actual technology cooperation innovation network by using the simulation analysis method. The research results show that the harmfulness of the degree value priority attack in the technical cooperation innovation network of new energy vehicles is stronger than the weighted degree value priority attack and random attack, and the attacks of the State Grid and other hub nodes have an important impact on the network invulnerability. During the network optimization process of three types of connection preferences, the “weak”-“weak” connection is the best connection mode given the situation of an unweighted network without considering the weight of the connected edge. However, the “strong”-“weak” connection is the best mode given the situation of a weighted network considering the weight of the connected edge. In addition, compared with the weighted network situation, the “strong”-“weak” connection has better network optimization results given the situation of an unweighted network. Finally, we propose counter measures and suggestions to promote the innovation network invulnerability capabilities of technical cooperation in new energy vehicles.

International normalised ratio as an independent predictor of mortality in limb necrotising fasciitis with sepsis

INTRODUCTION

Necrotising fasciitis with sepsis is a life threatening disease. The aim of this study was to analyse the association between international normalised ratio (INR) and mortality in sepsis patients with necrotising fasciitis.

METHODS

A retrospective review was undertaken of 106 patients suffering from necrotising fasciitis with sepsis between November 2007 and December 2016. Data on comorbidities, clinical manifestations, laboratory findings, causative microbiological organisms, APACHE II (Acute Physiology and Chronic Health Evaluation II) score and outcomes were extracted. Logistic regression was carried out to examine the factors affecting mortality.

RESULTS

Forty patients (37.7%) died. There was no significant difference in the white blood count (WBC) for the survivor and non-survivor groups. Non-survivors had a lower mean oxygenation index (OI) (288.7mmHg vs 329.4mmHg, p=0.032) and platelet count (PC) (139.5 vs 214.8 x 10⁹/l, p=0.028), and a higher mean INR (1.9 vs 1.3, p=0.000), activated partial thromboplastin time (APTT) (54.6 vs 44.2 seconds, p=0.005) and serum creatinine (2.3mg/dl vs 1.4mg/dl, p=0.007). Mortality in patients with INR >1.5 was significantly higher than in those with INR <1.5 when all risk factors (WBC, PC, OI, INR, APTT, creatinine) were considered (odds ratio: 4.414, 95% confidence interval: 1.263-15.428, p=0.020). Even after adjusting for age, sex, bacteraemia, diabetes and hepatic disorders, the data still exhibited elevated mortality for patients with INR >1.5 (odds ratio: 5.600, 95% confidence interval: 1.415-22.166, p=0.014).

CONCLUSIONS

INR is a significant independent predictor of mortality in sepsis patients diagnosed with necrotising fasciitis.

The invasion status of lymphovascular space and lymph nodes in cervical cancer assessed by mono-exponential and bi-exponential DWI-related parameters

AIM

To investigate whether mono-exponential and bi-exponential diffusion-weighted imaging (DWI)-related parameters of the primary tumour can evaluate the status of lymphovascular space invasion (LVSI) and lymph node metastasis (LNM) in patients with cervical carcinoma preoperatively.

MATERIALS AND METHODS

Eighty patients with cervical carcinoma were enrolled, who underwent preoperative multi b-value DWI and radical hysterectomy. They were classified into LVSI(+) versus LVSI(-) and LNM(+) versus LNM(-) according to postoperative pathology. The apparent diffusion coefficient (ADC), pure molecular diffusion (D), pseudo-diffusion coefficient (D∗), and perfusion fraction (f) were calculated from the whole tumour (__whole) and tumour margin (__margin). All parameters were compared between LVSI(+) and LVSI(-) and between LNM(+) and LNM(-). Logistic regression analysis and receiver operating characteristic (ROC) curve analysis were performed to evaluate the diagnostic performance of these parameters.

RESULTS

f__margin and D∗__whole showed significant differences in differentiating LVSI(+) from LVSI(-) tumours (p=0.002, 0.008, respectively), while LNM(+) tumours presented with significantly higher ADC__margin than that of LNM(-) tumours (p=0.009). The other parameters were not independent related factors with the status of LVSI or LNM according to logistic regression analysis (p>0.05). The area under the ROC curve of f__margin combined with D∗__whole in discriminating LVSI(+) from LVSI(-) was 0.826 (95% confidence interval [CI]: 0.691-0.961), while ADC__margin in differentiating LNM(+) from LNM(-) was 0.788 (95% CI: 0.648-0.928).

CONCLUSIONS

The parameters generated from mono-exponential and bi-exponential DWI of the primary cervical carcinoma could help discriminate its status regarding LVSI (f__margin and D∗__whole) and LNM (ADC__margin).

SAT0564 BONE TEXTURE ANALYSIS WITH DEEP LEARNING IN HAND RADIOGRAPHS FOR ASSESSING THE RISK OF RHEUMATOID ARTHRITIS

Background:

Conventional x-rays are essential to identify radiographic changes of rheumatoid arthritis (RA) in structure and bone texture. Limited evidence suggests that the bone texture analysis may quantify the radiographic changes in RA;1however, current techniques such as the fractal dimension characterize fixed texture features. Deep learning offers novel methods to ‘learn’ radiographic texture features relevant to RA.

Objectives:

To develop a deep learning model to assess the radiographic bone texture in the distal metacarpal bone relevant to RA.

Methods:

We collected 3,738 conventional hand radiographs from 2,128 individuals (RA, n = 908; non-RA, n = 1220). The second, third, and fourth metacarpal bone images were segmented using a curve Graph Convolutional Network (GCN), and the distal third was used as the input to train a texture model to classify RA. The texture model was based on the Deep Texture Encoding Network (Deep-TEN) architecture (figure 1),2which put an encoding layer on top of a pre-trained 18-layered residual network (ResNet18). The vectors produced by the model represent the orderless texture features that were used to generate a texture score for RA. Five texture models are trained using 5-fold cross-validation and are ensembled during inference by averaging the model outputs to produce the final score. We then validate the model using hand radiographs of 166 RA patients and 166 non-RA patients. Overall model performance was measured by area under the curve of the receiver operator curve (AUROC). Multivariate logistic regression was used to estimate the odds ratio (OR) and 95% confidence interval (CI) of RA.

Figure 1.

Schematic representation of deep learning models to extract and encode texture features for RA classification.

Results:

We included 140 women and 26 men with RA (mean age, 55.9±1.8 years) and 166 non-RA individuals (F: M, 140:26; mean age, 55.5 ± 1.8 years). The mean texture score was 0.49 (95% CI, 0.48–0.50) in RA patients, which is significantly higher than non-RA patients (0.42, 95% CI, 0.40–0.43; p<0.01). The AUROC of the model was 0.68. In the multivariate logistic regression model, a high texture score (>0.43) is associated with an OR (95% CI) of 3.42 (2.48–4.72) for RA, adjusted by age and sex.

Conclusion:

This study indicates that the texture model can delineate radiographic changes in texture relevant to RA and, coupled with automatic joint detection and segmentation, it has the potential to aid early RA diagnosis and monitor radiographic progression.

References:

[1]Zandieh S, Haller J, Bernt R, et al. Fractal analysis of subchondral bone changes of the hand in rheumatoid arthritis. Medicine (Baltimore) 2017;96(11):e6344.

[2]Zhang H, Xue J, Dana K. Deep TEN: Texture Encoding Network. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017:708-17.

Disclosure of Interests:

None declared

OP0301 PREDICTION OF LOW BONE MINERAL DENSITY AND FRAX SCORE BY ASSESSING HIP BONE TEXTURE WITH DEEP LEARNING

Background:

Osteoporosis is a widespread health concern associated with an increased risk of fractures in individuals with low bone mineral density (BMD). Dual-energy x-ray absorptiometry (DXA) is the gold standard to measure BMD, but methods based on the assessment of plain films, such as the digital radiogrammetry,1are also available. We describe a novel approach based on the assessment of hip texture with deep learning to estimate BMD.

Objectives:

To compare the BMD estimated by assessing hip texture using a deep learning model and that measured by DXA.

Methods:

In this study, we identified 1,203 patients who underwent DXA of left hip and hip plain film within six months. The dataset was split into a training set with 1,024 patients and a testing set with 179 patients. Hip images were obtained and regions of interest (ROI) around left hips were segmented using a tool based on the curve Graph Convolutional Network. The ROIs are processed using a Deep Texture Encoding Network (Deep-TEN) model,2which comprises the first 3 blocks of Residual Network with 18 layers (ResNet-18) model followed by a dictionary encoding operator (Figure 1). The encoded features are processed using a fully connected layer to estimate BMD. Five-fold cross-validation was conducted. Pearson’s correlation coefficient was used to assess the correlation between predicted and reference BMD. We also test the performance of the model to identify osteoporosis (T-score ≤ -2.5)

Figure 1.

Schematic representation of deep learning models to extract and encode texture features for estimation of hip bone density.

Results:

We included 151 women and 18 men in the testing dataset (mean age, 66.1 ± 1.7 years). The mean predicted BMD was 0.724 g/cm2compared with the mean BMD measured by DXA of 0.725 g/cm2(p = 0.51). Pearson’s correlation coefficient between predicted and true BMD was 0.88. The performance of the model to detect osteoporosis/osteopenia was shown in Table 1. The positive predictive value was 87.46% for a T-score ≤ -1 and 83.3% for a T-score ≤ -2.5. Furthermore, the mean FRAX® 10-year major fracture risk did not differ significantly between scores based on predicted (6.86%) and measured BMD (7.67%, p=0.52). The 10-year probability of hip fracture was lower in the predicted score (1.79%) than the measured score (2.43%, p = 0.01).

Table 1.

Performance matrices of the deep texture model to detect osteoporosis/osteopenia

T-score ≤ -1T-score ≤ -2.5Sensitivity91.11%(95% CI, 83.23% to 96.08%)33.33%(95% CI, 17.29% to 52.81%)Specificity86.08%(95% CI, 76.45% to 92.84%)98.56%(95% CI, 94.90% to 99.83%)Positive predictive value88.17%(95% CI, 81.10% to 92.83%)83.33%(95% CI, 53.58% to 95.59%)Negative predictive value89.47%(95% CI, 81.35% to 94.31%)87.26%(95% CI, 84.16% to 89.83%)

Conclusion:

This study demonstrates the potential of the bone texture model to detect osteoporosis and to predict the FRAX score using plain hip radiographs.

References:

[1]Zandieh S, Haller J, Bernt R, et al. Fractal analysis of subchondral bone changes of the hand in rheumatoid arthritis. Medicine (Baltimore) 2017;96(11):e6344.

[2]Zhang H, Xue J, Dana K. Deep TEN: Texture Encoding Network. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017:708-17.

Disclosure of Interests:

None declared

OP0062 PREDICTIVE VALUE OF BONE TEXTURE FEATURES EXTRACTED BY DEEP LEARNING MODELS FOR THE DETECTION OF OSTEOARTHRITIS: DATA FROM THE OSTEOARTHRITIS INITIATIVE

Background:

Osteoarthritis is a degenerative disorder characterized by radiographic features of asymmetric loss of joint space, subchondral sclerosis, and osteophyte formation. Conventional plain films are essential to detect structural changes in osteoarthritis. Recent evidence suggests that fractal- and entropy-based bone texture parameters may improve the prediction of radiographic osteoarthritis.1In contrast to the fixed texture features, deep learning models allow the comprehensive texture feature extraction and recognition relevant to osteoarthritis.

Objectives:

To assess the predictive value of deep learning-extracted bone texture features in the detection of radiographic osteoarthritis.

Methods:

We used data from the Osteoarthritis Initiative, which is a longitudinal study with 4,796 patients followed up and assessed for osteoarthritis. We used a training set of 25,978 images from 3,086 patients to develop the textual model. We use the BoneFinder software2to do the segmentation of distal femur and proximal tibia. We used the Deep Texture Encoding Network (Deep-TEN)3to encode the bone texture features into a vector, which is fed to a 5-way linear classifier for Kellgren and Lawrence grading for osteoarthritis classification. We also developed a Residual Network with 18 layers (ResNet18) for comparison since it deals with contours as well. Spearman’s correlation coefficient was used to assess the correlation between predicted and reference KL grades. We also test the performance of the model to identify osteoarthritis (KL grade≥2).

Results:

We obtained 6,490 knee radiographs from 446 female and 326 male patients who were not in the training sets to validate the performance of the models. The distribution of the KL grades in the training and testing sets were shown in Table 1. The Spearman’s correlation coefficient was 0.60 for the Deep-TEN and 0.67 for the ResNet18 model. Table 2 shows the performance of the models to detect osteoarthritis. The positive predictive value for Deep-TEN and ResNet18 model classification for OA was 81.37% and 87.46%, respectively.

Table 1

Distribution of KL grades in the training and testing sets.

KL grades01234TotalTraining set1089341.9%458218.7%611423.5%332012.8%7993.1%25,978Testing set247238.1%135320.8%169626.1%77511.9%1943.0%6,490Table 2

Performance matrices of the Deep-Ten and ResNet18 models to detect osteoarthritis

Deep-TENResNet18Sensitivity62.29%(95% CI, 60.42%–64.13%)59.14%(95% CI, 57.24%–61.01%)Specificity90.07%(95% CI, 89.07%–91.00%)94.09%(95% CI, 93.30%–94.82%)Positive predictive value81.37%(95% CI, 79.81%–82.84%)87.46%(95% CI, 85.96%–88.82%)Negative predictive value77.42%(95% CI, 77.64%–79.65%)76.77%(95% CI, 75.93%–77.59%)

Conclusion:

This study demonstrates that the bone texture model performs reasonably well to detect radiographic osteoarthritis with a similar performance to the bone contour model.

References:

[1]Bertalan Z, Ljuhar R, Norman B, et al. Combining fractal- and entropy-based bone texture analysis for the prediction of osteoarthritis: data from the multicenter osteoarthritis study (MOST). Osteoarthritis Cartilage 2018;26:S49.

[2]Lindner C, Wang CW, Huang CT, et al. Fully Automatic System for Accurate Localisation and Analysis of Cephalometric Landmarks in Lateral Cephalograms. Sci Rep 2016;6:33581.

[3]Zhang H, Xue J, Dana K. Deep TEN: Texture Encoding Network. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017:708-17.

Disclosure of Interests:

None declared

0659 A Safe and Precise Tongue Base Surgery for Obstructive Sleep Apnea: Real-Time Intraoperative Ultrasound-Assisted Transoral Robotic Surgery

Introduction

Lingual artery (LA) injury is a devastating complication of tongue base surgery. Compared with the anatomic findings of computed tomography angiography (CTA), intraoperative blade of mouth gag might change the thickness of base of tongue (BOT) and anatomy of LA. We aimed to investigate the position of LA in the BOT with intraoperative ultrasound (IOU) imaging during transoral robotic surgery (TORS), and evaluate the bleeding complications when assisted with / without IOU.

Methods

Adult obstructive sleep apnea (OSA) patients who received TORS in BOT resection were recruited since 2016. Assessment tools were pre-op over-night hospital polysomnography (PSG) and anatomy-based Friedman Staging System. Ultrasound imaging was utilized to identify anatomic parameters of LA in BOT, including distance to midline, arterial depth and diameter.

Results

Ninety-three OSA patients (82 male, 88.2%) were analyzed. The mean age was 42.2±10.0 years old and body mass index was 29.2±4.5 kg/m2. The average apnea hypopnea index (AHI) was 58.1±21.4 events/hour. There were 66 (71.0%), 24 (25.8%) and 3 (3.2%) patients in Friedman stages II, III and IV, respectively. Seventy patients underwent TORS with IOU had shorter operation time (191.7±3.8 minutes) than 23 patients without IOU (220.1±6.6 minutes), less total blood loss (11.3±10.8 versus 19.6±26.7 ml), and more BOT tissue reduction volume (7.1±2.5 versus 3.9±1.6 ml). Significant predictors of arterial depth were higher AHI level during rapid-eye-movement (REM) sleep stage (p=0.038), bigger tonsil size (p=0.034) and more elevated Friedman tongue position (p=0.012). Postoperative complication associated with LA injury was not found in the patients with use of IOU.

Conclusion

When tongue retracted with blade, the distance to midline and depth of LA were altered in BOT. With IOU assisted, surgeon could identify LA position confidently. It is expectable to maximize efficiency and minimize catastrophic bleeding complications when OSA patients received TORS in BOT resection.

Support

nil

0032 Timing of Daily Rhythm of Cardiac Autonomic Control Contributes to Weight Loss Resistance, Independent of Daily Energy Intake and Physical Activity Level

Introduction

Obesity is a major health problem. Many treatments have been designed to help overweight/obese people to lose weight, but their effectiveness is highly variable. The same treatments may work for some persons while others have no responses — weight loss resistance. We tested whether the daily rhythm of cardiac autonomic control contributes to weight loss resistance.

Methods

We studied 39 overweight/obese Caucasian women (BMI&gt;25; age: 21–62 years old) who completed (1) an obesity dietary treatment of up to 30 weeks with weekly assessments of body weight, and (2) ambulatory monitoring of electrocardiogram (ECG) for up to 3.5 days. Heartbeat intervals were derived from ECG. Cardiac autonomic control was assessed in each 1-h bin by examining the temporal correlation in heartbeat fluctuations — a nonlinear measure that quantifies the delicate dynamic interplay between sympathetic and vagal outflows. Daily rhythm was estimated using the cosinor analysis.

Results

Weight loss was highly variable (range: 0.68%-21.78 % of initial body weight). The correlation in heartbeat fluctuations displayed a 24-h rhythm (p&lt;0.0001) with fewer correlations (more random) during the nighttime. The phase (peak timing) of the rhythm was highly variable, i.e., 10AM to 8PM for most participants, and after midnight in four participants. Weight loss evolution depended on the phase (p=0.006) in a nonlinear manner. Specifically, participants with the phase between 2PM-8PM lost weight faster than those with phases before 2PM and those after 8PM. The effect was independent of total energy intake, physical activity level, and sleep/wake schedules.

Conclusion

Cardiac autonomic control in overweight/obese women displayed a daily rhythm. The timing of the rhythm had previously un-identified contributions to weight loss. The inter-individual differences in the timing may reflect different circadian regulation of autonomic function and its interaction with the daily behavioral cycle.

Support

This work was supported by NIH grants R01AG048108, RF1AG059867, RF1AG064312, R01AG017917, and R01NS078009.

1027 Differences in Polysomnography-Based Sleep Disorders Between HIV-Infected Persons and Matched Controls

Introduction

Sleep disturbance is a prevalent problem among HIV-infected persons. The recognition of comorbid sleep disorders in patients with HIV is currently hampered by limited knowledge of sleep-related symptoms, sleep architecture, and types of sleep disorders. We aimed to compare the differences in sleep-related symptoms and polysomnography-based sleep disorders between HIV-infected persons and controls.

Methods

The study included 170 men with a Pittsburgh sleep quality index (PSQI) greater than 5, composed of 44 HIV-infected men and 126 male controls who were frequency-matched by sex, age (-/+ 3.0 years) and BMI (-/+ 3.0 kg/m2). For all participants an overnight sleep study using a Somte V1 monitor was conducted. Differences in sleep-related symptoms and sleep disorders between HIV-infected patients and controls were examined using t-tests or Chi-square tests.

Results

HIV-infected persons with sleep disturbances more often had psychological disturbances (72.7% vs. 40.5%, p&lt;0.001) and suspected rapid eye movement (REM) behavior disorder (RBD) (25.0% vs. 4.8%, p&lt;0.01) than that of controls. The sleep-disordered breathing (SDB) in HIV-infected persons was less common than that in controls (56.8% vs. 87.3%, p&lt;0.001). The mean percentage of REM sleep among HIV-infected patients was higher than that among the controls (20.6% vs. 16.6%, p&lt;0.001). Enuresis was more common in HIV-infected persons than controls (40.9% vs. 22.2%, p=0.02).

Conclusion

Psychological disturbances and SDB can be the possible explanations of sleep disturbances in HIV-infected persons, in which suspected RBD is notable. Further studies are warranted to examine underlying factors of suspected RBD among HIV-infected persons with sleep disturbances.

Support

This work was supported by the Ministry of Science and Technology, Executive Yuan of Taiwan [MOST 105-3011-E-006-002], and National Cheng Kung University Hospital [NCKUH-10702022]

0434 Clinical Application of Computer Aided Cloud Sleep Scoring System

Introduction

According to a survey by World Sleep Society, 45% of the population suffered from sleep disorders. The best way to diagnose these patients is to use Polysomnography (PSG), recording their physiological signals throughout the night. Mostly, sleep technologists manually score sleep stages. Manual scoring is quite subjective and time-consuming. Although the technologist’s judgments are based on scoring standards of the American Academy of Sleep Medicine, fine-tuning scoring results because of different considerations in different sleep centers may be happened. In order to assess the consistency of scoring standards in sleep technologists, we tried to establish a cloud sleep scoring system and evaluate its feasibility in 4 sleep centers in southern Taiwan.

Methods

We constructed a computer-aided cloud sleep scoring system. Each sleep technologist could score the same test data of PSG online without being restricted by places and hardware equipment. After comparing scoring results of all participants, the scoring system could provide the following reports, including an overall agreement, agreement of each sleep stage and each sleep index. Besides, multi-person scoring results of each epoch with displaying physiological signals were analyzed.

Results

Seven sleep technologists from 4 hospitals in Tainan, Taiwan joined this study. Standard deviations (SDs) of each sleep stage included 2.64 in Wake stage, 6.90 in N1, 8.31 in N2, 6.87 in N3, 1.38 in REM, respectively. SDs of sleep indexes were 2.64 in sleep efficiency, 2.14 in sleep onset time, 8.35 in wake after sleep onset time, 10.03 in total sleep time, individually. The overall agreement was 89.6%. The satisfaction of this scoring system operation was 85.7%.

Conclusion

With the cloud sleep scoring system assistance, it was feasible to evaluate the scoring consistency among sleep technologists in different sleep centers.

Support

This work is supported by the Ministry of Science and Technology, Taiwan. (MOST 108-2634-F-006-012)