Closed-eye pupil monitoring system in patients with neurological disorders: a prospective, single-arm study

Closed-eye pupil monitoring system in patients with neurological disorders: a prospective, single-arm study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Closed-eye pupil monitoring system in patients with neurological disorders: a prospective, single-arm study Zhiying Shang, Shuting Chen, Zhixiang Hong, Bo Xu, Xiaoyan Bai, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7423596/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objective This study aims to develop a non-invasive dynamic pupil monitoring system for patients with neurological disorders. Traditional methods (such as flashlight examination or infrared devices) rely on patients keeping their eyes open, making them unsuitable for comatose, sedated, or patients with abnormal eyelids, and they also have limitations in intermittent monitoring and subjective interpretation. This study leverages the ability of near-infrared light to penetrate eyelids, combined with deep learning technology, to design a novel closed-eye monitoring device that achieves dynamic capture through sequential infrared projection-eyelid reflection imaging. Methods A prospective single-arm trial design was adopted, enrolling 44 patients in the Department of Neurosurgery at Nanjing Drum Tower Hospital from April to July 2025. Safety was assessed by the incidence of adverse events, and technical accuracy was quantified using diameter prediction error and image segmentation performance. Results Results showed: the incidence of device-related adverse events was zero, the average error rate for pupil diameter prediction was 7.18%, and 92.5% of prediction values fell within the consistency range, indicating good model stability. However, image segmentation performance (Dice coefficient 0.47) and accuracy under extreme anatomical conditions still require optimization. Conclusion This system enables high-precision, safe, and non-invasive pupil monitoring for patients with neurological diseases, overcoming the reliance on patient cooperation in traditional methods, and providing an innovative solution for those unable to cooperate with examinations. Future studies should further validate reliability through multi-center, large-scale trials and optimize dynamic parameter quantification methods for specific neurological diseases. Registry: ChiCTR, TRN: ChiCTR2500105504, Registration date: 1 January 2024. Closed-eye pupil monitoring Neurological diseases Near-infrared light Deep learning Non-invasive monitoring Figures Figure 1 Figure 2 Introduction The pupil serves as a critical window for assessing the central nervous system, particularly brain status, brainstem function [ 1 ] , and autonomic nervous system status [ 2 , 3 ] . Its dynamic changes hold significant value in the monitoring, prognosis [ 4 – 6 ] , and mortality [ 7 ] of neurological disorders. Pupil size and Pupillary Light Reflex (PLR) are regulated by the sympathetic and parasympathetic nervous systems [ 8 , 9 ] , and abnormal manifestations (such as anisocoria, delayed or absent reflexes, or abnormal trajectories) are characteristic of various neurological disorders [ 10 ] . For example, in severe brain injury, pupillary abnormalities often indicate increased intracranial pressure [ 11 ] or risk of brain herniation, and are closely associated with poor prognosis [ 12 , 13 ] ; in epilepsy patients, pupil size is an effective physiological indicator of memory encoding and recall [ 14 ] . Liu's study [ 15 ] used high-resolution functional Magnetic Resonance Imaging(fMRI) combined with real-time pupil measurement technology to reveal specific alterations in the PLR in an Alzheimer's disease mouse model, confirming that this phenomenon is closely associated with functional deterioration in the pontine reticular nucleus, hippocampus, and cholinergic neurotransmission pathways. This finding provides a new biomarker target for non-invasive diagnosis of Alzheimer's disease. However, the current standard pupil assessment methods relied upon in clinical practice—whether manual pupil pen examination [ 16 ] or quantitative infrared pupillometry [ 17 ] —have limitations, particularly in populations of patients with neurological disorders. The primary constraint is that these methods require patients to actively or passively open their eyes. This is often unfeasible for patients in a coma, under deep sedation [ 18 , 19 ] , with eyelid edema [ 20 ] , or with eye injuries, leading to the loss of critical monitoring data. Additionally, traditional monitoring methods are intermittent and may miss transient, dynamic pathophysiological changes. The subjectivity of manual examinations [ 21 ] and the susceptibility of measurement results to environmental light interference also limit their reliability and consistency in busy clinical settings, particularly in intensive care units [ 22 ] . For patients with eyelid abnormalities or eye movement disorders, traditional clinical assessment methods often struggle to yield reliable data. Deep learning-based image analysis technology offers an innovative solution to this challenge [ 23 ] . This technology enables automatic measurement of eyelid morphology by analyzing patient photographs, achieving high precision and good reproducibility while objectively quantifying changes in eyelid morphology before and after surgery. Therefore, addressing the clinical gap in pupil monitoring for patients with neurological disorders, closed-eye pupil monitoring technology leverages image analysis techniques based on deep learning. By utilizing the property of near-infrared light to penetrate the eyelids, this technology enables non-invasive dynamic capture [ 24 ] . Its advantage lies in overcoming the reliance of traditional methods on eye-opening cooperation, thereby making continuous monitoring of comatose and sedated patients feasible. This technology provides real-time continuous data, reduces environmental light interference through physical shielding, and enables objective quantification of parameters such as pupil diameter and PLR. Based on this principle, our research team designed a novel closed-eye pupil monitoring device using temporal infrared projection-eyelid reflection imaging technology: light penetrates the eyeball and is reflected by the pupil, with the closed-eye pupil dynamics captured by a front-facing camera. Despite the promising technical prospects, there is currently a lack of systematic validation data for this device in real-world populations with neurological disorders. There is an urgent need to clarify its clinical applicability, safety, and the specific pupillary parameter characteristics under closed-eye conditions. To this end, following preliminary testing conducted at the Department of Neurosurgery, Nanjing Drum Tower Hospital from December 2024 to January 2025, we are now initiating this prospective single-arm study to systematically assess the clinical feasibility and safety of this device in patients with neurological disorders. This study will provide critical empirical evidence to support the clinical translation of closed-eye pupil monitoring technology. Methods Study design This study was a prospective single-arm trial conducted at Nanjing Gulou Hospital on April 15, 2025, and approved by the Ethics Committee of Nanjing Drum Tower Hospital (2024-727-02). Informed consent was obtained from all individual participants included in the study. Since this study was a prospective, single-arm controlled trial, no statistical calculation of sample size is required. Subject screening was conducted in the neurosurgery inpatient department. The costs of subject examinations and any subsequent adverse event treatments were covered by the investigators, and there were no other incentives for participating in the study. Two experienced neurosurgery department heads independently screened and assessed participants using a consensus method. Participants meeting the minimum criteria for this study were enrolled and underwent examinations. All participants signed informed consent forms. Follow-up assessments were conducted 72 hours after collecting pupil data. The device's efficacy and accuracy were evaluated based on preoperative and postoperative comparisons of participant pupil diameter, while its safety was assessed through the incidence and severity of adverse events (AEs). Patients selection The primary inclusion criteria are as follows: (1) Males and females aged 18 years or older; (2) Diagnosed with a neurological disorder (such as traumatic brain injury, cerebrovascular disease, neurodegenerative disease, etc.) by two senior neurosurgeons from the neurosurgery team; (3) Complete eyelid anatomy with no congenital malformations that impair infrared light penetration; (4) Voluntarily signed informed consent form. Individuals meeting any of the following criteria will be excluded from study participation: presence of ocular lesions (including iris abnormalities, active uveitis, cataracts, intraocular lens implantation within the past month, or eyelid defects/infections); presence of periorbital skin damage, implantation of an intracranial pressure monitoring device, or allergy to study-related materials. Terminally ill patients (expected survival period < 48 hours), pregnant women, and individuals participating in other interventional clinical trials are also excluded. The procedure During the testing process, we measured each volunteer's eyes one by one, and only operated on one eye at a time to ensure data accuracy and safety. An infrared pulsed Light-Emitting Diode (LED) was installed inside an opaque sleeve, which was tightly fitted to the volunteer's temple. The key step in the test was to determine the optimal position of the infrared LED on the temple. By continuously capturing infrared images of the pupil in the open-eye state, we dynamically adjusted the LED's position until a clear and distinguishable pupil image was obtained. After positioning was completed, we verified whether infrared light could penetrate the eyelid by having the volunteer close their eyes. To control variables, participants were required to maintain fixation on a fixed point throughout the test to minimize interference from pupil movement. To ensure data accuracy, volunteers were instructed to maintain visual fixation throughout the process to reduce pupil movement. By comparing images taken with eyes open and closed, we found that the eyelids absorb and scatter the infrared light emitted by the pupil. Two experienced operators manually calibrated the pupil contour with eyes open and calculated the equivalent diameter as the baseline value. We collected 1–3 valid data points from each group of subjects, and all samples included in the analysis underwent strict quality control (invalid data with incomplete blinking movements or blurred images were excluded). Among the 50 volunteers, we successfully imaged the pupils of 44 closed-eye subjects. Among these 44 volunteers, we obtained at least one pair of images clearly showing the pupils before and after the white light pulse, totaling 185 such image pairs analyzed. Outcome assessments The primary evaluation criteria are the safety and technical accuracy of the treatment. Safety is primarily assessed through an active telephone follow-up mechanism by the research team, tracking potential delayed adverse events within 24 to 72 hours after device use. Adverse events are reported and categorized by the investigators. If an adverse event results in death, life-threatening conditions, requires hospitalization or prolongs hospitalization, causes or may cause permanent damage to bodily structure or function, or causes or may cause fetal disease, congenital defects, or abnormalities in offspring, it is classified as a serious adverse event (SAE). Technical accuracy is assessed through diameter prediction error, segmentation performance, and model stability. In terms of diameter prediction, absolute error levels are evaluated using mean absolute error (MAE) and Root Mean Square Error (RMSE), and a systematic comparison framework between AI-automated measurements and manual gold standard measurements is established using the Bland-Altman consistency analysis method under both closed-eye and open-eye conditions. Image segmentation performance was evaluated using pixel accuracy (PixelAcc), Dice coefficient, and intersection-over-union (IoU). PixelAcc reflects overall classification accuracy, while Dice and IoU focus on assessing the precision of target region segmentation. Model stability analysis used the proportion of low-error samples as an indicator and calculated its 95% confidence interval. All data collection strictly followed standardized quality control procedures to ensure that abnormal data was screened and excluded prior to analysis. Statistical analysis Descriptive statistical analysis was performed on all parameters. Continuous variables were described using mean ± standard deviation (Mean ± SD) combined with median and extremes to characterize their distribution; The average absolute error (in millimeters) between closed-eye measurements and the manual gold standard was quantified using MAE, while RMSE was used to highlight the contribution of larger errors to the degree of dispersion. Additionally, the Bland-Altman four-step consistency analysis method was employed—calculating the mean (d̄) and standard deviation of measurement differences, determining the 95% consistency limits (LoA), and plotting a difference-mean scatter plot— — using a clinically acceptable threshold (± 0.5 millimeters) for judgment. Image processing and data analysis were performed using Python 3.13. P < 0.05 was considered statistically significant (two-tailed test). Results Patient characteristics This study included a total of 44 patients who met the inclusion criteria, with enrollment occurring between April 15, 2025, and July 10, 2025. Demographic characteristics of the patients were as follows(Table 1 ): 29 males (65.91%) and 15 females (34.09%). Neurological examination results showed that 21 patients (47.73%) had normal muscle strength in all four limbs (Grade 5), 16 patients (36.36%) exhibited Grade 4 muscle strength in all four limbs, 6 patients (13.64%) presented with Grade 5 muscle strength in both upper limbs accompanied by Grade 4 muscle strength in both lower limbs, and 1 case (2.27%) had bilateral upper limb muscle strength of grade 4 with asymmetric lower limb muscle strength (right lower limb grade 4, left lower limb grade 5). In the pupillary light reflex assessment, 42 cases (95.45%) had a sensitive response to light, and 2 cases (4.55%) had a delayed response to light. Additionally, the study identified three patients with special conditions: one was in the postoperative edema phase, one had hearing impairment that did not affect experimental communication, and one developed eyelid thickening postoperatively due to an intracranial space-occupying lesion. None of these special conditions affected the normal conduct of the trial. Table 1 General information about patients Baseline characteristic Age 48.73 ± 18.69 Gender Male 29(65.91%) Female 15(34.09%) Body Mass Index(BMI)/(kg/m 2 ) 24.07 ± 4.44 Muscle Normal 21(47.73%) Muscle strength in limbs: Grade 4 16(36.36%) Grade 5 in both upper limbs, grade 4 in both lower limbs 6(13.64%) Grade 4 in both upper limbs, grade 4 in the right lower limb, and grade 5 in the left lower limb 1(2.27%) Surgery Preoperative 21(47.73%) Postoperative 23(52.27%) PLR Sensitive 42(95.45%) Sluggish 2(4.55%) Glasgow Coma Scale(GCS) E4V5M6 44(100%) Diagnosis Intracranial space-occupying lesion 23(52.27%) Intraspinal space-occupying lesion 17(38.64%) Other neurological disorders 4(9.09%) Glasgow Coma Scale(GCS) : Eye opening response (4 points),Verbal response (5 points), Motor response (6 points), A maximum score of 15 indicates clear consciousness; 13–15 is mildly impaired consciousness; 9–12 is moderately impaired consciousness; and 8 or less is coma. Safety No device-related adverse events occurred, nor were there any SAEs. Accuracy After the initial data collection was completed on July 10, we used ‘python predict.py --model-path E:\models\1\850epoch_model.pth --input-dir [input directory] --output-dir [output directory] --batch-size 1 --threshold 0.68’ to predict the collected mask images(Fig. 1 ). To address the prediction results, the research team fixed errors in the test program related to mask area and brightness calculations, removed channel dimension issues caused by ‘.squeeze()’, and tested models with 1000 to 3000 epochs, but found that there were misjudgments in the bright areas on both sides. The research team then introduced attention mechanisms and residual connections, and adjusted the learning rate from 3e-4 to 1e-4. However, during testing, they still found that the accuracy was low, and speculated that this might be related to the model saving or calling process. Additionally, it was discovered that statistical differences in ‘BatchNorm2d’ between training and testing (training batch_size = 4, testing batch_size = 1) caused model output instability. To address this, we replaced it with `InstanceNorm2d(track_running_stats = False)`, improving testing consistency. Additionally, the optimized model's binary threshold was adjusted from the original 0.50–0.60 to 0.61–0.71, and the test results of each epoch model were recorded in a CSV file for analysis. Ultimately, the 850-epoch model (threshold 0.68) (MAE of 184.1 pixels², MAPE of 9.1%) was determined to be the primary recommended model(Table 2 ). In summary, the average error between the pupil diameter calculated based on the prediction mask (pred_d) and the true value (gt_d) is -0.40 pixels, with an error rate of 7.18%. Most prediction results are concentrated within ± 2 pixels. The minimum error rate is 0.26%, the maximum error rate is 20.76%, and the standard deviation is 6.54%, indicating that some samples exhibit significant prediction deviations. The high error rates in edge prediction and certain samples (exceeding 19%) suggest that the model's stability in complex scenarios still requires optimization. Overall, the model performs reasonably well in measuring target diameter, but there is significant room for improvement in segmentation performance (Dice coefficient 0.47, IoU 0.33) and edge accuracy. Through statistical analysis of the model's prediction results, we found that the model performs well overall in radius prediction but still has room for optimization. The average deviation in radius prediction is -0.40 (error rate 7.18%), with a median deviation of -0.23 (error rate 4.06%), indicating a slight systematic underestimation bias in the model. The standard deviation of the error is 1.52 (error rate standard deviation 6.54%), suggesting that the prediction results remain relatively stable. Additionally, consistency analysis shows that the average difference (bias) between predicted and actual values is -0.13, with a standard deviation (SD) of 0.94, indicating a small overall bias and controllable variability. The 95% consistency limits range from − 1.98 to + 1.72, with 92.5% of samples falling within this range, which fully validates the reliability of the model's predictions. Overall, while the current model exhibits a slight underestimation trend, its predictive stability is good, with 92.5% of the prediction results falling within an acceptable error range. All outcomes have shown in the Table 3 . Table 2 Model statistical analysis Epoch Imange of samples MAE MAPE(%) Low error sample ratio(%) 850 185 184.1 9.1% 65.9% 1650 185 199.3 9.7% 54.5% 2550 185 229.8 12.3% 38.6% 1550 185 209.5 10.5% 45.5% 900 185 236.6 11.3% 52.3% 1700 185 212.0 10.7% 40.9% 950 185 250.8 12.4% 50.0% 800 185 384.0 20.4% 27.3% 1800 185 325.6 16.3% 22.7% Table 3 Outcomes Indicators Radius error (err) Error rate (err_rate) mean -0.40 7.18% median -0.23 4.06% standard deviation 1.52 6.54% Minimum value -3.64 0.26% Maximum value 2.25 20.76% Mean difference (Bias) -0.13 / Standard deviation of differences 0.94 / lower limit of consistency -1.98 / Upper limit of consistency + 1.72 / Inclusion ratio 92.5% / Discussion The closed-eye pupil monitoring system developed in this study represents an innovative breakthrough in the field of neural monitoring technology. The system is based on the principles of near-infrared spectroscopy [ 25 , 26 ] , leveraging the characteristics of this wavelength band, which has a deeper penetration depth in biological tissues and lower scattering effects compared to visible light, providing a physical foundation for obtaining clearer images of the pupil in a closed-eye state. Additionally, this study employs an experimental protocol involving infrared light illumination of the temples [ 27 ] , avoiding the potential visual discomfort and potential eye damage caused by direct light exposure in traditional pupil measurement methods, while enhancing the comfort experience of participants during actual testing. In terms of algorithm performance, the closed-eye pupil measurement system developed in this study achieved clinically significant measurement accuracy, with an average error rate of 7.18%, comparable to the 5–8% error rate of currently widely used quantitative infrared pupillometers [ 16 , 28 ] . Previous studies [ 16 ] have shown that traditional subjective visual estimation methods have an error rate as high as 19% within the pupil size range of 2–4 mm, making it prone to missing diagnoses of anisocoria and weak pupillary light reflexes. Quantitative infrared pupillometers, on the other hand, measure parameters such as pupil size, contraction latency, and speed, and generate a Neural Pupillary Index (NPI), providing objective evidence for clinical assessment [ 29 , 30 ] . A significant improvement in this study was the replacement of BatchNorm2d with InstanceNorm2d. This adjustment effectively addressed the issue of limited batch size in clinical settings due to patient privacy protection and data acquisition restrictions, enhancing the system's universality. However, in this study, we also observed individual cases with a maximum error rate of 20.76%, which, upon analysis, may be related to individual anatomical differences(such as eyelid thickness and subcutaneous fat distribution). One patient with post-surgical eyelid edema following intracranial space-occupying lesion surgery exhibited abnormal eyelid thickening and delayed light reflexes. During mask image processing, we may have introduced biases, leading to an increase in the maximum error rate. This suggests that future research should establish personalized correction models for eyelid optical characteristics to further optimize system performance. This study fills a technical gap in non-invasive monitoring within the field of neurocritical care. Compared to traditional methods such as invasive intracranial pressure monitoring [ 31 , 32 ] , this experiment is the first to systematically validate the feasibility of closed-eye measurements in a real clinical setting, overcoming the limitations of previous studies that were primarily focused on pupil monitoring in critically ill patients with their eyes open [ 12 , 13 , 33 ] ; Additionally, an end-to-end deep learning architecture was employed to enhance the performance of traditional image processing pipelines. Furthermore, all data were collected from real clinical scenarios, making the findings more clinically applicable. The results of this study show that 92.5% of the predicted values fall within the consistency range, demonstrating clinical reference value. Notably, no adverse events or serious adverse events were observed throughout the study, and its safety has been certified in accordance with the IEC 62471 photobiological safety standard. However, the study results indicate that the Dice coefficient is 0.47, suggesting that the current algorithm used in the image segmentation process has limitations. By analyzing the provided mask image samples, we believe the primary errors stem from: 1) partial volume effects at the eyelid-sclera junction; 2) shadow artifacts caused by eyelashes; 3) motion blur caused by eyelid micro-movements. To address these technical bottlenecks, we will implement multiple measures in future studies, including incorporating temporal information to leverage video continuity features, adopting adversarial learning strategies to enhance edge retention capabilities, and integrating optical coherence tomography (OCT) technology for auxiliary calibration. Although this study has made significant progress in validating clinical feasibility and safety, limitations remain. The study design was a single-arm trial with a small sample size from a single medical center, which may affect the representativeness and generalizability of the results. Participants excluded patients with eyelid defects or drug interference, and quantification methods for dynamic parameters (e.g., PLR velocity) are not yet mature. Additionally, short-term follow-up cannot assess long-term safety. Conclusion This study confirms that a deep learning-based closed-eye pupil monitoring system can achieve non-invasive pupil measurement in patients with neurological diseases with high accuracy and good safety, with no device-related adverse events observed. This technology overcomes the reliance of traditional pupil monitoring on patient cooperation with open eyes, providing an innovative continuous monitoring solution for patients who cannot cooperate with examinations, such as those in a coma or under sedation, and holds significant clinical translation value. Although image segmentation accuracy and measurement performance under extreme anatomical conditions (such as abnormally thickened eyelids) still require optimization, the overall performance has already reached a clinically acceptable range. Future research will involve multi-center, large-scale clinical validation to enhance the reliability of research conclusions; targeted analyses for specific neurological conditions (such as brain injury, epilepsy, Alzheimer's disease, etc.) to establish disease-specific pupil parameter profiles; and refinement of quantification methods for dynamic parameters (such as PLR velocity and contraction latency) to enhance the comprehensiveness of neurological function assessment. Declarations Acknowledgements None. Author contributions ZYS, STC and ZXH designed research, performed research and analyzed data; ZYS, JYL, SL and LWZ wrote the paper. BX, XYB and LC peformed research. All authors contributed todrafting of the original manuscript. Funding Project of China Hospital Reform and Development Research Institute, Nanjing University; Aid project of Jiangsu Ningai Medical Development & Medical Aid Foundation(NDYGN2024017); 2023 Open Research Fund for the Key Laboratory of the Ministry of Education for Geriatric Long-Term Care (Naval Military Medical University)(LNYB-2023-13) 2023 Special Funds Project for Transformation of Scientific and Technological Achievements of Nanjing Gulou Hospital(2023-GNYZ-YB-07); 2024 Nanjing Health Science and Technology Development Special Funds Grant(ZKX24025) Data availability The datasets generated and analyzed during the current study are not publicly available due to restrictions specified in the signed partnership agreement. However, they can be accessed from the corresponding author upon reasonable request. Ethics approval and consent to participate Approved by the Ethics Committee of Nanjing Drum Tower Hospital (2024-727-02) and informed consent was obtained from all individual participants included in the study. Competing interests The authors declare no competing interests. Previous presentations None. Clinical trial number ChiCTR2500105504 Acknowledgements:Project of China Hospital Reform and Development Research Institute, Nanjing University; Aid project of Jiangsu Ningai Medical Development & Medical Aid Foundation(NDYGN2024017); 2023 Open Research Fund for the Key Laboratory of the Ministry of Education for Geriatric Long-Term Care (Naval Military Medical University)(LNYB-2023-13) 2023 Special Funds Project for Transformation of Scientific and Technological Achievements of Nanjing Gulou Hospital(2023-GNYZ-YB-07); 2024 Nanjing Health Science and Technology Development Special Funds Grant(ZKX24025) References Grujic, N., Polania, R., & Burdakov, D. (2024). Neurobehavioral meaning of pupil size. Neuron , 112 (20), 3381～3395. https://doi.org/10.1016/j.neuron.2024.05.029. The autonomic nervous system and the eye. (1985). Lancet (London, England) , 2 (8455), 591–592. Wang, Y., Zekveld, A. A., Naylor, G., Ohlenforst, B., Jansma, E. P., Lorens, A., Lunner, T., & Kramer, S. E. (2016). Parasympathetic Nervous System Dysfunction, as Identified by Pupil Light Reflex, and Its Possible Connection to Hearing Impairment. PLOS ONE , 11 (4), e0153566. https://doi.org/10.1371/journal.pone.0153566. Choi, S. C., Narayan, R. K., Anderson, R. L., & Ward, J. D. (1988). Enhanced specificity of prognosis in severe head injury. Journal of neurosurgery, 69(3), 381–385. https://doi.org/10.3171/jns.1988.69.3.0381. Emami, P., Czorlich, P., Fritzsche, F. S., Westphal, M., Rueger, J. M., Lefering, R., & Hoffmann, M. (2017). Impact of Glasgow Coma Scale score and pupil parameters on mortality rate and outcome in pediatric and adult severe traumatic brain injury: A retrospective, multicenter cohort study. Journal of Neurosurgery, 126(3), 760～767. https://doi.org/10.3171/2016.1.jns152385. Marmarou, A., Lu, J., Butcher, I., McHugh, G. S., Murray, G. D., Steyerberg, E. W., Mushkudiani, N. A., Choi, S., & Maas, A. I. R. (2007). Prognostic Value of The Glasgow Coma Scale And Pupil Reactivity in Traumatic Brain Injury Assessed Pre-Hospital And on Enrollment: An IMPACT Analysis. Journal of Neurotrauma , 24 (2), 270～280. https://doi.org/10.1089/neu.2006.0029. Greer, D. M., Shemie, S. D., Lewis, A., Torrance, S., Varelas, P., Goldenberg, F. D., Bernat, J. L., Souter, M., Topcuoglu, M. A., Alexandrov, A. W., Baldisseri, M., Bleck, T., Citerio, G., Dawson, R., Hoppe, A., Jacobe, S., Manara, A., Nakagawa, T. A., Pope, T. M., Silvester, W., … Sung, G. (2020). Determination of Brain Death/Death by Neurologic Criteria: The World Brain Death Project. JAMA , 324(11), 1078–1097. https://doi.org/10.1001/jama.2020.11586. Wilhelm H. (2011). Disorders of the pupil. Handbook of clinical neurology, 102, 427–466. https://doi.org/10.1016/B978-0-444-52903-9.00022-4 Kawasaki A. (2025). Disorders of the Pupil. Continuum (Minneapolis, Minn.), 31(2), 527–542. https://doi.org/10.1212/cont.0000000000001542 Barry, C., & Wang, E. (2023). Racially fair pupillometry measurements for RGB smartphone cameras using the far red spectrum. Scientific reports , 13 (1), 13841. https://doi.org/10.1038/s41598-023-40796-0. Kolias, A. G., Adams, H., Timofeev, I. S., Corteen, E. A., Hossain, I., Czosnyka, M., Timothy, J., Anderson, I., Bulters, D. O., Belli, A., Eynon, C. A., Wadley, J., Mendelow, A. D., Mitchell, P. M., Wilson, M. H., Critchley, G., Sahuquillo, J., Unterberg, A., Posti, J. P., … Buxton, M. (2022). Evaluation of Outcomes Among Patients With Traumatic Intracranial Hypertension Treated With Decompressive Craniectomy vs Standard Medical Care at 24 Months: A Secondary Analysis of the RESCUEicp Randomized Clinical Trial. JAMA Neurology , 79 (7), 664. https://doi.org/10.1001/jamaneurol.2022.1070. Oddo, M., Taccone, F. S., Petrosino, M., Badenes, R., Blandino-Ortiz, A., Bouzat, P., Caricato, A., Chesnut, R. M., Feyling, A. C., Ben-Hamouda, N., Hemphill, J. C., Koehn, J., Rasulo, F., Suarez, J. I., Elli, F., Vargiolu, A., Rebora, P., Galimberti, S., Citerio, G., … Wang, R. (2023). The Neurological Pupil index for outcome prognostication in people with acute brain injury (ORANGE): A prospective, observational, multicentre cohort study. The Lancet Neurology , 22 (10), 925～933. https://doi.org/10.1016/S1474-4422(23)00271-5. Petrosino, M., Gouvêa Bogossian, E., Rebora, P., Galimberti, S., Chesnut, R., Bouzat, P., Oddo, M., Taccone, F. S., Citerio, G., & ORANGE Participants and Investigators. (2025). Neurological Pupil Index and Intracranial Hypertension in Patients With Acute Brain Injury: A Secondary Analysis of the ORANGE Study. JAMA Neurology , 82 (2), 176. https://doi.org/10.1001/jamaneurol.2024.4189. Cimbalnik, J., Dolezal, J., Topçu, Ç., Lech, M., Marks, V. S., Joseph, B., Dobias, M., Van Gompel, J., Worrell, G., & Kucewicz, M. (2022). Intracranial electrophysiological recordings from the human brain during memory tasks with pupillometry. Scientific Data , 9 (1). https://doi.org/10.1038/s41597-021-01099-z. Liu, X., Hike, D., Choi, S., Man, W., Ran, C., Zhou, X. A., Jiang, Y., & Yu, X. (2024). Identifying the bioimaging features of Alzheimer’s disease based on pupillary light response-driven brain-wide fMRI in awake mice. Nature Communications , 15 (1), 9657. https://doi.org/10.1038/s41467-024-53878-y. Couret, D., Boumaza, D., Grisotto, C., Triglia, T., Pellegrini, L., Ocquidant, P., Bruder, N. J., & Velly, L. J. (2016). Reliability of standard pupillometry practice in neurocritical care: An observational, double-blinded study. Critical Care , 20 (1), 99. https://doi.org/10.1186/s13054-016-1239-z. Favre, E., Bernini, A., Morelli, P., Pasquier, J., Miroz, J.-P., Abed-Maillard, S., Ben-Hamouda, N., & Oddo, M. (2020). Neuromonitoring of delirium with quantitative pupillometry in sedated mechanically ventilated critically ill patients. Critical Care , 24 (1). https://doi.org/10.1186/s13054-020-2796-8. Chan, W. P., Prescott, B. R., Barra, M. E., Chung, D. Y., Kim, I. S., Saglam, H., Hutch, M. R., Shin, M., Zafar, S. F., Benjamin, E. J., Smirnakis, S. M., Dupuis, J., Greer, D. M., & Ong, C. J. (2022). Dexmedetomidine and Other Analgosedatives Alter Pupil Characteristics in Critically Ill Patients. Critical Care Explorations , 4 (5), e0691. https://doi.org/10.1097/cce.0000000000000691. Gouvêa Bogossian, E., Blandino Ortiz, A., Esposito, V., Caricato, A., Righy Shinotsuka, C., Monléon Lopez, B., Giannì, G., Macchini, E., De Pablo Sanchez, R., Pisapia, L., Turon, R., Gonçalves, B., Badenes, R., Kurtz, P., & Taccone, F. S. (2023). Neurological Pupil Index and Delayed Cerebral Ischemia after Subarachnoid Hemorrhage: A Retrospective Multicentric Study. Neurocritical Care , 39 (1), 116～124. https://doi.org/10.1007/s12028-023-01744-y. Ong, C., Hutch, M., Barra, M., Kim, A., Zafar, S., & Smirnakis, S. (2019). Effects of Osmotic Therapy on Pupil Reactivity: Quantification Using Pupillometry in Critically Ill Neurologic Patients. Neurocritical Care , 30 (2), 307～315. https://doi.org/10.1007/s12028-018-0620-y. Olson, D. M., Stutzman, S., Saju, C., Wilson, M., Zhao, W., & Aiyagari, V. (2016). Interrater Reliability of Pupillary Assessments. Neurocritical Care , 24 (2), 251～257. https://doi.org/10.1007/s12028-015-0182-1. Robba, C., Graziano, F., Rebora, P., Elli, F., Giussani, C., Oddo, M., Meyfroidt, G., Helbok, R., Taccone, F. S., Prisco, L., Vincent, J.-L., Suarez, J. I., Stocchetti, N., Citerio, G., Abdelaty, M., Abed Maillard, S., Ahmed, H., Albrecht, L., Alsudani, A., … Zerbi, S. M. (2021). Intracranial pressure monitoring in patients with acute brain injury in the intensive care unit (SYNAPSE-ICU): An international, prospective observational cohort study. The Lancet Neurology , 20 (7), 548～558. https://doi.org/10.1016/s1474-4422(21)00138-1. Lou, L., Cao, J., Wang, Y., Gao, Z., Jin, K., Xu, Z., Zhang, Q., Huang, X., & Ye, J. (2021). Deep learning-based image analysis for automated measurement of eyelid morphology before and after blepharoptosis surgery. Annals of medicine, 53(1), 2278–2285. https://doi.org/10.1080/07853890.2021.2009127. Farraj, Y., Buxboim, A., Cohen, J. E., Kan-Tor, Y., Glasner Hagege, S., Weiss, D., Goldman, V., & Beatus, T. (2021). Measuring pupil size and light response through closed eyelids. Biomedical Optics Express , 12 (10), 6485. https://doi.org/10.1364/BOE.435508. Afara, I. O., Shaikh, R., Nippolainen, E., Querido, W., Torniainen, J., Sarin, J. K., Kandel, S., Pleshko, N., & Töyräs, J. (2021). Characterization of connective tissues using near-infrared spectroscopy and imaging. Nature Protocols , 16 (2), 1297～1329. https://doi.org/10.1038/s41596-020-00468-z. Byrne, H. J., Knief, P., Keating, M. E., & Bonnier, F. (2016). Spectral pre and post processing for infrared and Raman spectroscopy of biological tissues and cells. Chemical Society Reviews , 45 (7), 1865～1878. https://doi.org/10.1039/C5CS00440C. Gordon, K. B., Char, D. H., & Sagerman, R. H. (1995). Late effects of radiation on the eye and ocular adnexa. International journal of radiation oncology, biology, physics , 31 (5), 1123～1139. https://doi.org/10.1016/0360-3016(95)00062-4. T, K., E, T., J, B., & Em, K. (2003). Comparison of a digital and a handheld infrared pupillometer for determining scotopic pupil diameter. Journal of Cataract and Refractive Surgery , 29 (1). https://doi.org/10.1016/s0886-3350(02)01898-9. Boulter, J. H., Shields, M. M., Meister, M. R., Murtha, G., Curry, B. P., & Dengler, B. A. (2021). The Expanding Role of Quantitative Pupillometry in the Evaluation and Management of Traumatic Brain Injury. Frontiers in Neurology , 12 , 685313. https://doi.org/10.3389/fneur.2021.685313. P, M., M, O., & N, B.-H. (2019). Role of automated pupillometry in critically ill patients. Minerva Anestesiologica , 85 (9). https://doi.org/10.23736/S0375-9393.19.13437-2. Gwj, H., G, C., P, H., A, K., G, M., C, R., N, S., & R, C. (2022). Intracranial pressure: Current perspectives on physiology and monitoring. Intensive Care Medicine , 48 (10). https://doi.org/10.1007/s00134-022-06786-y. Kb, E., & Pk, E. (2020). Measuring intracranial pressure by invasive, less invasive or non-invasive means: Limitations and avenues for improvement. Fluids and Barriers of the CNS , 17 (1). https://doi.org/10.1186/s12987-020-00195-3. Mm, B., Aj, S., Jc, X., S, S.-N., W, Y., & Li, G. (2021). Quantitative Pupillometry in the Intensive Care Unit. Journal of Intensive Care Medicine , 36 (4). https://doi.org/10.1177/0885066619881124. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7423596","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":511303533,"identity":"e8830b13-1285-4969-a804-f7579461ff45","order_by":0,"name":"Zhiying Shang","email":"","orcid":"","institution":"Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zhiying","middleName":"","lastName":"Shang","suffix":""},{"id":511303534,"identity":"d01bd029-1778-41d0-9dcf-fe49cd5d1ad8","order_by":1,"name":"Shuting Chen","email":"","orcid":"","institution":"Kuang Yaming College, Nanjing University","correspondingAuthor":false,"prefix":"","firstName":"Shuting","middleName":"","lastName":"Chen","suffix":""},{"id":511303535,"identity":"531b624f-e1bd-467e-a7f4-47c96668b542","order_by":2,"name":"Zhixiang Hong","email":"","orcid":"","institution":"Kuang Yaming College, Nanjing University","correspondingAuthor":false,"prefix":"","firstName":"Zhixiang","middleName":"","lastName":"Hong","suffix":""},{"id":511303536,"identity":"2596c4ca-e8fc-4025-a92c-2652347f65a0","order_by":3,"name":"Bo Xu","email":"","orcid":"","institution":"Drum Tower Clinical Medical College，Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Bo","middleName":"","lastName":"Xu","suffix":""},{"id":511303537,"identity":"90eb5d61-1b8e-4ad8-b753-7fd66cd47ce0","order_by":4,"name":"Xiaoyan Bai","email":"","orcid":"","institution":"Drum Tower Clinical Medical College，Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiaoyan","middleName":"","lastName":"Bai","suffix":""},{"id":511303538,"identity":"936c1a6c-f7a1-4ec2-b712-0dc05da8c91a","order_by":5,"name":"Jingyu Li","email":"","orcid":"","institution":"Nanjing University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jingyu","middleName":"","lastName":"Li","suffix":""},{"id":511303539,"identity":"52177668-20de-46dc-a21a-97b8798f09e1","order_by":6,"name":"Shuang Liang","email":"","orcid":"","institution":"Nanjing University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Shuang","middleName":"","lastName":"Liang","suffix":""},{"id":511303540,"identity":"ceca5569-0d8a-41e1-ac6a-f0213aad1460","order_by":7,"name":"Lingwei Zhang","email":"","orcid":"","institution":"Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lingwei","middleName":"","lastName":"Zhang","suffix":""},{"id":511303541,"identity":"bb721c65-e1d1-4c9c-b2e1-0b71477d2afd","order_by":8,"name":"Lu Chen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7ElEQVRIie3RMQrCMBSA4UigLsGurxTqFSKFgiB4lQShLhWdXFUK8QoVL+GkqyLoUjpncKhLZ0Fw1VQnl1g3wfwkQ8L7yBCETKZfDNTOUYfYGG+fB7StQhgKPWcu2Fdk79M0pa+bT8RexoecC8ynMrrlbYG8hmS160j3yOnQp1xYfJYMNtQRyHckw26iIRSiALggPIbBGhThK8ksTLRkeFOkXFFRkkkFEllqnvqEpFZJGP1EQIYBsIx5UBcBQAatRXqOXR2xk14Bl/GddPe4cGHcaTaOvd1VR97C8Pym2rQqULOX6rMmk8n0Rz0Az+xIBLDW3TUAAAAASUVORK5CYII=","orcid":"","institution":"Nanjing Medical University","correspondingAuthor":true,"prefix":"","firstName":"Lu","middleName":"","lastName":"Chen","suffix":""}],"badges":[],"createdAt":"2025-08-21 07:53:27","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7423596/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7423596/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":91085371,"identity":"b3cd75e8-2589-4963-acef-d7716de70df0","added_by":"auto","created_at":"2025-09-11 12:21:30","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":244090,"visible":true,"origin":"","legend":"\u003cp\u003ePartial mask image\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7423596/v1/dd04c4df7e320431fd29819c.png"},{"id":91085368,"identity":"64292a79-09bb-4d07-ae53-538687baf16d","added_by":"auto","created_at":"2025-09-11 12:21:30","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":77636,"visible":true,"origin":"","legend":"\u003cp\u003eMAE\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7423596/v1/270ab68503fad3de4192c49a.jpg"},{"id":92607418,"identity":"9fddff5d-b3b3-4231-9fa8-287f42b6066f","added_by":"auto","created_at":"2025-10-01 15:23:51","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1216279,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7423596/v1/3c7a6fc4-2880-4877-a0f4-0e0e9a15259e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Closed-eye pupil monitoring system in patients with neurological disorders: a prospective, single-arm study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe pupil serves as a critical window for assessing the central nervous system, particularly brain status, brainstem function\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e, and autonomic nervous system status\u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e. Its dynamic changes hold significant value in the monitoring, prognosis\u003csup\u003e[\u003cspan additionalcitationids=\"CR5\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e, and mortality\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e of neurological disorders. Pupil size and Pupillary Light Reflex (PLR) are regulated by the sympathetic and parasympathetic nervous systems\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e, and abnormal manifestations (such as anisocoria, delayed or absent reflexes, or abnormal trajectories) are characteristic of various neurological disorders\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e. For example, in severe brain injury, pupillary abnormalities often indicate increased intracranial pressure\u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e or risk of brain herniation, and are closely associated with poor prognosis\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e; in epilepsy patients, pupil size is an effective physiological indicator of memory encoding and recall\u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e. Liu's study\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003eused high-resolution functional Magnetic Resonance Imaging(fMRI) combined with real-time pupil measurement technology to reveal specific alterations in the PLR in an Alzheimer's disease mouse model, confirming that this phenomenon is closely associated with functional deterioration in the pontine reticular nucleus, hippocampus, and cholinergic neurotransmission pathways. This finding provides a new biomarker target for non-invasive diagnosis of Alzheimer's disease. However, the current standard pupil assessment methods relied upon in clinical practice\u0026mdash;whether manual pupil pen examination\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e or quantitative infrared pupillometry\u003csup\u003e[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/sup\u003e\u0026mdash;have limitations, particularly in populations of patients with neurological disorders. The primary constraint is that these methods require patients to actively or passively open their eyes. This is often unfeasible for patients in a coma, under deep sedation\u003csup\u003e[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e, with eyelid edema\u003csup\u003e[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e, or with eye injuries, leading to the loss of critical monitoring data. Additionally, traditional monitoring methods are intermittent and may miss transient, dynamic pathophysiological changes. The subjectivity of manual examinations\u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e and the susceptibility of measurement results to environmental light interference also limit their reliability and consistency in busy clinical settings, particularly in intensive care units\u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e. For patients with eyelid abnormalities or eye movement disorders, traditional clinical assessment methods often struggle to yield reliable data. Deep learning-based image analysis technology offers an innovative solution to this challenge\u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003e. This technology enables automatic measurement of eyelid morphology by analyzing patient photographs, achieving high precision and good reproducibility while objectively quantifying changes in eyelid morphology before and after surgery.\u003c/p\u003e\u003cp\u003eTherefore, addressing the clinical gap in pupil monitoring for patients with neurological disorders, closed-eye pupil monitoring technology leverages image analysis techniques based on deep learning. By utilizing the property of near-infrared light to penetrate the eyelids, this technology enables non-invasive dynamic capture\u003csup\u003e[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e. Its advantage lies in overcoming the reliance of traditional methods on eye-opening cooperation, thereby making continuous monitoring of comatose and sedated patients feasible. This technology provides real-time continuous data, reduces environmental light interference through physical shielding, and enables objective quantification of parameters such as pupil diameter and PLR. Based on this principle, our research team designed a novel closed-eye pupil monitoring device using temporal infrared projection-eyelid reflection imaging technology: light penetrates the eyeball and is reflected by the pupil, with the closed-eye pupil dynamics captured by a front-facing camera.\u003c/p\u003e\u003cp\u003eDespite the promising technical prospects, there is currently a lack of systematic validation data for this device in real-world populations with neurological disorders. There is an urgent need to clarify its clinical applicability, safety, and the specific pupillary parameter characteristics under closed-eye conditions. To this end, following preliminary testing conducted at the Department of Neurosurgery, Nanjing Drum Tower Hospital from December 2024 to January 2025, we are now initiating this prospective single-arm study to systematically assess the clinical feasibility and safety of this device in patients with neurological disorders. This study will provide critical empirical evidence to support the clinical translation of closed-eye pupil monitoring technology.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eStudy design\u003c/h2\u003e\u003cp\u003e This study was a prospective single-arm trial conducted at Nanjing Gulou Hospital on April 15, 2025, and approved by the Ethics Committee of Nanjing Drum Tower Hospital (2024-727-02). Informed consent was obtained from all individual participants included in the study. Since this study was a prospective, single-arm controlled trial, no statistical calculation of sample size is required. Subject screening was conducted in the neurosurgery inpatient department. The costs of subject examinations and any subsequent adverse event treatments were covered by the investigators, and there were no other incentives for participating in the study. Two experienced neurosurgery department heads independently screened and assessed participants using a consensus method. Participants meeting the minimum criteria for this study were enrolled and underwent examinations. All participants signed informed consent forms. Follow-up assessments were conducted 72 hours after collecting pupil data. The device's efficacy and accuracy were evaluated based on preoperative and postoperative comparisons of participant pupil diameter, while its safety was assessed through the incidence and severity of adverse events (AEs).\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003ePatients selection\u003c/h3\u003e\n\u003cp\u003eThe primary inclusion criteria are as follows: (1) Males and females aged 18 years or older; (2) Diagnosed with a neurological disorder (such as traumatic brain injury, cerebrovascular disease, neurodegenerative disease, etc.) by two senior neurosurgeons from the neurosurgery team; (3) Complete eyelid anatomy with no congenital malformations that impair infrared light penetration; (4) Voluntarily signed informed consent form.\u003c/p\u003e\u003cp\u003eIndividuals meeting any of the following criteria will be excluded from study participation: presence of ocular lesions (including iris abnormalities, active uveitis, cataracts, intraocular lens implantation within the past month, or eyelid defects/infections); presence of periorbital skin damage, implantation of an intracranial pressure monitoring device, or allergy to study-related materials. Terminally ill patients (expected survival period\u0026thinsp;\u0026lt;\u0026thinsp;48 hours), pregnant women, and individuals participating in other interventional clinical trials are also excluded.\u003c/p\u003e\n\u003ch3\u003eThe procedure\u003c/h3\u003e\n\u003cp\u003eDuring the testing process, we measured each volunteer's eyes one by one, and only operated on one eye at a time to ensure data accuracy and safety. An infrared pulsed Light-Emitting Diode (LED) was installed inside an opaque sleeve, which was tightly fitted to the volunteer's temple. The key step in the test was to determine the optimal position of the infrared LED on the temple. By continuously capturing infrared images of the pupil in the open-eye state, we dynamically adjusted the LED's position until a clear and distinguishable pupil image was obtained. After positioning was completed, we verified whether infrared light could penetrate the eyelid by having the volunteer close their eyes. To control variables, participants were required to maintain fixation on a fixed point throughout the test to minimize interference from pupil movement. To ensure data accuracy, volunteers were instructed to maintain visual fixation throughout the process to reduce pupil movement. By comparing images taken with eyes open and closed, we found that the eyelids absorb and scatter the infrared light emitted by the pupil. Two experienced operators manually calibrated the pupil contour with eyes open and calculated the equivalent diameter as the baseline value. We collected 1\u0026ndash;3 valid data points from each group of subjects, and all samples included in the analysis underwent strict quality control (invalid data with incomplete blinking movements or blurred images were excluded). Among the 50 volunteers, we successfully imaged the pupils of 44 closed-eye subjects. Among these 44 volunteers, we obtained at least one pair of images clearly showing the pupils before and after the white light pulse, totaling 185 such image pairs analyzed.\u003c/p\u003e\n\u003ch3\u003eOutcome assessments\u003c/h3\u003e\n\u003cp\u003eThe primary evaluation criteria are the safety and technical accuracy of the treatment. Safety is primarily assessed through an active telephone follow-up mechanism by the research team, tracking potential delayed adverse events within 24 to 72 hours after device use. Adverse events are reported and categorized by the investigators. If an adverse event results in death, life-threatening conditions, requires hospitalization or prolongs hospitalization, causes or may cause permanent damage to bodily structure or function, or causes or may cause fetal disease, congenital defects, or abnormalities in offspring, it is classified as a serious adverse event (SAE).\u003c/p\u003e\u003cp\u003eTechnical accuracy is assessed through diameter prediction error, segmentation performance, and model stability. In terms of diameter prediction, absolute error levels are evaluated using mean absolute error (MAE) and Root Mean Square Error (RMSE), and a systematic comparison framework between AI-automated measurements and manual gold standard measurements is established using the Bland-Altman consistency analysis method under both closed-eye and open-eye conditions. Image segmentation performance was evaluated using pixel accuracy (PixelAcc), Dice coefficient, and intersection-over-union (IoU). PixelAcc reflects overall classification accuracy, while Dice and IoU focus on assessing the precision of target region segmentation. Model stability analysis used the proportion of low-error samples as an indicator and calculated its 95% confidence interval. All data collection strictly followed standardized quality control procedures to ensure that abnormal data was screened and excluded prior to analysis.\u003c/p\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003eStatistical analysis\u003c/h2\u003e\u003cp\u003eDescriptive statistical analysis was performed on all parameters. Continuous variables were described using mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (Mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD) combined with median and extremes to characterize their distribution; The average absolute error (in millimeters) between closed-eye measurements and the manual gold standard was quantified using MAE, while RMSE was used to highlight the contribution of larger errors to the degree of dispersion. Additionally, the Bland-Altman four-step consistency analysis method was employed\u0026mdash;calculating the mean (d̄) and standard deviation of measurement differences, determining the 95% consistency limits (LoA), and plotting a difference-mean scatter plot\u0026mdash; \u0026mdash; using a clinically acceptable threshold (\u0026plusmn;\u0026thinsp;0.5 millimeters) for judgment. Image processing and data analysis were performed using Python 3.13. P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant (two-tailed test).\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003ePatient characteristics\u003c/h2\u003e\u003cp\u003eThis study included a total of 44 patients who met the inclusion criteria, with enrollment occurring between April 15, 2025, and July 10, 2025. Demographic characteristics of the patients were as follows(Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e): 29 males (65.91%) and 15 females (34.09%). Neurological examination results showed that 21 patients (47.73%) had normal muscle strength in all four limbs (Grade 5), 16 patients (36.36%) exhibited Grade 4 muscle strength in all four limbs, 6 patients (13.64%) presented with Grade 5 muscle strength in both upper limbs accompanied by Grade 4 muscle strength in both lower limbs, and 1 case (2.27%) had bilateral upper limb muscle strength of grade 4 with asymmetric lower limb muscle strength (right lower limb grade 4, left lower limb grade 5). In the pupillary light reflex assessment, 42 cases (95.45%) had a sensitive response to light, and 2 cases (4.55%) had a delayed response to light. Additionally, the study identified three patients with special conditions: one was in the postoperative edema phase, one had hearing impairment that did not affect experimental communication, and one developed eyelid thickening postoperatively due to an intracranial space-occupying lesion. None of these special conditions affected the normal conduct of the trial.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eGeneral information about patients\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBaseline\u003c/p\u003e\u003cp\u003echaracteristic\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e48.73\u0026thinsp;\u0026plusmn;\u0026thinsp;18.69\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGender\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e29(65.91%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e15(34.09%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBody Mass Index(BMI)/(kg/m\u003csup\u003e2\u003c/sup\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e24.07\u0026thinsp;\u0026plusmn;\u0026thinsp;4.44\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMuscle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNormal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e21(47.73%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMuscle strength in limbs: Grade 4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e16(36.36%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGrade 5 in both upper limbs, grade 4\u003c/p\u003e\u003cp\u003ein both lower limbs\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e6(13.64%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGrade 4 in both upper limbs, grade 4\u003c/p\u003e\u003cp\u003ein the right lower limb, and grade 5 in\u003c/p\u003e\u003cp\u003ethe left lower limb\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e1(2.27%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSurgery\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePreoperative\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e21(47.73%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePostoperative\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e23(52.27%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePLR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSensitive\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e42(95.45%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSluggish\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e2(4.55%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGlasgow Coma Scale(GCS)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eE4V5M6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e44(100%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiagnosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIntracranial space-occupying lesion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e23(52.27%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIntraspinal space-occupying lesion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e17(38.64%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOther neurological disorders\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e4(9.09%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eGlasgow Coma Scale(GCS)\u003c/em\u003e: Eye opening response (4 points),Verbal response (5 points), Motor response (6 points), A maximum score of 15 indicates clear consciousness; 13\u0026ndash;15 is mildly impaired consciousness; 9\u0026ndash;12 is moderately impaired consciousness; and 8 or less is coma.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eSafety\u003c/h3\u003e\n\u003cp\u003eNo device-related adverse events occurred, nor were there any SAEs.\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eAccuracy\u003c/h2\u003e\u003cp\u003eAfter the initial data collection was completed on July 10, we used \u0026lsquo;python predict.py --model-path E:\\models\\1\\850epoch_model.pth --input-dir [input directory] --output-dir [output directory] --batch-size 1 --threshold 0.68\u0026rsquo; to predict the collected mask images(Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). To address the prediction results, the research team fixed errors in the test program related to mask area and brightness calculations, removed channel dimension issues caused by \u0026lsquo;.squeeze()\u0026rsquo;, and tested models with 1000 to 3000 epochs, but found that there were misjudgments in the bright areas on both sides.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe research team then introduced attention mechanisms and residual connections, and adjusted the learning rate from 3e-4 to 1e-4. However, during testing, they still found that the accuracy was low, and speculated that this might be related to the model saving or calling process. Additionally, it was discovered that statistical differences in \u0026lsquo;BatchNorm2d\u0026rsquo; between training and testing (training batch_size\u0026thinsp;=\u0026thinsp;4, testing batch_size\u0026thinsp;=\u0026thinsp;1) caused model output instability. To address this, we replaced it with `InstanceNorm2d(track_running_stats\u0026thinsp;=\u0026thinsp;False)`, improving testing consistency. Additionally, the optimized model's binary threshold was adjusted from the original 0.50\u0026ndash;0.60 to 0.61\u0026ndash;0.71, and the test results of each epoch model were recorded in a CSV file for analysis. Ultimately, the 850-epoch model (threshold 0.68) (MAE of 184.1 pixels\u0026sup2;, MAPE of 9.1%) was determined to be the primary recommended model(Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn summary, the average error between the pupil diameter calculated based on the prediction mask (pred_d) and the true value (gt_d) is -0.40 pixels, with an error rate of 7.18%. Most prediction results are concentrated within \u0026plusmn;\u0026thinsp;2 pixels. The minimum error rate is 0.26%, the maximum error rate is 20.76%, and the standard deviation is 6.54%, indicating that some samples exhibit significant prediction deviations. The high error rates in edge prediction and certain samples (exceeding 19%) suggest that the model's stability in complex scenarios still requires optimization. Overall, the model performs reasonably well in measuring target diameter, but there is significant room for improvement in segmentation performance (Dice coefficient 0.47, IoU 0.33) and edge accuracy. Through statistical analysis of the model's prediction results, we found that the model performs well overall in radius prediction but still has room for optimization. The average deviation in radius prediction is -0.40 (error rate 7.18%), with a median deviation of -0.23 (error rate 4.06%), indicating a slight systematic underestimation bias in the model. The standard deviation of the error is 1.52 (error rate standard deviation 6.54%), suggesting that the prediction results remain relatively stable. Additionally, consistency analysis shows that the average difference (bias) between predicted and actual values is -0.13, with a standard deviation (SD) of 0.94, indicating a small overall bias and controllable variability. The 95% consistency limits range from \u0026minus;\u0026thinsp;1.98 to +\u0026thinsp;1.72, with 92.5% of samples falling within this range, which fully validates the reliability of the model's predictions. Overall, while the current model exhibits a slight underestimation trend, its predictive stability is good, with 92.5% of the prediction results falling within an acceptable error range. All outcomes have shown in the Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eModel statistical analysis\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEpoch\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eImange of samples\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMAE\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMAPE(%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eLow error sample ratio(%)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e850\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e185\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e184.1\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e9.1%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e65.9%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e199.3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e9.7%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e54.5%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2550\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e229.8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e12.3%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e38.6%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1550\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e209.5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e10.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e45.5%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e900\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e236.6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e11.3%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e52.3%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1700\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e212.0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e10.7%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e40.9%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e950\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e250.8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e12.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e50.0%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e800\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e384.0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e20.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e27.3%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1800\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e325.6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e16.3%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e22.7%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eOutcomes\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIndicators\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRadius error (err)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eError rate (err_rate)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003emean\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-0.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7.18%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003emedian\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-0.23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.06%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003estandard deviation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.52\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6.54%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMinimum value\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-3.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.26%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMaximum value\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e20.76%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMean difference (Bias)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-0.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e/\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStandard deviation of differences\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e/\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003elower limit of consistency\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-1.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e/\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUpper limit of consistency\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e+\u0026thinsp;1.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e/\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInclusion ratio\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e92.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e/\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe closed-eye pupil monitoring system developed in this study represents an innovative breakthrough in the field of neural monitoring technology. The system is based on the principles of near-infrared spectroscopy\u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e, leveraging the characteristics of this wavelength band, which has a deeper penetration depth in biological tissues and lower scattering effects compared to visible light, providing a physical foundation for obtaining clearer images of the pupil in a closed-eye state. Additionally, this study employs an experimental protocol involving infrared light illumination of the temples\u003csup\u003e[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/sup\u003e, avoiding the potential visual discomfort and potential eye damage caused by direct light exposure in traditional pupil measurement methods, while enhancing the comfort experience of participants during actual testing.\u003c/p\u003e\u003cp\u003eIn terms of algorithm performance, the closed-eye pupil measurement system developed in this study achieved clinically significant measurement accuracy, with an average error rate of 7.18%, comparable to the 5\u0026ndash;8% error rate of currently widely used quantitative infrared pupillometers\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]\u003c/sup\u003e. Previous studies\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e have shown that traditional subjective visual estimation methods have an error rate as high as 19% within the pupil size range of 2\u0026ndash;4 mm, making it prone to missing diagnoses of anisocoria and weak pupillary light reflexes. Quantitative infrared pupillometers, on the other hand, measure parameters such as pupil size, contraction latency, and speed, and generate a Neural Pupillary Index (NPI), providing objective evidence for clinical assessment\u003csup\u003e[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u003c/sup\u003e. A significant improvement in this study was the replacement of BatchNorm2d with InstanceNorm2d. This adjustment effectively addressed the issue of limited batch size in clinical settings due to patient privacy protection and data acquisition restrictions, enhancing the system's universality. However, in this study, we also observed individual cases with a maximum error rate of 20.76%, which, upon analysis, may be related to individual anatomical differences(such as eyelid thickness and subcutaneous fat distribution). One patient with post-surgical eyelid edema following intracranial space-occupying lesion surgery exhibited abnormal eyelid thickening and delayed light reflexes. During mask image processing, we may have introduced biases, leading to an increase in the maximum error rate. This suggests that future research should establish personalized correction models for eyelid optical characteristics to further optimize system performance.\u003c/p\u003e\u003cp\u003eThis study fills a technical gap in non-invasive monitoring within the field of neurocritical care. Compared to traditional methods such as invasive intracranial pressure monitoring\u003csup\u003e[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]\u003c/sup\u003e, this experiment is the first to systematically validate the feasibility of closed-eye measurements in a real clinical setting, overcoming the limitations of previous studies that were primarily focused on pupil monitoring in critically ill patients with their eyes open\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e; Additionally, an end-to-end deep learning architecture was employed to enhance the performance of traditional image processing pipelines. Furthermore, all data were collected from real clinical scenarios, making the findings more clinically applicable. The results of this study show that 92.5% of the predicted values fall within the consistency range, demonstrating clinical reference value. Notably, no adverse events or serious adverse events were observed throughout the study, and its safety has been certified in accordance with the IEC 62471 photobiological safety standard. However, the study results indicate that the Dice coefficient is 0.47, suggesting that the current algorithm used in the image segmentation process has limitations. By analyzing the provided mask image samples, we believe the primary errors stem from: 1) partial volume effects at the eyelid-sclera junction; 2) shadow artifacts caused by eyelashes; 3) motion blur caused by eyelid micro-movements. To address these technical bottlenecks, we will implement multiple measures in future studies, including incorporating temporal information to leverage video continuity features, adopting adversarial learning strategies to enhance edge retention capabilities, and integrating optical coherence tomography (OCT) technology for auxiliary calibration.\u003c/p\u003e\u003cp\u003eAlthough this study has made significant progress in validating clinical feasibility and safety, limitations remain. The study design was a single-arm trial with a small sample size from a single medical center, which may affect the representativeness and generalizability of the results. Participants excluded patients with eyelid defects or drug interference, and quantification methods for dynamic parameters (e.g., PLR velocity) are not yet mature. Additionally, short-term follow-up cannot assess long-term safety.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study confirms that a deep learning-based closed-eye pupil monitoring system can achieve non-invasive pupil measurement in patients with neurological diseases with high accuracy and good safety, with no device-related adverse events observed. This technology overcomes the reliance of traditional pupil monitoring on patient cooperation with open eyes, providing an innovative continuous monitoring solution for patients who cannot cooperate with examinations, such as those in a coma or under sedation, and holds significant clinical translation value. Although image segmentation accuracy and measurement performance under extreme anatomical conditions (such as abnormally thickened eyelids) still require optimization, the overall performance has already reached a clinically acceptable range. Future research will involve multi-center, large-scale clinical validation to enhance the reliability of research conclusions; targeted analyses for specific neurological conditions (such as brain injury, epilepsy, Alzheimer's disease, etc.) to establish disease-specific pupil parameter profiles; and refinement of quantification methods for dynamic parameters (such as PLR velocity and contraction latency) to enhance the comprehensiveness of neurological function assessment.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u0026nbsp;\u003c/strong\u003eNone.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u0026nbsp;\u003c/strong\u003eZYS, STC and ZXH designed research, performed research and analyzed data; \u0026nbsp;ZYS, JYL, SL and LWZ wrote the paper. BX, XYB and LC peformed research. All authors contributed todrafting of the original manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eProject of China Hospital Reform and Development Research Institute, Nanjing University;\u003c/p\u003e\n\u003cp\u003eAid project of Jiangsu Ningai Medical Development \u0026amp; Medical Aid Foundation(NDYGN2024017);\u003c/p\u003e\n\u003cp\u003e2023 Open Research Fund for the Key Laboratory of the Ministry of Education for Geriatric Long-Term Care (Naval Military Medical University)(LNYB-2023-13)\u003c/p\u003e\n\u003cp\u003e2023 Special Funds Project for Transformation of Scientific and Technological Achievements of Nanjing Gulou Hospital(2023-GNYZ-YB-07);\u003c/p\u003e\n\u003cp\u003e2024 Nanjing Health Science and Technology Development Special Funds Grant(ZKX24025)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u0026nbsp;\u003c/strong\u003eThe datasets generated and analyzed during the current study are not publicly available due to restrictions specified in the signed partnership agreement. However, they can be accessed from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e Approved by the Ethics Committee of Nanjing Drum Tower Hospital (2024-727-02) and informed consent was obtained from all individual participants included in the study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrevious presentations\u003c/strong\u003e None.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number\u003c/strong\u003e\u0026nbsp; ChiCTR2500105504\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eAcknowledgements:Project of China Hospital Reform and Development Research Institute, Nanjing University;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAid project of Jiangsu Ningai Medical Development \u0026amp; Medical Aid Foundation(NDYGN2024017);\u003c/p\u003e\n\u003cp\u003e2023 Open Research Fund for the Key Laboratory of the Ministry of Education for Geriatric Long-Term Care (Naval Military Medical University)(LNYB-2023-13)\u003c/p\u003e\n\u003cp\u003e2023 Special Funds Project for Transformation of Scientific and Technological Achievements of Nanjing Gulou Hospital(2023-GNYZ-YB-07);\u003c/p\u003e\n\u003cp\u003e2024 Nanjing Health Science and Technology Development Special Funds Grant(ZKX24025)\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eGrujic, N., Polania, R., \u0026amp; Burdakov, D. (2024). Neurobehavioral meaning of pupil size. \u003cem\u003eNeuron\u003c/em\u003e, \u003cem\u003e112\u003c/em\u003e(20), 3381～3395. https://doi.org/10.1016/j.neuron.2024.05.029.\u003c/li\u003e\n\u003cli\u003eThe autonomic nervous system and the eye. (1985). \u003cem\u003eLancet (London, England)\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(8455), 591\u0026ndash;592.\u003c/li\u003e\n\u003cli\u003eWang, Y., Zekveld, A. A., Naylor, G., Ohlenforst, B., Jansma, E. P., Lorens, A., Lunner, T., \u0026amp; Kramer, S. E. (2016). Parasympathetic Nervous System Dysfunction, as Identified by Pupil Light Reflex, and Its Possible Connection to Hearing Impairment. \u003cem\u003ePLOS ONE\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(4), e0153566. https://doi.org/10.1371/journal.pone.0153566.\u003c/li\u003e\n\u003cli\u003eChoi, S. C., Narayan, R. K., Anderson, R. L., \u0026amp; Ward, J. D. (1988). Enhanced specificity of prognosis in severe head injury. \u003cem\u003eJournal of neurosurgery,\u003c/em\u003e 69(3), 381\u0026ndash;385. https://doi.org/10.3171/jns.1988.69.3.0381.\u003c/li\u003e\n\u003cli\u003eEmami, P., Czorlich, P., Fritzsche, F. S., Westphal, M., Rueger, J. M., Lefering, R., \u0026amp; Hoffmann, M. (2017). Impact of Glasgow Coma Scale score and pupil parameters on mortality rate and outcome in pediatric and adult severe traumatic brain injury: A retrospective, multicenter cohort study. \u003cem\u003eJournal of Neurosurgery,\u003c/em\u003e 126(3), 760～767. https://doi.org/10.3171/2016.1.jns152385.\u003c/li\u003e\n\u003cli\u003eMarmarou, A., Lu, J., Butcher, I., McHugh, G. S., Murray, G. D., Steyerberg, E. W., Mushkudiani, N. A., Choi, S., \u0026amp; Maas, A. I. R. (2007). Prognostic Value of The Glasgow Coma Scale And Pupil Reactivity in Traumatic Brain Injury Assessed Pre-Hospital And on Enrollment: An IMPACT Analysis. \u003cem\u003eJournal of Neurotrauma\u003c/em\u003e, \u003cem\u003e24\u003c/em\u003e(2), 270～280. https://doi.org/10.1089/neu.2006.0029.\u003c/li\u003e\n\u003cli\u003eGreer, D. M., Shemie, S. D., Lewis, A., Torrance, S., Varelas, P., Goldenberg, F. D., Bernat, J. L., Souter, M., Topcuoglu, M. A., Alexandrov, A. W., Baldisseri, M., Bleck, T., Citerio, G., Dawson, R., Hoppe, A., Jacobe, S., Manara, A., Nakagawa, T. A., Pope, T. M., Silvester, W., \u0026hellip; Sung, G. (2020). Determination of Brain Death/Death by Neurologic Criteria: The World Brain Death Project. \u003cem\u003eJAMA\u003c/em\u003e, 324(11), 1078\u0026ndash;1097. https://doi.org/10.1001/jama.2020.11586.\u003c/li\u003e\n\u003cli\u003eWilhelm H. (2011). Disorders of the pupil. \u003cem\u003eHandbook of clinical neurology,\u003c/em\u003e 102, 427\u0026ndash;466. https://doi.org/10.1016/B978-0-444-52903-9.00022-4\u003c/li\u003e\n\u003cli\u003eKawasaki A. (2025). Disorders of the Pupil. \u003cem\u003eContinuum (Minneapolis, Minn.),\u003c/em\u003e 31(2), 527\u0026ndash;542. https://doi.org/10.1212/cont.0000000000001542\u003c/li\u003e\n\u003cli\u003eBarry, C., \u0026amp; Wang, E. (2023). Racially fair pupillometry measurements for RGB smartphone cameras using the far red spectrum. \u003cem\u003eScientific reports\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(1), 13841. https://doi.org/10.1038/s41598-023-40796-0.\u003c/li\u003e\n\u003cli\u003eKolias, A. G., Adams, H., Timofeev, I. S., Corteen, E. A., Hossain, I., Czosnyka, M., Timothy, J., Anderson, I., Bulters, D. O., Belli, A., Eynon, C. A., Wadley, J., Mendelow, A. D., Mitchell, P. M., Wilson, M. H., Critchley, G., Sahuquillo, J., Unterberg, A., Posti, J. P., \u0026hellip; Buxton, M. (2022). Evaluation of Outcomes Among Patients With Traumatic Intracranial Hypertension Treated With Decompressive Craniectomy vs Standard Medical Care at 24 Months: A Secondary Analysis of the RESCUEicp Randomized Clinical Trial. \u003cem\u003eJAMA Neurology\u003c/em\u003e, \u003cem\u003e79\u003c/em\u003e(7), 664. https://doi.org/10.1001/jamaneurol.2022.1070.\u003c/li\u003e\n\u003cli\u003eOddo, M., Taccone, F. S., Petrosino, M., Badenes, R., Blandino-Ortiz, A., Bouzat, P., Caricato, A., Chesnut, R. M., Feyling, A. C., Ben-Hamouda, N., Hemphill, J. C., Koehn, J., Rasulo, F., Suarez, J. I., Elli, F., Vargiolu, A., Rebora, P., Galimberti, S., Citerio, G., \u0026hellip; Wang, R. (2023). The Neurological Pupil index for outcome prognostication in people with acute brain injury (ORANGE): A prospective, observational, multicentre cohort study. \u003cem\u003eThe Lancet Neurology\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e(10), 925～933. https://doi.org/10.1016/S1474-4422(23)00271-5.\u003c/li\u003e\n\u003cli\u003ePetrosino, M., Gouv\u0026ecirc;a Bogossian, E., Rebora, P., Galimberti, S., Chesnut, R., Bouzat, P., Oddo, M., Taccone, F. S., Citerio, G., \u0026amp; ORANGE Participants and Investigators. (2025). Neurological Pupil Index and Intracranial Hypertension in Patients With Acute Brain Injury: A Secondary Analysis of the ORANGE Study. \u003cem\u003eJAMA Neurology\u003c/em\u003e, \u003cem\u003e82\u003c/em\u003e(2), 176. https://doi.org/10.1001/jamaneurol.2024.4189.\u003c/li\u003e\n\u003cli\u003eCimbalnik, J., Dolezal, J., Top\u0026ccedil;u, \u0026Ccedil;., Lech, M., Marks, V. S., Joseph, B., Dobias, M., Van Gompel, J., Worrell, G., \u0026amp; Kucewicz, M. (2022). Intracranial electrophysiological recordings from the human brain during memory tasks with pupillometry. \u003cem\u003eScientific Data\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(1). https://doi.org/10.1038/s41597-021-01099-z.\u003c/li\u003e\n\u003cli\u003eLiu, X., Hike, D., Choi, S., Man, W., Ran, C., Zhou, X. A., Jiang, Y., \u0026amp; Yu, X. (2024). Identifying the bioimaging features of Alzheimer\u0026rsquo;s disease based on pupillary light response-driven brain-wide fMRI in awake mice. \u003cem\u003eNature Communications\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(1), 9657. https://doi.org/10.1038/s41467-024-53878-y.\u003c/li\u003e\n\u003cli\u003eCouret, D., Boumaza, D., Grisotto, C., Triglia, T., Pellegrini, L., Ocquidant, P., Bruder, N. J., \u0026amp; Velly, L. J. (2016). Reliability of standard pupillometry practice in neurocritical care: An observational, double-blinded study. \u003cem\u003eCritical Care\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e(1), 99. https://doi.org/10.1186/s13054-016-1239-z.\u003c/li\u003e\n\u003cli\u003eFavre, E., Bernini, A., Morelli, P., Pasquier, J., Miroz, J.-P., Abed-Maillard, S., Ben-Hamouda, N., \u0026amp; Oddo, M. (2020). Neuromonitoring of delirium with quantitative pupillometry in sedated mechanically ventilated critically ill patients. \u003cem\u003eCritical Care\u003c/em\u003e, \u003cem\u003e24\u003c/em\u003e(1). https://doi.org/10.1186/s13054-020-2796-8.\u003c/li\u003e\n\u003cli\u003eChan, W. P., Prescott, B. R., Barra, M. E., Chung, D. Y., Kim, I. S., Saglam, H., Hutch, M. R., Shin, M., Zafar, S. F., Benjamin, E. J., Smirnakis, S. M., Dupuis, J., Greer, D. M., \u0026amp; Ong, C. J. (2022). Dexmedetomidine and Other Analgosedatives Alter Pupil Characteristics in Critically Ill Patients. \u003cem\u003eCritical Care Explorations\u003c/em\u003e, \u003cem\u003e4\u003c/em\u003e(5), e0691. https://doi.org/10.1097/cce.0000000000000691.\u003c/li\u003e\n\u003cli\u003eGouv\u0026ecirc;a Bogossian, E., Blandino Ortiz, A., Esposito, V., Caricato, A., Righy Shinotsuka, C., Monl\u0026eacute;on Lopez, B., Giann\u0026igrave;, G., Macchini, E., De Pablo Sanchez, R., Pisapia, L., Turon, R., Gon\u0026ccedil;alves, B., Badenes, R., Kurtz, P., \u0026amp; Taccone, F. S. (2023). Neurological Pupil Index and Delayed Cerebral Ischemia after Subarachnoid Hemorrhage: A Retrospective Multicentric Study. \u003cem\u003eNeurocritical Care\u003c/em\u003e, \u003cem\u003e39\u003c/em\u003e(1), 116～124. https://doi.org/10.1007/s12028-023-01744-y.\u003c/li\u003e\n\u003cli\u003eOng, C., Hutch, M., Barra, M., Kim, A., Zafar, S., \u0026amp; Smirnakis, S. (2019). Effects of Osmotic Therapy on Pupil Reactivity: Quantification Using Pupillometry in Critically Ill Neurologic Patients. \u003cem\u003eNeurocritical Care\u003c/em\u003e, \u003cem\u003e30\u003c/em\u003e(2), 307～315. https://doi.org/10.1007/s12028-018-0620-y.\u003c/li\u003e\n\u003cli\u003eOlson, D. M., Stutzman, S., Saju, C., Wilson, M., Zhao, W., \u0026amp; Aiyagari, V. (2016). Interrater Reliability of Pupillary Assessments. \u003cem\u003eNeurocritical Care\u003c/em\u003e, \u003cem\u003e24\u003c/em\u003e(2), 251～257. https://doi.org/10.1007/s12028-015-0182-1.\u003c/li\u003e\n\u003cli\u003eRobba, C., Graziano, F., Rebora, P., Elli, F., Giussani, C., Oddo, M., Meyfroidt, G., Helbok, R., Taccone, F. S., Prisco, L., Vincent, J.-L., Suarez, J. I., Stocchetti, N., Citerio, G., Abdelaty, M., Abed Maillard, S., Ahmed, H., Albrecht, L., Alsudani, A., \u0026hellip; Zerbi, S. M. (2021). Intracranial pressure monitoring in patients with acute brain injury in the intensive care unit (SYNAPSE-ICU): An international, prospective observational cohort study. \u003cem\u003eThe Lancet Neurology\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e(7), 548～558. https://doi.org/10.1016/s1474-4422(21)00138-1.\u003c/li\u003e\n\u003cli\u003eLou, L., Cao, J., Wang, Y., Gao, Z., Jin, K., Xu, Z., Zhang, Q., Huang, X., \u0026amp; Ye, J. (2021). Deep learning-based image analysis for automated measurement of eyelid morphology before and after blepharoptosis surgery. Annals of medicine, 53(1), 2278\u0026ndash;2285. https://doi.org/10.1080/07853890.2021.2009127.\u003c/li\u003e\n\u003cli\u003eFarraj, Y., Buxboim, A., Cohen, J. E., Kan-Tor, Y., Glasner Hagege, S., Weiss, D., Goldman, V., \u0026amp; Beatus, T. (2021). Measuring pupil size and light response through closed eyelids. \u003cem\u003eBiomedical Optics Express\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(10), 6485. https://doi.org/10.1364/BOE.435508.\u003c/li\u003e\n\u003cli\u003eAfara, I. O., Shaikh, R., Nippolainen, E., Querido, W., Torniainen, J., Sarin, J. K., Kandel, S., Pleshko, N., \u0026amp; T\u0026ouml;yr\u0026auml;s, J. (2021). Characterization of connective tissues using near-infrared spectroscopy and imaging. \u003cem\u003eNature Protocols\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(2), 1297～1329. https://doi.org/10.1038/s41596-020-00468-z.\u003c/li\u003e\n\u003cli\u003eByrne, H. J., Knief, P., Keating, M. E., \u0026amp; Bonnier, F. (2016). Spectral pre and post processing for infrared and Raman spectroscopy of biological tissues and cells. \u003cem\u003eChemical Society Reviews\u003c/em\u003e, \u003cem\u003e45\u003c/em\u003e(7), 1865～1878. https://doi.org/10.1039/C5CS00440C.\u003c/li\u003e\n\u003cli\u003eGordon, K. B., Char, D. H., \u0026amp; Sagerman, R. H. (1995). Late effects of radiation on the eye and ocular adnexa. \u003cem\u003eInternational journal of radiation oncology, biology, physics\u003c/em\u003e, \u003cem\u003e31\u003c/em\u003e(5), 1123～1139. https://doi.org/10.1016/0360-3016(95)00062-4.\u003c/li\u003e\n\u003cli\u003eT, K., E, T., J, B., \u0026amp; Em, K. (2003). Comparison of a digital and a handheld infrared pupillometer for determining scotopic pupil diameter. \u003cem\u003eJournal of Cataract and Refractive Surgery\u003c/em\u003e, \u003cem\u003e29\u003c/em\u003e(1). https://doi.org/10.1016/s0886-3350(02)01898-9.\u003c/li\u003e\n\u003cli\u003eBoulter, J. H., Shields, M. M., Meister, M. R., Murtha, G., Curry, B. P., \u0026amp; Dengler, B. A. (2021). The Expanding Role of Quantitative Pupillometry in the Evaluation and Management of Traumatic Brain Injury. \u003cem\u003eFrontiers in Neurology\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e, 685313. https://doi.org/10.3389/fneur.2021.685313.\u003c/li\u003e\n\u003cli\u003eP, M., M, O., \u0026amp; N, B.-H. (2019). Role of automated pupillometry in critically ill patients. \u003cem\u003eMinerva Anestesiologica\u003c/em\u003e, \u003cem\u003e85\u003c/em\u003e(9). https://doi.org/10.23736/S0375-9393.19.13437-2.\u003c/li\u003e\n\u003cli\u003eGwj, H., G, C., P, H., A, K., G, M., C, R., N, S., \u0026amp; R, C. (2022). Intracranial pressure: Current perspectives on physiology and monitoring. \u003cem\u003eIntensive Care Medicine\u003c/em\u003e, \u003cem\u003e48\u003c/em\u003e(10). https://doi.org/10.1007/s00134-022-06786-y.\u003c/li\u003e\n\u003cli\u003eKb, E., \u0026amp; Pk, E. (2020). Measuring intracranial pressure by invasive, less invasive or non-invasive means: Limitations and avenues for improvement. \u003cem\u003eFluids and Barriers of the CNS\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e(1). https://doi.org/10.1186/s12987-020-00195-3.\u003c/li\u003e\n\u003cli\u003eMm, B., Aj, S., Jc, X., S, S.-N., W, Y., \u0026amp; Li, G. (2021). Quantitative Pupillometry in the Intensive Care Unit. \u003cem\u003eJournal of Intensive Care Medicine\u003c/em\u003e, \u003cem\u003e36\u003c/em\u003e(4). https://doi.org/10.1177/0885066619881124.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Closed-eye pupil monitoring, Neurological diseases, Near-infrared light, Deep learning, Non-invasive monitoring","lastPublishedDoi":"10.21203/rs.3.rs-7423596/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7423596/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cb\u003eObjective\u003c/b\u003e This study aims to develop a non-invasive dynamic pupil monitoring system for patients with neurological disorders. Traditional methods (such as flashlight examination or infrared devices) rely on patients keeping their eyes open, making them unsuitable for comatose, sedated, or patients with abnormal eyelids, and they also have limitations in intermittent monitoring and subjective interpretation. This study leverages the ability of near-infrared light to penetrate eyelids, combined with deep learning technology, to design a novel closed-eye monitoring device that achieves dynamic capture through sequential infrared projection-eyelid reflection imaging.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMethods\u003c/b\u003e A prospective single-arm trial design was adopted, enrolling 44 patients in the Department of Neurosurgery at Nanjing Drum Tower Hospital from April to July 2025. Safety was assessed by the incidence of adverse events, and technical accuracy was quantified using diameter prediction error and image segmentation performance.\u003c/p\u003e\u003cp\u003e\u003cb\u003eResults\u003c/b\u003e Results showed: the incidence of device-related adverse events was zero, the average error rate for pupil diameter prediction was 7.18%, and 92.5% of prediction values fell within the consistency range, indicating good model stability. However, image segmentation performance (Dice coefficient 0.47) and accuracy under extreme anatomical conditions still require optimization.\u003c/p\u003e\u003cp\u003e\u003cb\u003eConclusion\u003c/b\u003e This system enables high-precision, safe, and non-invasive pupil monitoring for patients with neurological diseases, overcoming the reliance on patient cooperation in traditional methods, and providing an innovative solution for those unable to cooperate with examinations. Future studies should further validate reliability through multi-center, large-scale trials and optimize dynamic parameter quantification methods for specific neurological diseases.\u003c/p\u003e\u003cp\u003eRegistry: ChiCTR, TRN: ChiCTR2500105504, Registration date: 1 January 2024.\u003c/p\u003e","manuscriptTitle":"Closed-eye pupil monitoring system in patients with neurological disorders: a prospective, single-arm study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-11 12:21:25","doi":"10.21203/rs.3.rs-7423596/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"cbf040b2-d6e0-47cf-819e-f11eb2bf4f6e","owner":[],"postedDate":"September 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-10-01T14:56:05+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-11 12:21:25","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7423596","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7423596","identity":"rs-7423596","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}