Prognosis of patients in prolonged coma after severe carbon monoxide poisoning

BACKGROUND

Late recovery in patients following prolonged coma from carbon monoxide poisoning have been reported, but the probability is unclear. The purpose of this research was to assess the prognosis of patients in prolonged coma after severe carbon monoxide poisoning and related clinical and imaging features.

METHODS

There were 13 patients who had been in a state of coma for >7 days after acute carbon monoxide poisoning in the retrospective observational study, and demographic data, clinical data, laboratory data, complications, and image data were collected. Outcome was assessed by means of the Glasgow outcome scale after 1 year. The relationship between complications and imaging manifestations and prognosis was also analyzed.

RESULTS

One year after severe carbon monoxide poisoning, two patients (15.4%) had died (GOS 1), nine (69.2%) were in a persistent vegetative state (GOS 2), one (7.7%) was moderately disabled (GOS 4), and one (7.7%) achieved a good recovery (GOS 5) with minimal disability.

CONCLUSIONS

Most patients with prolonged coma after severe carbon monoxide poisoning had a poor prognosis, although the younger patients had a better prognosis. Respiratory failure, hypotension and renal failure during the course of the disease were associated with a poor prognosis. The prognosis of patients with injuries in two sites in early CT was poor. Multiple lesions (≥3) and extensive white matter damage (Fazekas grade (PVH or DWMH) = 3) on MRI of chronic phase were also associated with a poor prognosis.

[Association of blood lead level with cognition impairment among elderly aged 65 years and older in 9 longevity areas of China]

Objective: To investigate the association between blood lead concentrations and cognition impairment among Chinese older adults aged 65 or over. Method: Data was collected in 9 longevity areas from Heathy Aging and Biomarkers Cohort Study between 2017 and 2018. This study included 1 684 elderly aged 65 years and older. Information about demographic characteristics, socioeconomic factors, health status and cognitive function score of respondents were collected by questionnaire survey and physical examination. Venous blood of the subjects was collected to detect the blood lead concentration. Subjects were stratified into four groups (Q₁-Q₄) by quartile of blood lead concentration. Multivariate logistic regression model was used to analyze the association between blood lead concentration and cognitive impairment. The linear or non-linear association between blood lead concentration and cognitive impairment were described by restrictive cubic splines (RCS). Results: Among the 1 684 respondents, 843 (50.1%) were female and 191 (11.3%) suffered from cognition impairment. After adjusting for confounding factors, the OR value and 95%CI of cognition impairment was 1.05 (1.01-1.10) for every 10 μg/L increase in blood lead concentration in elderly; Compared with the elderly in Q₁, the elderly with higher blood lead concentration had an increased risk of cognitive impairment. The OR value and 95%CI of Q₂, Q₃ and Q₄ groups were 1.19 (0.69-2.05), 1.45 (0.84-2.51) and 1.92 (1.13-3.27), respectively. Conclusion: Higher blood lead concentration is associated with cognitive impairment among the elderly aged 65 years and older in 9 longevity areas in China.

[Association of blood oxidative stress level with hypertriglyceridemia in the elderly aged 65 years and older in 9 longevity areas of China]

Objective: To investigate the association of blood oxidative stress level with hypertriglyceridemia in the elderly aged 65 years and older in China. Methods: A total of 2 393 participants aged 65 years and older were recruited in 9 longevity areas from Heathy Aging and Biomarkers Cohort Study, during 2017 to 2018. Information on demographics characteristic, life style and health status were collected by questionnaire and physical examination, and venous blood was collected to detect the levels of blood oxidative stress and hypertriglyceridemia. The linear or non-linear association between oxidative stress and hypertriglyceridemia was described by restrictive cubic splines (RCS) fitting multiple linear regression model. The generalized linear mixed effect model was conducted to assess the association between oxidative stress and hypertriglyceridemia. Results: A total of 2 393 participants, mean age was 84.6 years, the youngest was 65 and the oldest was 112, the male was 47.9%(1 145/2 393), the triglyceride level was (1.4±0.8) mmol/L. The hypertriglyceridemia detection rate was 9.99%(239/2 393). The results of multiple linear regression model with restrictive cubic spline fitting showed that MDA level was linear association with triglyceride level; SOD level was nonlinear association with triglyceride level. MDA level had significantly association with hypertriglyceridemia, and the corresponding OR value was 1.063 (95%CI: 1.046,1.081) with 1 nmol/ml increment of blood MDA; SOD level had significantly association with hypertriglyceridemia, and the corresponding OR value was 0.986(95%CI: 0.983,0.989) with 1 U/ml increment of blood SOD. Conclusion: Among the elderly aged 65 and older in 9 longevity areas in China, MDA and SOD levels were associated with the risk of hypertriglyceridemia.

First-phase ejection fraction predicts response to cardiac resynchronization therapy and adverse outcomes

Background

Cardiac Resynchronization Therapy (CRT) is an important therapeutic treatment for chronic heart failure. However, even in carefully selected cases up to 40% of patient fail to respond. First-phase ejection fraction (EF1), the ejection fraction up to the time of maximal ventricular contraction, is a novel and more sensitive echocardiographic measure of early systolic function.

Purpose

We examined the value of EF1, to predict response to CRT and clinical outcomes after CRT.

Methods

We analysed echocardiograms from 197 patients who underwent CRT between 2009 and 2018 and were followed to determine clinical outcomes.

Results

Volumetric response rate (reduction in end-systolic volume ≥15%) was 92.3% vs. 12.1%, for those with EF1 in the highest vs. lowest tertiles (p<0.001). A cut-off value of 11.9% for EF1 had >85% sensitivity and specificity for prediction of response to CRT; on multivariate logistic regression analysis incorporating previously defined predictors, EF1 was the strongest predictor of response (OR: 1.563, 95% CI: 1.371–1.782, p<0.001) (table 1). EF1 was also the strongest predictor of a clinical composite score (OR: 1.115, 95% CI: 1.044–1.191, p=0.001). Improvement in EF1 at 6 months after CRT implantation (6.5±5.8% vs 1.8±4.3% in responders vs. non-responders, p<0.001) was the best predictor of heart failure re-hospitalization and death after median follow-up of 29.2 months (HR: 0.819, 95% CI:0.765–0.876, p<0.001).

Conclusion

EF1 is a promising marker to identify patients likely to respond to CRT and most likely provides a measure of myocardial viability that determines response.

Funding Acknowledgement

Type of funding source: Foundation. Main funding source(s): British Heart Foundation, Wellcome/EPSRC Centre for Medical Engineering

A joint analysis strategy reveals genetic changes associated with artificial selection between egg-type and meat-type ducks

Egg-type ducks and meat-type ducks are predominantly commercial or indigenous and have been subjected to artificial directional selection. These two duck types differ substantially in body shape, production performance and reproductivity. However, the genetic changes associated with phenotypic differences remain unclear. Here, we compared the two duck types at the genomic and transcriptomic levels. We identified a large number of SNPs and genes in genomic divergent regions in terms of F_ST and θπ values. The corresponding genes were mainly enriched in embryonic development function and metabolic pathway. RNA-seq analysis also revealed differential gene expression in the liver and gonads. The differentially expressed genes were functionally associated with signal transmission and substance metabolism respectively. Furthermore, we found that seven genes were related to differentiation between the two types by both g genome and transcriptome analysis and were plausible candidate genes. These genes were annotated to GO categories of cell development and disease immunity. These findings will enable a better understanding of the artificial selection history of meat and egg ducks and provide a valuable resource for future research on the breeding of these two lineages.

3D Bioinspired Microstructures for Switchable Repellency in both Air and Liquid

In addition to superhydrophobicity/superoleophobicity, surfaces with switchable water/oil repellency have also aroused considerable attention because of their potential values in microreactors, sensors, and microfluidics. Nevertheless, almost all those as-prepared surfaces are only applicable for liquids with higher surface tension (γ > 25.0 mN m-1) in air. In this work, inspired by some natural models, such as lotus leaf, springtail skin, and filefish skin, switchable repellency for liquids (γ = 12.0-72.8 mN m-1) in both air and liquid is realized via employing 3D deformable multiply re-entrant microstructures. Herein, the microstructures are fabricated by a two-photon polymerization based 3D printing technique and the reversible deformation is elaborately tuned by evaporation-induced bending and immersion-induced fast recovery (within 30 s). Based on 3D controlled microstructural architectures, this work offers an insightful explanation of repellency/penetration behavior at any three-phase interface and starts some novel ideas for manipulating opposite repellency by designing/fabricating stimuli-responsive microstructures.

LncRNA KCNQ1OT1 delayed fracture healing through the Wnt/β-catenin pathway

OBJECTIVE

The aim of this study was to explore the effect of long non-coding ribonucleic acid (lncRNA) KCNQ1 overlapping transcript 1 (KCNQ1OT1) on fracture healing and its possible mechanism.

PATIENTS AND METHODS

Abnormal lncRNAs were compared between patients with delayed fracture healing and those with normal fracture healing using gene expression profiling method. LncRNA expression in patients was verified by quantitative Reverse Transcription-Polymerase Chain Reaction (qRT-PCR). Subsequently, the model of tibial fracture was successfully established in rabbits. The effect of lncRNA KCNQ1OT1 expression on tibial fracture healing in rabbits was explored. Meanwhile, the effects of lncRNA KCNQ1OT1 on cell proliferation and apoptosis were investigated by knockdown and overexpression experiments with HC-a as a cell model. Furthermore, Western blotting was used to explore the expressions of proteins in signaling pathway affected by lncRNA KCNQ1OT1.

RESULTS

Gene expression profiling and qRT-PCR revealed that lncRNA KCNQ1OT1 was significantly down-regulated in bone tissues of patients with delayed fracture healing. Compared with the control group, knocking down lncRNA KCNQ1OT1 remarkably reduced the serum levels of alkaline phosphatase (ALP) and osteoprotegerin (OPG) in rabbits, and markedly decreased bone trabecular growth index (p<0.05). In HC-a cells, overexpression of lncRNA KCNQ1OT1 activated the Wnt/β-catenin signaling pathway, which could be suppressed by knocking down lncRNA KCNQ1OT1. Cell Counting Kit-8 (CCK-8) assay and 5(6)-carboxyfluorescein diacetate, succinimidyl ester (CFSE) results manifested that lncRNA KCNQ1OT1 remarkably promoted the proliferation and inhibited apoptosis of HC-a cells by activating the Wnt/β-catenin signaling pathway.

CONCLUSIONS

LncRNA KCNQ1OT1 plays a vital role in delayed fracture healing. Moreover, it can induce cell proliferation and inhibit cell apoptosis by activating the Wnt/β-catenin signaling pathway. Therefore, KCNQ1OT1 may be used as a biomarker to predict the occurrence of delayed fracture healing.

[Risk factors for death and the clinical features of different subtypes of patients with pulmonary arterial hypertension related to congenital heart disease]

Objective: To explore the risk factors for death in patients with pulmonary arterial hypertension related to congenital heart disease (PAH-CHD) and the clinical characteristics of different subtypes in patients with PAH-CHD. Methods: It was a retrospective study. A total of 507 PAH-CHD patients, who were hospitalized in the Department of Pediatric Cardiology of Beijing Anzhen Hospital between September 2005 and May 2019, were included. Patients were divided into 4 subgroups: (1) Eisenmenger syndrome(ES) group. (2) PAH associated with prevalent systemic-to-pulmonary shunts(SP) group. (3) PAH associated with small defects(SD) group. (4) PAH after defect correction(CD) group. According to the complexity of cardiac malformation, patients were divided into simple-medium complex malformation group and complex malformation group. According to the location of shunts, patients were divided into pre-tricuspid group, post-tricuspid group, and mixed group or complex deformity group. Baseline clinical data of patients were collected from the electronic medical record system. Demographic data (age, gender, etc.), percutaneous oxygen saturation(SPO(2)), New York Heart Association(NYHA) cardiac function classification, 6 minutes walking distance(6MWD), and B type natriuretic peptide(BNP), systolic pulmonary arterial pressure(sPAP) estimated by echocardiography and mean pulmonary artery pressure (mPAP), mean right atrial pressure(mRAP), cardiac index(CI), and calculated pulmonary vascular resistance (PVR) estimated by right heart catheterization were compared among various groups. The results of regular follow-up of all enrolled patients were collected, including the status of monotherapy or combination of PAH-targeted drugs during the follow-up period, cardiac-related adverse events(hemopysis, syncope, edema, arrhythmia, etc.) and primary endpoint event(all-cause death) were obtained and analyzed. Risk factors for all-cause death were analyzed using univariate and multivariate Cox regression analysis model. Results: The median age at diagnosis was 23.1(13.9,32.1) years, 345 cases(68.0%) were female. Two hunderds and thirty-five cases(46.4%) were diagnosed with ES; 193 cases(38.1%) were diagnosed with CD, 47 cases (9.3%) were diagnosed with SD. Among them, 32 cases(6.3%) were in the SP group. All 507 patients underwent echocardiography examination, there were significant differences in sPAP among different clinical subgroups(P<0.001). A total of 289 patients(57.0%) received right heart catheterization examination, the results showed that the ES group had the highest mPAP and PVR and the lowest mRAP(all P<0.001), the CD group had the highest mRAP and CI(both P<0.001). The 6MWD in the ES group was significantly shorter than that in the SP, SD, and CD groups(all P<0.001). The proportion of patients with NYHA class Ⅲ/Ⅳ was higher in SD group than in SP group(P<0.001), which was similar between SD, ES and CD groups (P values were 0.077 and 0.072, respectively). At admission, the proportion of patients with NYHA class Ⅰ/Ⅱwas the highest in SP group(96.9% (31/32) ), followed by CD group (85.5%(165/193)) and the ES group(85.1%(200/235)), and the SD group(75.0%(35/47)). The BNP level at admission was also higher in SD group than in SP, ES and CD groups(P<0.001). Of the 507 patients, 379(74.8%) patients received PAH-targeted drug therapy at the last follow-up, and the treatment plan was mainly monotherapy(75.7%(287/379)). The median follow-up time was 3.6(2.0, 5.6) years and 37(7.3%) patients died, including 13 in the CD group, 17 in the ES group, and 7 in the SD group. No deaths occurred in the SP group. Right heart failure was the most common cause of death(11(29.7%)), followed by severe hemoptysis dyspnea(7(18.9%)), sudden cardiac death(6(16.2%)), and pulmonary hypertensive crisis(4(10.8%)). Kaplan-Meier curve showed that survival rates of end-point-free events at 1, 3, 5 and 10 years after diagnosis of PAH were 98.0%, 95.4%, 89.9%, and 84.4%, respectively; there were statistically significant differences in survival among the subgroups(P=0.026); there was no significant difference in the survival rate between the ES group and the CD group(P=0.918), and both were higher than the SD group(P values were 0.011 and 0.013, respectively). Univariate Cox regression analysis showed that NYHA class Ⅲ/Ⅳ and BNP>100 ng/L at admission were the risk factors for all-cause death in patients with PAH-CHD(HR=6.452, 95%CI 3.378-12.346, P<0.001, and HR=2.481, 95%CI 1.225-5.025, P=0.012). Multivariate Cox regression analysis showed that NYHA class Ⅲ/Ⅳ was an independent risk factor for all-cause death in patients with PAH-CHD(HR=4.998, 95%CI 1.246-20.055, P=0.023). Conclusions: PAH-CHD patients with different clinical subtypes have different clinical symptoms, cardiac functional class, hemodynamic characteristics, and mid to long-term survival rates. SP patients have the best prognosis, outcome of ES and CD patients is similar, and SD patients have the worst prognosis. NYHA class Ⅲ/Ⅳ is an independent risk factor for all-cause death in patients with PAH-CHD.

[Characteristics of rare diseases in Zhejiang province, 2007-2017]

Objective: To understand the characteristics of 24 388 inpatients with rare diseases in Zhejiang province during 2007-2017 and provide evidence for rare disease prevention and control. Methods: Inpatient data of rare diseases and the number of hospitalization in each year were collected in 10 hospitals of class Ⅲ (A) in Zhejiang province from 2007 to 2017, and descriptive statistical analysis was used. Results: A total of 24 388 cases of rare diseases were found, accounting for 2.69‰ (24 388/9 054 201) of total hospitalized cases. The top 3 types of rare diseases were "diseases of blood and blood-forming organs and certain disorders involving immune mechanism" (32.81%, 8 001/24 388), "congenital malformations, deformations and chromosomal abnormalities" (24.87%, 6 065/24 388) and "diseases of the nervous system" (19.01%, 4 635/24 388). The number of rare disease cases increased year by year from 2007 to 2017 with an average annual growth of 19.69%, however, the proportion of rare disease cases in the annual number of hospitalized cases only showed upward trend during 2016-2017, the time distribution of different types of rare diseases had different characteristics. The male to female ratio of rare diseases cases was 1.35∶1(13 990/10 398), "diseases of the digestive system" (4.45∶1, 1 180/265), "consequences of injury, poisoning and other external causes" (3.51∶1, 281/80) and "diseases of the nervous system" (2.26∶1, 3 213/1 422) had the highest male to female ratio. The distributions of rare disease types and diseases of different types in different age groups varied. The top 10 rare diseases accounted for 53.55% (13 060/24 388) of the total cases, and the top 3 diseases were adult idiopathic neutropenia (14.41%, 3 515/24 388), corticobasal degeneration (7.60%, 1 854/24 388) and henock-schoenlein purpura (6.01%, 1 466/24 388). Conclusion: The analysis on the characteristics of 24 388 rare disease cases in Zhejiang during 2007-2017 provided reference evidence for the promotion of rare disease research, monitoring, building registration database, and development of the prevention and control strategy for rare diseases in China.

WHITE STRIPE LEAF8, encoding a deoxyribonucleoside kinase, is involved in chloroplast development in rice

WSL8 encoding a deoxyribonucleoside kinase (dNK) that catalyzes the first step in the salvage pathway of nucleotide synthesis plays an important role in early chloroplast development in rice. The chloroplast is an organelle that converts light energy into chemical energy; therefore, the normal differentiation and development of chloroplast are pivotal for plant survival. Deoxyribonucleoside kinases (dNKs) play an important role in the salvage pathway of nucleotides. However, the relationship between dNKs and chloroplast development remains elusive. Here, we identified a white stripe leaf 8 (wsl8) mutant that exhibited a white stripe leaf phenotype at seedling stage (before the four-leaf stage). The mutant showed a significantly lower chlorophyll content and defective chloroplast morphology, whereas higher reactive oxygen species than the wild type. As the leaf developed, the chlorotic mutant plants gradually turned green, accompanied by the restoration in chlorophyll accumulation and chloroplast ultrastructure. Map-based cloning revealed that WSL8 encodes a dNK on chromosome 5. Compared with the wild type, a C-to-G single base substitution occurred in the wsl8 mutant, which caused a missense mutation (Leu 349 Val) and significantly reduced dNK enzyme activity. A subcellular localization experiment showed the WSL8 protein was targeted in the chloroplast and its transcripts were expressed in various tissues, with more abundance in young leaves and nodes. Ribosome and RNA-sequencing analysis indicated that some components and genes related to ribosome biosynthesis were down-regulated in the mutant. An exogenous feeding experiment suggested that the WSL8 performed the enzymic activity of thymidine kinase, especially functioning in the salvage synthesis of thymidine monophosphate. Our results highlight that the salvage pathway mediated by the dNK is essential for early chloroplast development in rice.

563 Early systolic function is impaired in patients with heart failure and preserved ejection fraction and is a more important predictor of clinical outcomes than diastolic function

Funding Acknowledgements

British Heart Foundation

Aims

First-phase ejection fraction (EF1), the ejection fraction up to the time of maximal ventricular contraction may be impaired in heart failure with preserved ejection fraction (HFpEF), with homeostatic mechanisms preserving the overall ejection fraction. We examined whether EF1 is impaired and whether it predicts adverse outcomes in patients with HFpEF.

Methods and Results

We examined EF1 in patients with HFpEF diagnosed according to European Society of Cardiology guidelines, in subjects with pre-clinical diastolic dysfunction and in control subjects with no evidence of heart failure. The predictive value of EF1 for a combined end-point of re-hospitalisation for heart failure and death from any cause was compared to that of conventional echocardiographic and other indices. There was a progressive impairment of EF1 in patients with diastolic dysfunction and HFpEF compared to those without evidence of heart failure (P &lt; 0.001). In 180 HFpEF patients followed for a median of 14.7 months, 101 patients reached the primary endpoint (61 deaths and 40 hospitalisations). EF1 was the most powerful predictor of events (table 1). A cut-off value of 19.4% gave hazard ratios (for EF1 &lt; 19.4% compared to ≥19.4%) of 3.04 (figure 1), (95% confidence interval 2.014 – 4.604, P &lt; 0.001) unadjusted, and 2.976 (1.887 – 4.695, P &lt; 0.001) adjusted for age, gender, other echocardiographic indices (including EF, E/e’ ratio, stroke volume and left atrial volume index) and N-terminal pro-brain natriuretic peptide.

Conclusion

Early systolic function is impaired in HFpEF and is a powerful predictor of clinical outcomes. Therapies targeted at improving early systolic function may improve outcomes in HFpEF.

Table 1 HR CI (95%) P value Age 1.020 1.001 - 1.039 0.042 Female Gender 1.989 1.254 - 3.156 0.003 ln NT-proBNP 1.354 1.096 - 1.672 0.005 Haemoglobin 0.992 0.980 - 1.005 0.225 eGFR 0.997 0.988 - 1.006 0.568 Structural Heart Disease 0.748 0.439 - 1.274 0.285 Diastolic Dysfunction 0.694 0.445 - 1.082 0.107 Ejection Fraction 1.031 0.996 - 1.068 0.082 Stroke Volume index 0.963 0.940 - 0.986 0.002 TPAVF 0.995 0.980 - 1.005 0.230 EF1 0.900 0.869 - 0.932 &lt;0.001 Multivariate analysis of predictors of events

Abstract 563 Figure 1 Kaplan-Meier Curve

Increased expression of miR-330-3p: a novel independent indicator of poor prognosis in human breast cancer

OBJECTIVE

Previous study has reported that miR-330-3p was highly expressed in breast cancer (BC) patients. However, the effect of miR-330-3p in BC progression remains largely unclear. The purpose of this study was to investigate the clinical significance of miR-330-3p expression in BC.

PATIENTS AND METHODS

The expression of miR-330-3p was detected by quantitative Real-time PCR in BC tissues and matched normal breast tissues. The association of miR-330-3p expression with clinicopathological factors of BC patients was also analyzed by x2-test. Prognosis value of patients with BC was assessed by Kaplan-Meier method and Cox proportional hazards model, respectively.

RESULTS

Quantitative real-time PCR analysis showed that the expression level of miR-330-3p was significantly higher in BC specimens than that in corresponding noncancerous tissues (p < 0.01). The levels of miR-330-3p were positively correlated with the status of TNM stage (p = 0.011) and lymph node metastasis (p = 0.006). Kaplan-Meier analysis revealed that 5-year overall survival of BC patients with high miR-330-3p expression was shorter compared to those patients with low miR-330-3p expression (p < 0.0001). Moreover, univariate and multivariate regression analysis demonstrated that miR-330-3p was an independent prognostic factor in BC.

CONCLUSIONS

Our data suggest that miR-330-3p upregulation maybe concurrently associated with prognosis in patients with BC, suggesting that miR-330-3p may be a potential prognostic biomarker and therapeutic target for patients with BC.

Whole exome sequencing identified a second pathogenic variant in HOMER2 for autosomal dominant non-syndromic deafness

Hearing loss is one of the most common sensory disorders worldwide, and about half of all occurrences are attributable to genetic factors. Here, we have identified a novel pathogenic variant in HOMER2 in a Chinese family with autosomal dominant, non-syndromic hearing loss. This is the second family reported globally with hearing loss caused by a variant in HOMER2. The pathogenic variant c.840_841insC in HOMER2 (NM_199330), segregating with the hearing-loss phenotype in the family, leads to a premature stop codon producing a truncated protein. The coiled-coil domain in the C-terminal of HOMER2 protein is essential for protein multimerization and HOMER2-CDC42 interaction. We compared the phenotypes in the two families and found that hearing impairment in this Chinese family was more severe. Furthermore, we found that the ability of this insertion mutant type HOMER2 (HOMER2_MU ) to multimerize decreased more significantly than wild-type HOMER2 (HOMER2_WT ) and the reported c.554G>C (NM_004839) mutant HOMER2. HOMER2_MU protein tended to be distributed in a diffuse manner, whereas HOMER2_WT and the reported mutant HOMER2 tended to cluster together. Our research provides a validating second family for variants in HOMER2 causing non-syndromic sensorineural hearing loss. HOMER2 homo-/hetero-multimerization might be the first step in exerting its normal function.

Recognizing basal cell carcinoma on smartphone-captured digital histopathology images with a deep neural network

BACKGROUND

Pioneering effort has been made to facilitate the recognition of pathology in malignancies based on whole-slide images (WSIs) through deep learning approaches. It remains unclear whether we can accurately detect and locate basal cell carcinoma (BCC) using smartphone-captured images.

OBJECTIVES

To develop deep neural network frameworks for accurate BCC recognition and segmentation based on smartphone-captured microscopic ocular images (MOIs).

METHODS

We collected a total of 8046 MOIs, 6610 of which had binary classification labels and the other 1436 had pixelwise annotations. Meanwhile, 128 WSIs were collected for comparison. Two deep learning frameworks were created. The 'cascade' framework had a classification model for identifying hard cases (images with low prediction confidence) and a segmentation model for further in-depth analysis of the hard cases. The 'segmentation' framework directly segmented and classified all images. Sensitivity, specificity and area under the curve (AUC) were used to evaluate the overall performance of BCC recognition.

RESULTS

The MOI- and WSI-based models achieved comparable AUCs around 0·95. The 'cascade' framework achieved 0·93 sensitivity and 0·91 specificity. The 'segmentation' framework was more accurate but required more computational resources, achieving 0·97 sensitivity, 0·94 specificity and 0·987 AUC. The runtime of the 'segmentation' framework was 15·3 ± 3·9 s per image, whereas the 'cascade' framework took 4·1 ± 1·4 s. Additionally, the 'segmentation' framework achieved 0·863 mean intersection over union.

CONCLUSIONS

Based on the accessible MOIs via smartphone photography, we developed two deep learning frameworks for recognizing BCC pathology with high sensitivity and specificity. This work opens a new avenue for automatic BCC diagnosis in different clinical scenarios. What's already known about this topic? The diagnosis of basal cell carcinoma (BCC) is labour intensive due to the large number of images to be examined, especially when consecutive slide reading is needed in Mohs surgery. Deep learning approaches have demonstrated promising results on pathological image-related diagnostic tasks. Previous studies have focused on whole-slide images (WSIs) and leveraged classification on image patches for detecting and localizing breast cancer metastases. What does this study add? Instead of WSIs, microscopic ocular images (MOIs) photographed from microscope eyepieces using smartphone cameras were used to develop neural network models for recognizing BCC automatically. The MOI- and WSI-based models achieved comparable areas under the curve around 0·95. Two deep learning frameworks for recognizing BCC pathology were developed with high sensitivity and specificity. Recognizing BCC through a smartphone could be considered a future clinical choice.