Development and validation of a collaborative robotic platform based on monocular vision for oral surgery: an in vitro study.

PURPOSE: Surgical robots effectively improve the accuracy and safety of surgical procedures. Current optical-navigated oral surgical robots are typically developed based on binocular vision positioning systems, which are susceptible to factors including obscured visibility, limited workplace, and ambient light interference. Hence, the purpose of this study was to develop a lightweight robotic platform based on monocular vision for oral surgery that enhances the precision and efficiency of surgical procedures. METHODS: A monocular optical positioning system (MOPS) was applied to oral surgical robots, and a semi-autonomous robotic platform was developed utilizing monocular vision. A series of vitro experiments were designed to simulate dental implant procedures to evaluate the performance of optical positioning systems and assess the robotic system accuracy. The singular configuration detection and avoidance test, the collision detection and processing test, and the drilling test under slight movement were conducted to validate the safety of the robotic system. RESULTS: The position error and rotation error of MOPS were 0.0906 ± 0.0762 mm and 0.0158 ± 0.0069 degrees, respectively. The attitude angle of robotic arms calculated by the forward and inverse solutions was accurate. Additionally, the robot's surgical calibration point exhibited an average error of 0.42 mm, with a maximum error of 0.57 mm. Meanwhile, the robot system was capable of effectively avoiding singularities and demonstrating robust safety measures in the presence of minor patient movements and collisions during vitro experiment procedures. CONCLUSION: The results of this in vitro study demonstrate that the accuracy of MOPS meets clinical requirements, making it a promising alternative in the field of oral surgical robots. Further studies will be planned to make the monocular vision oral robot suitable for clinical application.

Elucidating the association of obstructive sleep apnea with brain structure and cognitive performance.

BACKGROUND: Obstructive sleep apnea (OSA) is a pervasive, chronic sleep-related respiratory condition that causes brain structural alterations and cognitive impairments. However, the causal association of OSA with brain morphology and cognitive performance has not been determined. METHODS: We conducted a two-sample bidirectional Mendelian randomization (MR) analysis to investigate the causal relationship between OSA and a range of neurocognitive characteristics, including brain cortical structure, brain subcortical structure, brain structural change across the lifespan, and cognitive performance. Summary-level GWAS data for OSA from the FinnGen consortium was used to identify genetically predicted OSA. Data regarding neurocognitive characteristics were obtained from published meta-analysis studies. Linkage disequilibrium score regression analysis was employed to reveal genetic correlations between OSA and related traits. RESULTS: Our MR study provided evidence that OSA was found to significantly increase the volume of the hippocampus (IVW ß (95% CI) = 158.997 (76.768 to 241.227), P = 1.51e-04), with no heterogeneity and pleiotropy detected. Nominally causal effects of OSA on brain structures, such as the thickness of the temporal pole with or without global weighted, amygdala structure change, and cerebellum white matter change covering lifespan, were observed. Bidirectional causal links were also detected between brain cortical structure, brain subcortical, cognitive performance, and OSA risk. LDSC regression analysis showed no significant correlation between OSA and hippocampus volume. CONCLUSIONS: Overall, we observed a positive association between genetically predicted OSA and hippocampus volume. These findings may provide new insights into the bidirectional links between OSA and neurocognitive features, including brain morphology and cognitive performance.

The Skeletal Stability of Combined Surgery First Approach and Clear Aligners in Skeletal Class III Malocclusion Correction: A Randomized Controlled Trial.

The surgery first approach (SFA) and clear aligners technique can address traditional treatment defects, such as prolonged waiting times for surgery and a less desirable facial appearance due to wire aligners. However, the curative effect of the combination remains uncertain. The randomized controlled study aimed to evaluate the skeletal stability of the SFA compared to the conventional orthodontic first approach (OFA), both of which were applied with clear aligners. A total of 74 participants were randomly allocated to two groups: the SFA group (experimental) and the OFA group (control). The skeletal deviation was calculated using reconstruction models from computed tomography scans taken immediately and 6 months after surgery. The largest median deviations were detected in the y-axis of the mandible for both two groups, separately 1.36 mm in the experimental group and 1.19 mm in the control group. Apart from the maxillary yaw dimension (p = 0.005), there were no significant differences between the two groups in terms of linear and angular deviation. The experimental group had an overall treatment time of 18.05 ± 2.53 months, while the control group took 22.83 ± 3.60 months (p < 0.05). Therefore, the combined surgery-first and clear aligners treatment can achieve comparable skeletal stability to the conventional approach, while also saving significant time.

Integrating Virtual Surgical Planning and 3D-Printed Tools with Iliac Bone Grafts for Orbital and Zygomatic Reconstruction in Hemifacial Microsomia Patients.

Hemifacial Microsomia (HFM) is the second most common congenital craniofacial malformation syndrome, and the complexity of HFM makes its treatment challenging. The present study aimed to introduce a new approach of utilization of virtual surgical planning (VSP) and 3D-printed surgical adjuncts for maxillofacial reconstruction. Five HFM patients were included in this study. All participants were provided with a full VSP, including the design of osteotomy lines, the design and fabrication of 3D-printed cutting guides, fixation plates, and titanium mesh for implantation. With the assistance of 3D-printed cutting guides and fixation plates, the orbital deformities were corrected, and a 3D-printed titanium mesh combined with iliac cancellous bone graft was applied to reconstruct the zygomatic arch. The surgical accuracy, effectiveness, and bone absorption rate were evaluated. All patients completed the entirely digital treatment process without experiencing severe complications. The surgical adjuncts were effective in aligning the movement of the bone segments with the surgical plan, resulting in mean 3D deviations (1.0681 ± 0.15 mm) and maximum 3D deviations (3.1127 ± 0.44 mm). The image fusion results showed that the patients' postoperative position of the maxilla, zygoma, and orbital rim was consistent with the virtual surgical plan, with only a slight increase in the area of bone grafting. The postoperative measurements showed significant improvement in the asymmetry indices of Er (AI of Er: from 17.91 ± 3.732 to 5.427 ± 1.389 mm, p = 0.0001) and FZ (AI of FZ: from 7.581 ± 1.435 to 4.070 ± 1.028 mm, p = 0.0009) points. In addition, the observed bone resorption rate at the 6-month follow-up across the five patients was 45.24% ± 3.13%. In conclusion, the application of VSP and 3D-printed surgical adjuncts demonstrates significant value in enhancing the precision and effectiveness of surgical treatments for HFM. A 3D-printed titanium mesh combined with iliac cancellous bone graft can be considered an ideal alternative for the reconstruction of the zygomatic arch.

Carbohydrazide-Assisted Morphology and Structure Controlling for Lead-Free Cs2AgBiBr6 Double Perovskite Solar Cells.

The stability and toxicity problems have haunted the development and applications of metal halide perovskite materials, for which the lead-free inorganic double perovskite Cs2AgBiBr6 has emerged as a promising substitute in recent years. However, poor film quality has severely limited its photovoltaic performance that could have been induced by some key factors such as high annealing temperature. Herein, we present a facile strategy to fabricate high-quality pinhole-free Cs2AgBiBr6 films with large grain sizes by introducing carbohydrazide (CBH) into the precursor. Detailed characterizations have shown that the carbonyl group (C═O) in CBH plays the critical role in coordinating with Ag+ and Bi3+ cations during the film formation process. As another consequence, the as-fabricated devices have exhibited significantly higher reproducibility for fabrication. By optimizing the amount of CBH, the power conversion efficiency (PCE) relatively increased 37 to 1.57%, which remained 95.0% in an ambient environment for a 1000-h test. Hopefully, this work could facilitate the current technologies in the exploration of high-performance lead-free perovskites such as Cs2AgBiBr6 and better understanding of the mechanism in the additive engineering as well.

Fatty acids metabolism affects the therapeutic effect of anti-PD-1/PD-L1 in tumor immune microenvironment in clear cell renal cell carcinoma.

BACKGROUND: Clear cell renal cell carcinoma (ccRCC) is a highly invasive and metastatic subtype of kidney malignancy and is correlated with metabolic reprogramming for adaptation to the tumor microenvironment comprising infiltrated immune cells and immunomodulatory molecules. The role of immune cells in the tumor microenvironment (TME) and their association with abnormal fatty acids metabolism in ccRCC remains poorly understood. METHOD: RNA-seq and clinical data of KIRC from The Cancer Genome Atlas (TCGA) and E-MTAB-1980 from the ArrayExpress dataset. The Nivolumab group and Everolimus group of the CheckMate 025 study, the Atezolizumab arm of IMmotion150 and the Atezolizumab plus Bevacizumab group of IMmotion151 cohort were obtained for subsequent analysis. After differential expression genes identification, the signature was constructed through univariate Cox proportional hazard regression and simultaneously the least absolute shrinkage and selection operator (Lasso) analysis and the predictive performance of our signature was assessed by using receiver operating characteristic (ROC), Kaplan-Meier (KM) survival analysis, nomogram, drug sensitivity analysis, immunotherapeutic effect analysis and enrichment analysis. Immunohistochemistry (IHC), qPCR and western blot were performed to measure related mRNA or protein expression. Biological features were evaluated by wound healing, cell migration and invasion assays and colony formation test and analyzed using coculture assay and flow cytometry. RESULTS: Twenty fatty acids metabolism-related mRNA signatures were constructed in TCGA and possessed a strong predictive performance demonstrated through time-dependent ROC and KM survival analysis. Notably, the high-risk group exhibited an impaired response to anti-PD-1/PD-L1 (Programmed death-1 receptor/Programmed death-1 receptor-ligand) therapy compared to the low-risk group. The overall levels of the immune score were higher in the high-risk group. Additionally, drug sensitivity analysis observed that the model could effectively predict efficacy and sensitivity to chemotherapy. Enrichment analysis revealed that the IL6-JAK-STAT3 signaling pathway was a major pathway. IL4I1 could promote ccRCC cells' malignant features through JAK1/STAT3 signaling pathway and M2-like macrophage polarization. CONCLUSION: The study elucidates that targeting fatty acids metabolism can affect the therapeutic effect of PD-1/PD-L1 in TME and related signal pathways. The model can effectively predict the response to several treatment options, underscoring its potential clinical utility.

Iron metabolism and ferroptosis in diabetic bone loss: from mechanism to therapy.

Osteoporosis, one of the most serious and common complications of diabetes, has affected the quality of life of a large number of people in recent years. Although there are many studies on the mechanism of diabetic osteoporosis, the information is still limited and there is no consensus. Recently, researchers have proven that osteoporosis induced by diabetes mellitus may be connected to an abnormal iron metabolism and ferroptosis inside cells under high glucose situations. However, there are no comprehensive reviews reported. Understanding these mechanisms has important implications for the development and treatment of diabetic osteoporosis. Therefore, this review elaborates on the changes in bones under high glucose conditions, the consequences of an elevated glucose microenvironment on the associated cells, the impact of high glucose conditions on the iron metabolism of the associated cells, and the signaling pathways of the cells that may contribute to diabetic bone loss in the presence of an abnormal iron metabolism. Lastly, we also elucidate and discuss the therapeutic targets of diabetic bone loss with relevant medications which provides some inspiration for its cure.

One-step synthesis of SiO2 nanomesh for antireflection and self-cleaning of solar cell.

SiO2 nanomaterials are widely used for antireflection and self-cleaning, but the preparation process is usually complex and time-consuming. Hence, we present a facile one-step synthesis of a hydrophobic two-dimensional SiO2 nanomesh by tuning the reaction temperature using dodecylamine as a catalyst. SiO2 nanomesh has the advantages of an adjustable refractive index, simple preparation process, and low cost, which affords both antireflection and self-cleaning functions for solar cells. Two types of perovskite solar cells were used to verify the stability and universality of the SiO2 nanomesh coatings. The antireflection effect of the SiO2 nanomesh is found to increase the current density of both perovskite solar cells fabricated at 500 °C and 150 °C, with the efficiency increased by 4.48% and 4.79%, respectively.

Crosstalk of necroptosis and pyroptosis defines tumor microenvironment characterization and predicts prognosis in clear cell renal carcinoma.

Pyroptosis and necroptosis are two recently identified forms of immunogenic cell death in the tumor microenvironment (TME), indicating a crucial involvement in tumor metastasis. However, the characteristics of necroptosis and pyroptosis that define tumor microenvironment and prognosis in ccRCC patients remain unknown. We systematically investigated the transcriptional variation and expression patterns of Necroptosis and Pyroptosis related genes (NPRGs). After screening the necroptosis-pyroptosis clusters, the potential functional annotation for clusters was explored by GSVA enrichment analysis. The Necroptosis-Pyroptosis Genes (NPG) scores were used for the prognosis model construction and validation. Then, the correlations of NPG score with clinical features, cancer stem cell (CSC) index, tumor mutation burden (TMB), TME, and Immune Checkpoint Genes (ICGs) were also individually explored to evaluate the prognosis predictive values in ccRCC. Microarray screenings identified 27 upregulated and 1 downregulated NPRGs. Ten overall survival associated NPRGs were filtered to construct the NPG prognostic model indicating a better prognostic signature for ccRCC patients with lower NPG scores (P< 0.001), which was verified using the external cohort. Univariate and multivariate analyses along with Kaplan-Meier survival analysis demonstrated that NPG score prognostic model could be applied as an independent prognostic factor, and AUC values of nomogram from 1- to 5- year overall survival with good agreement in calibration plots suggested that the proposed prognostic signature possessed good predictive capabilities in ccRCC. A high-/sNPG score is proven to be connected with tumor growth and immune-related biological processes, according to enriched GO, KEGG, and GSEA analyses. Comparing patients with a high-NPG score to those with a low-NPG score revealed significant differences in clinical characteristics, growth and recurrence of malignancies (CSC index), TME cell infiltration, and immunotherapeutic response (P< 0.005), potentially making the NPG score multifunctional in the clinical therapeutic setting. Furthermore, AIM2, CASP4, GSDMB, NOD2, and RBCK1 were also found to be highly expressed in ccRCC cell lines and tumor tissues, and GASP4 and GSDMB promote ccRCC cells' proliferation, migration, and invasion. This study firstly suggests that targeting the NPG score feature for TME characterization may lend novel insights into its clinical applications in the prognostic prediction of ccRCC.

Identification of a novel cuproptosis-related gene signature and integrative analyses in patients with lower-grade gliomas.

Background: Cuproptosis is a newly discovered unique non-apoptotic programmed cell death distinguished from known death mechanisms like ferroptosis, pyroptosis, and necroptosis. However, the prognostic value of cuproptosis and the correlation between cuproptosis and the tumor microenvironment (TME) in lower-grade gliomas (LGGs) remain unknown. Methods: In this study, we systematically investigated the genetic and transcriptional variation, prognostic value, and expression patterns of cuproptosis-related genes (CRGs). The CRG score was applied to quantify the cuproptosis subtypes. We then evaluated their values in the TME, prognostic prediction, and therapeutic responses in LGG. Lastly, we collected five paired LGG and matched normal adjacent tissue samples from Sun Yat-sen University Cancer Center (SYSUCC) to verify the expression of signature genes by quantitative real-time PCR (qRT-PCR) and Western blotting (WB). Results: Two distinct cuproptosis-related clusters were identified using consensus unsupervised clustering analysis. The correlation between multilayer CRG alterations with clinical characteristics, prognosis, and TME cell infiltration were observed. Then, a well-performed cuproptosis-related risk model (CRG score) was developed to predict LGG patients' prognosis, which was evaluated and validated in two external cohorts. We classified patients into high- and low-risk groups according to the CRG score and found that patients in the low-risk group showed significantly higher survival possibilities than those in the high-risk group (P<0.001). A high CRG score implies higher TME scores, more significant TME cell infiltration, and increased mutation burden. Meanwhile, the CRG score was significantly correlated with the cancer stem cell index, chemoradiotherapy sensitivity-related genes and immune checkpoint genes, and chemotherapeutic sensitivity, indicating the association with CRGs and treatment responses. Univariate and multivariate Cox regression analyses revealed that the CRG score was an independent prognostic predictor for LGG patients. Subsequently, a highly accurate predictive model was established for facilitating the clinical application of the CRG score, showing good predictive ability and calibration. Additionally, crucial CRGs were further validated by qRT-PCR and WB. Conclusion: Collectively, we demonstrated a comprehensive overview of CRG profiles in LGG and established a novel risk model for LGG patients' therapy status and prognosis. Our findings highlight the potential clinical implications of CRGs, suggesting that cuproptosis may be the potential therapeutic target for patients with LGG.

Corrigendum: The risk of heart disease-related death among anaplastic astrocytoma patients after chemotherapy: A SEER population-based analysis.

[This corrects the article DOI: 10.3389/fonc.2022.870843.].

The Risk of Heart Disease-Related Death Among Anaplastic Astrocytoma Patients After Chemotherapy: A SEER Population-Based Analysis.

Background: Despite improved overall survival outcomes, chemotherapy has brought concerns for heart disease-related death (HDRD) among cancer patients. The effect of chemotherapy on the risk of HDRD in anaplastic astrocytoma (AA) patients remains unclear. Methods: We obtained 7,129 AA patients from the Surveillance, Epidemiology, and End Results (SEER) database from 1975 to 2016. Kaplan-Meier and Cox regression analysis were conducted to evaluate the effect of chemotherapy on the HDRD risk. Based on the competing risk model, we calculated the cumulative incidences of HDRD and non-HDRD and performed univariate and multivariate regression analyses. Then, a 1:1 propensity score matching (PSM) was used to improve the comparability between AA patients with and without chemotherapy. Landmark analysis at 216 and 314 months was employed to minimize immortal time bias. Results: AA patients with chemotherapy were at a lower HDRD risk compared to those patients without chemotherapy (adjusted HR=0.782, 95%CI=0.736-0.83, P<0.001). For competing risk regression analysis, the cumulative incidence of HDRD in non-chemotherapy exceeded HDRD in the chemotherapy group (P<0.001) and multivariable analysis showed a lower HDRD risk in AA patients with chemotherapy (adjusted SHR=0.574, 95%CI=0.331-0.991, P=0.046). In the PSM-after cohort, there were no significant association between chemotherapy and the increased HDRD risk (adjusted SHR=0.595, 95%CI=0.316-1.122, P=0.11). Landmark analysis showed that AA patients who received chemotherapy had better heart disease-specific survival than those in the non-chemotherapy group (P=0.007) at the follow-up time points of 216 months. No difference was found when the follow-up time was more than 216 months. Conclusion: AA patients with chemotherapy are associated with a lower risk of HDRD compared with those without chemotherapy. Our findings may help clinicians make a decision about the management of AA patients and provide new and important evidence for applying chemotherapy in AA patients as the first-line treatment. However, more research is needed to confirm these findings and investigate the correlation of the risk of HDRD with different chemotherapy drugs and doses.

Comprehensive Analysis of m5C Methylation Regulatory Genes and Tumor Microenvironment in Prostate Cancer.

Background: 5-Methylcytidine (m5C) methylation is an emerging epigenetic modification in recent years, which is associated with the development and progression of various cancers. However, the prognostic value of m5C regulatory genes and the correlation between m5C methylation and the tumor microenvironment (TME) in prostate cancer remain unknown. Methods: In the current study, the genetic and transcriptional alterations and prognostic value of m5C regulatory genes were investigated in The Cancer Genome Atlas and Gene Expression Omnibus datasets. Then, an m5C prognostic model was established by LASSO Cox regression analysis. Gene set variation analyses (GSVA), gene set enrichment analysis (GSEA), clinical relevance, and TME analyses were conducted to explain the biological functions and quantify the TME scores between high-risk and low-risk subgroups. m5C regulatory gene clusters and m5C immune subtypes were identified using consensus unsupervised clustering analysis. The Cell-type Identification By Estimating Relative Subsets of RNA Transcripts algorithm was used to calculate the contents of immune cells. Results: TET3 was upregulated at transcriptional levels in PCa compared with normal tissues, and a high TET3 expression was associated with poor prognosis. An m5C prognostic model consisting of 3 genes (NSUN2, TET3, and YBX1) was developed and a nomogram was constructed for improving the clinical applicability of the model. Functional analysis revealed the enrichment of pathways and the biological processes associated with RNA regulation and immune function. Significant differences were also found in the expression levels of m5C regulatory genes, TME scores, and immune cell infiltration levels between different risk subgroups. We identified two distinct m5C gene clusters and found their correlation with patient prognosis and immune cell infiltration characteristics. Naive B cells, CD8+ T cells, M1 macrophages and M2 macrophages were obtained and 2 m5C immune subtypes were identified. CTLA4, NSUN6, TET1, and TET3 were differentially expressed between immune subtypes. The expression of CTLA4 was found to be correlated with the degree of immune cell infiltration. Conclusions: Our comprehensive analysis of m5C regulatory genes in PCa demonstrated their potential roles in the prognosis, clinical features, and TME. These findings may improve our understanding of m5C regulatory genes in the tumor biology of PCa.

Deciphering a Novel Necroptosis-Related miRNA Signature for Predicting the Prognosis of Clear Cell Renal Carcinoma.

Clear cell renal cell carcinoma (ccRCC) is the most common histological and devastating subtype of renal cell carcinoma. Necroptosis is a form of programmed cell death that causes prominent inflammatory responses. miRNAs play a significant role in cancer progression through necroptosis. However, the prognostic value of necroptosis-related miRNAs remains ambiguous. In this study, 39 necroptosis-related miRNAs (NRMs) were extracted and 17 differentially expressed NRMs between normal and tumor samples were identified using data form The Cancer Genome Atlas (TCGA). After applying univariate Cox proportional hazard regression analysis and LASSO Cox regression model, six necroptosis-related miRNA signatures were identified in the training cohort and their expression levels were verified by qRT-PCR. Using the expression levels of these miRNAs, all patients were divided into the high- and low-risk groups. Patients in the high-risk group showed poor overall survival (P < 0.0001). Time-dependent ROC curves confirmed the good performance of our signature. The results were verified in the testing cohort and the entire TCGA cohort. Univariate and multivariate Cox regression models demonstrated that the risk score was an independent prognostic factor. Additionally, a predictive nomogram with good performance was constructed to enhance the implementation of the constructed signature in a clinical setting. We then employed miRBD, miRTarBase, and TargetScan to predict the target genes of six necroptosis-related miRNAs. Gene ontology and Kyoto Encyclopedia of Genes and Genomes analyses indicated that 392 potential target genes were enriched in cell proliferation-related biological processes. Six miRNAs and 59 differentially expressed target genes were used to construct an miRNA-mRNA interaction network, and 11 hub genes were selected for survival and tumor infiltration analysis. Drug sensitivity analysis revealed potential drugs that may contribute to cancer management. Hence, necroptosis-related genes play an important role in cancer biology. We developed, for the first time, a necroptosis-related miRNA signature to predict ccRCC prognosis.

Effects of interleukin 1ß on long noncoding RNA and mRNA expression profiles of human synovial fluid derived mesenchymal stem cells.

Synovial fluid-derived mesenchymal stem cells (SFMSCs) play important regulatory roles in the physiological balance of the temporomandibular joint. Interleukin (IL)-1ß regulates the biological behavior of SFMSCs; however, the effects of IL-1ß on long noncoding RNA (lncRNA) and mRNA expression in SFMSCs in the temporomandibular joint are unclear. Here, we evaluated the lncRNA and mRNA expression profiles of IL-1ß-stimulated SFMSCs. Using microarrays, we identified 264 lncRNAs (203 upregulated, 61 downregulated) and 258 mRNAs (201 upregulated, 57 downregulated) that were differentially expressed after treatment with IL-1ß (fold changes ≥ 2, P < 0.05). Kyoto Encyclopedia of Genes and Genomes pathway analysis found that one of the most significantly enriched pathways was the NF-κB pathway. Five paired antisense lncRNAs and mRNAs, eight paired enhancer lncRNAs and mRNAs, and nine paired long intergenic noncoding RNAs and mRNAs were predicted to be co-expressed. A network constructed by the top 30 K-score genes was visualized and evaluated. We found a co-expression relationship between RP3-467K16.4 and IL8 and between LOC541472 and IL6, which are related to NF-κB pathway activation. Overall, our results provide important insights into changes in lncRNA and mRNA expression in IL-1ß-stimulated SFMSCs, which can facilitate the identification of potential therapeutic targets.

A deep learning algorithm for detection of oral cavity squamous cell carcinoma from photographic images: A retrospective study.

BACKGROUND: The overall prognosis of oral cancer remains poor because over half of patients are diagnosed at advanced-stages. Previously reported screening and earlier detection methods for oral cancer still largely rely on health workers' clinical experience and as yet there is no established method. We aimed to develop a rapid, non-invasive, cost-effective, and easy-to-use deep learning approach for identifying oral cavity squamous cell carcinoma (OCSCC) patients using photographic images. METHODS: We developed an automated deep learning algorithm using cascaded convolutional neural networks to detect OCSCC from photographic images. We included all biopsy-proven OCSCC photographs and normal controls of 44,409 clinical images collected from 11 hospitals around China between April 12, 2006, and Nov 25, 2019. We trained the algorithm on a randomly selected part of this dataset (development dataset) and used the rest for testing (internal validation dataset). Additionally, we curated an external validation dataset comprising clinical photographs from six representative journals in the field of dentistry and oral surgery. We also compared the performance of the algorithm with that of seven oral cancer specialists on a clinical validation dataset. We used the pathological reports as gold standard for OCSCC identification. We evaluated the algorithm performance on the internal, external, and clinical validation datasets by calculating the area under the receiver operating characteristic curves (AUCs), accuracy, sensitivity, and specificity with two-sided 95% CIs. FINDINGS: 1469 intraoral photographic images were used to validate our approach. The deep learning algorithm achieved an AUC of 0·983 (95% CI 0·973-0·991), sensitivity of 94·9% (0·915-0·978), and specificity of 88·7% (0·845-0·926) on the internal validation dataset (n = 401), and an AUC of 0·935 (0·910-0·957), sensitivity of 89·6% (0·847-0·942) and specificity of 80·6% (0·757-0·853) on the external validation dataset (n = 402). For a secondary analysis on the internal validation dataset, the algorithm presented an AUC of 0·995 (0·988-0·999), sensitivity of 97·4% (0·932-1·000) and specificity of 93·5% (0·882-0·979) in detecting early-stage OCSCC. On the clinical validation dataset (n = 666), our algorithm achieved comparable performance to that of the average oral cancer expert in terms of accuracy (92·3% [0·902-0·943] vs 92.4% [0·912-0·936]), sensitivity (91·0% [0·879-0·941] vs 91·7% [0·898-0·934]), and specificity (93·5% [0·909-0·960] vs 93·1% [0·914-0·948]). The algorithm also achieved significantly better performance than that of the average medical student (accuracy of 87·0% [0·855-0·885], sensitivity of 83·1% [0·807-0·854], and specificity of 90·7% [0·889-0·924]) and the average non-medical student (accuracy of 77·2% [0·757-0·787], sensitivity of 76·6% [0·743-0·788], and specificity of 77·9% [0·759-0·797]). INTERPRETATION: Automated detection of OCSCC by deep-learning-powered algorithm is a rapid, non-invasive, low-cost, and convenient method, which yielded comparable performance to that of human specialists and has the potential to be used as a clinical tool for fast screening, earlier detection, and therapeutic efficacy assessment of the cancer.