A model for oversight of rare disease studies: The 25-year experience of the cystic fibrosis foundation data safety monitoring board.

This manuscript addresses the development and operating procedures of the Cystic Fibrosis Foundation Data Safety Monitoring Board (CFF-DSMB) and its role in the development and approval of new therapies through complex clinical trials with an emphasis on ensuring patient safety and study integrity. The authors describe the processes that have been developed over the last 25 years including the development of educational curricula for DSMB members and patient representation on DSMBs. The experience and success of the CFF-DSMB can serve as a model for providing high quality oversight of clinical trials for other groups who are dedicated to developing treatments for rare and complex diseases.

Microengineering 3D Collagen Matrices with Tumor-Mimetic Gradients in Fiber Alignment.

Collagen fibers in the 3D tumor microenvironment (TME) exhibit complex alignment landscapes that are critical in directing cell migration through a process called contact guidance. Previous in vitro work studying this phenomenon has focused on quantifying cell responses in uniformly aligned environments. However, the TME also features short-range gradients in fiber alignment that result from cell-induced traction forces. Although the influence of graded biophysical taxis cues is well established, cell responses to physiological alignment gradients remain largely unexplored. In this work, fiber alignment gradients in biopsy samples are characterized and recreated using a new microfluidic biofabrication technique to achieve tunable sub-millimeter to millimeter scale gradients. This study represents the first successful engineering of continuous alignment gradients in soft, natural biomaterials. Migration experiments on graded alignment show that HUVECs exhibit increased directionality, persistence, and speed compared to uniform and unaligned fiber architectures. Similarly, patterned MDA-MB-231 aggregates exhibit biased migration toward increasing fiber alignment, suggesting a role for alignment gradients as a taxis cue. This user-friendly approach, requiring no specialized equipment, is anticipated to offer new insights into the biophysical cues that cells interpret as they traverse the extracellular matrix, with broad applicability in healthy and diseased tissue environments.

Association between CFTR modulators and changes in iron deficiency markers in cystic fibrosis.

BACKGROUND: Iron deficiency (ID) is a common extrapulmonary manifestation in cystic fibrosis (CF). CF transmembrane conductance regulator (CFTR) modulator therapies, particularly highly-effective modulator therapy (HEMT), have drastically improved health status in a majority of people with CF. We hypothesize that CFTR modulator use is associated with improved markers of ID. METHODS: In a multicenter retrospective cohort study across 4 United States CF centers 2012-2022, the association between modulator therapies and ID laboratory outcomes was estimated using multivariable linear mixed effects models overall and by key subgroups. Summary statistics describe the prevalence and trends of ID, defined a priori as transferrin saturation (TSAT) <20 % or serum iron <60 µg/dL (<10.7 µmol/L). RESULTS: A total of 568 patients with 2571 person-years of follow-up were included in analyses. Compared to off modulator therapy, HEMT was associated with +8.4 % TSAT (95 % confidence interval [CI], +6.3-10.6 %; p < 0.0001) and +34.4 µg/dL serum iron (95 % CI, +26.7-42.1 µg/dL; p < 0.0001) overall; +5.4 % TSAT (95 % CI, +2.8-8.0 %; p = 0.0001) and +22.1 µg/dL serum iron (95 % CI, +13.5-30.8 µg/dL; p < 0.0001) in females; and +11.4 % TSAT (95 % CI, +7.9-14.8 %; p < 0.0001) and +46.0 µg/dL serum iron (95 % CI, +33.3-58.8 µg/dL; p < 0.0001) in males. Ferritin was not different in those taking modulator therapy relative to off modulator therapy. Hemoglobin was overall higher with use of modulator therapy. The prevalence of ID was high throughout the study period (32.8 % in those treated with HEMT). CONCLUSIONS: ID remains a prevalent comorbidity in CF, despite availability of HEMT. Modulator use, particularly of HEMT, is associated with improved markers for ID (TSAT, serum iron) and anemia (hemoglobin).

Comparing Simulation Training of Bronchoscopy-Guided Percutaneous Dilatational Tracheostomy Using Conventional Versus 3D Printed Simulators (TRAC-Sim Study).

Background: Individual implementation rate of bronchoscopy-guided percutaneous dilatational tracheostomy (PDT) varies among intensivists. Simulation training (ST) can increase the safety of medical procedures by reducing stress levels of the performing team. The aim of this study was to evaluate the benefit of ST in PDT regarding procedural time, quality of performance, and percepted feelings of safety of the proceduralist and to compare conventional simulators (CSIM) with simulators generated from 3D printers (3DSIM). Methods: We conducted a prospective, single-center, randomized, blinded cross-over study comparing the benefit of CSIM versus 3DSIM for ST of PDT. Participants underwent a standardized theoretical training and were randomized to ST with CSIM (group A) or 3DSIM (group B). After ST, participants' performance was assessed by two blinded examiners on a porcine trachea regarding time required for successful completion of PDT and correct performance (assessed by a performance score). Percepted feelings of safety were assessed before and after ST. This was followed by a second training and second assessment of the same aspects with crossed groups. Results: 44 participants were included: 24 initially trained with CSIM (group A) and 20 with 3DSIM (group B). Correctness of the PDT performance increased significantly in group B (p < .01) and not significantly in group A (p = .14). Mean procedural time required for performing a PDT after their second ST compared to the first assessment (p < .01) was lower with no difference between group A and group B and irrespective of the participants' previous experience regarding PDT, age, and sex. Moreover, percepted feelings of safety increased after the first ST in both groups (p < .001). Conclusions: ST can improve procedural skills, procedural time, and percepted feelings of safety of the proceduralist in simulated PDT.

The Diagnostic Utility of Repeat Fine-needle Aspirations of Benign Thyroid Nodules.

Objective This study aims to analyze the diagnostic utility of multiple repeat FNA on thyroid nodules with initially benign diagnosis. Methods In a 5-year period, 1658 thyroid nodules with initially benign FNAs were retrospectively reviewed and followed for subsequent resection and repeat biopsy. Results Out of 2150 thyroid nodules, 1658 (77.1%) were diagnosed as benign on FNAs. The average age was 57.4 years (range 11-93 years), and most were females (83.8%). Repeat FNA was performed on 183 benign nodules, of which 141 (8.5%) were sampled a second time and 42 (2.5%) had 2 or more repeat samplings. For the benign nodules without repeat FNAs, 124 had benign resection. Of cases with one-time repeat FNA, most (n=101) remained benign on repeat FNAs, 13 of which were benign on resection. Eleven had atypical repeat FNAs, 5 were resected, 4 of which were benign and one was atypical follicular neoplasm with HRAS and TERT promoter mutations. Of cases with multiple repeat FNA, most (n=35) were still benign on repeat FNAs, one had benign resection. Two had atypical repeat biopsies, one was PTC on resection with CCD6::RET fusion. The positive predictive value significantly decreased from 41.1% on single FNA to 8.3% on one-time repeat (p<0.001) and 16.7% on multiple repeat (p=0.002). The total cost for workup of previously benign nodules was $285,454. Conclusions Repeat FNA biopsies did not provide an additional diagnostic value in the evaluation of benign thyroid nodules, and often led to unwarranted follow-up procedures and significantly increased health care cost.

Education of schoolchildren in cardiopulmonary resuscitation - overview of the current literature.

PURPOSE OF REVIEW: Recognition of cardiac arrest and initiation of cardiopulmonary resuscitation (CPR) can be learned and adequately replicated by schoolchildren. Regular instruction of schoolchildren in CPR is therefore a core element to increase low bystander CPR rates. Thereby, schoolchildren CPR training evolved as own scientific field within the last decade. Aim was to describe current evidence in terms of epidemiology, teaching approaches and political aspects. RECENT FINDINGS: Schoolchildren demonstrate a high motivation to be trained in CPR. Teaching approaches that combine theoretical and practical learning sessions guarantee a sustainable learning effect. Schoolchildren can adequately perform chest compressions and mouth-to-mouth ventilation from the age of 12 years. Use of digital media is a highly promising teaching approach. CPR training conducted by teachers from the own school is effective and guarantees continuous development of CPR skills. Integration of schoolchildren CPR training into school curricula is the foundation for a sustainable increase of lay resuscitation rates in the population. Scientific and political promotion of schoolchildren CPR training is needed to sensitize the population and move bystander CPR in the social focus. SUMMARY: While bystander CPR rates are low in Europe comprehensive establishment of schoolchildren CPR training may sustainably increase survival after cardiac arrest.

Resource selection by New York City deer reveals the effective interface between wildlife, zoonotic hazards and humans.

Although the role of host movement in shaping infectious disease dynamics is widely acknowledged, methodological separation between animal movement and disease ecology has prevented researchers from leveraging empirical insights from movement data to advance landscape scale understanding of infectious disease risk. To address this knowledge gap, we examine how movement behaviour and resource utilization by white-tailed deer (Odocoileus virginianus) determines blacklegged tick (Ixodes scapularis) distribution, which depend on deer for dispersal in a highly fragmented New York City borough. Multi-scale hierarchical resource selection analysis and movement modelling provide insight into how deer's movements contribute to the risk landscape for human exposure to the Lyme disease vector-I. scapularis. We find deer select highly vegetated and accessible residential properties which support blacklegged tick survival. We conclude the distribution of tick-borne disease risk results from the individual resource selection by deer across spatial scales in response to habitat fragmentation and anthropogenic disturbances.

Inpainting surgical occlusion from laparoscopic video sequences for robot-assisted interventions.

Purpose: Medical technology for minimally invasive surgery has undergone a paradigm shift with the introduction of robot-assisted surgery. However, it is very difficult to track the position of the surgical tools in a surgical scene, so it is crucial to accurately detect and identify surgical tools. This task can be aided by deep learning-based semantic segmentation of surgical video frames. Furthermore, due to the limited working and viewing areas of these surgical instruments, there is a higher chance of complications from tissue injuries (e.g., tissue scars and tears). Approach: With the aid of digital inpainting algorithms, we present an application that uses image segmentation to remove surgical instruments from laparoscopic/endoscopic video. We employ a modified U-Net architecture (U-NetPlus) to segment the surgical instruments. It consists of a redesigned decoder and a pre-trained VGG11 or VGG16 encoder. The decoder was modified by substituting an up-sampling operation based on nearest-neighbor interpolation for the transposed convolution operation. Furthermore, these interpolation weights do not need to be learned to perform upsampling, which eliminates the artifacts generated by the transposed convolution. In addition, we use a very fast and adaptable data augmentation technique to further enhance performance. The instrument segmentation mask is filled in (i.e., inpainted) by the tool removal algorithms using the previously acquired tool segmentation masks and either previous instrument-containing frames or instrument-free reference frames. Results: We have shown the effectiveness of the proposed surgical tool segmentation/removal algorithms on a robotic instrument dataset from the MICCAI 2015 and 2017 EndoVis Challenge. We report a 90.20% DICE for binary segmentation, a 76.26% DICE for instrument part segmentation, and a 46.07% DICE for instrument type (i.e., all instruments) segmentation on the MICCAI 2017 challenge dataset using our U-NetPlus architecture, outperforming the results of earlier techniques used and tested on these data. In addition, we demonstrated the successful execution of the tool removal algorithm from surgical tool-free videos that contained moving surgical tools that were generated artificially. Conclusions: Our application successfully separates and eliminates the surgical tool to reveal a view of the background tissue that was otherwise hidden by the tool, producing results that are visually similar to the actual data.

Microengineering 3D Collagen Matrices with Tumor-Mimetic Gradients in Fiber Alignment.

In the tumor microenvironment (TME), collagen fibers facilitate tumor cell migration through the extracellular matrix. Previous studies have focused on studying the responses of cells on uniformly aligned or randomly aligned collagen fibers. However, the in vivo environment also features spatial gradients in alignment, which arise from the local reorganization of the matrix architecture due to cell-induced traction forces. Although there has been extensive research on how cells respond to graded biophysical cues, such as stiffness, porosity, and ligand density, the cellular responses to physiological fiber alignment gradients have been largely unexplored. This is due, in part, to a lack of robust experimental techniques to create controlled alignment gradients in natural materials. In this study, we image tumor biopsy samples and characterize the alignment gradients present in the TME. To replicate physiological gradients, we introduce a first-of-its-kind biofabrication technique that utilizes a microfluidic channel with constricting and expanding geometry to engineer 3D collagen hydrogels with tunable fiber alignment gradients that range from sub-millimeter to millimeter length scales. Our modular approach allows easy access to the microengineered gradient gels, and we demonstrate that HUVECs migrate in response to the fiber architecture. We provide preliminary evidence suggesting that MDA-MB-231 cell aggregates, patterned onto a specific location on the alignment gradient, exhibit preferential migration towards increasing alignment. This finding suggests that alignment gradients could serve as an additional taxis cue in the ECM. Importantly, our study represents the first successful engineering of continuous gradients of fiber alignment in soft, natural materials. We anticipate that our user-friendly platform, which needs no specialized equipment, will offer new experimental capabilities to study the impact of fiber-based contact guidance on directed cell migration.

Disparity refinement framework for learning-based stereo matching methods in cross-domain setting for laparoscopic images.

Purpose: Stereo matching methods that enable depth estimation are crucial for visualization enhancement applications in computer-assisted surgery. Learning-based stereo matching methods have shown great promise in making accurate predictions on laparoscopic images. However, they require a large amount of training data, and their performance may be degraded due to domain shifts. Approach: Maintaining robustness and improving the accuracy of learning-based methods are still open problems. To overcome the limitations of learning-based methods, we propose a disparity refinement framework consisting of a local disparity refinement method and a global disparity refinement method to improve the results of learning-based stereo matching methods in a cross-domain setting. Those learning-based stereo matching methods are pre-trained on a large public dataset of natural images and are tested on two datasets of laparoscopic images. Results: Qualitative and quantitative results suggest that our proposed disparity framework can effectively refine disparity maps when they are noise-corrupted on an unseen dataset, without compromising prediction accuracy when the network can generalize well on an unseen dataset. Conclusions: Our proposed disparity refinement framework could work with learning-based methods to achieve robust and accurate disparity prediction. Yet, as a large laparoscopic dataset for training learning-based methods does not exist and the generalization ability of networks remains to be improved, the incorporation of the proposed disparity refinement framework into existing networks will contribute to improving their overall accuracy and robustness associated with depth estimation.

KIDS SAVE LIVES: Basic Life Support Education for Schoolchildren: A Narrative Review and Scientific Statement From the International Liaison Committee on Resuscitation.

BACKGROUND: Basic life support education for schoolchildren has become a key initiative to increase bystander cardiopulmonary resuscitation rates. Our objective was to review the existing literature on teaching schoolchildren basic life support to identify the best practices to provide basic life support training in schoolchildren. METHODS: After topics and subgroups were defined, a comprehensive literature search was conducted. Systematic reviews and controlled and uncontrolled prospective and retrospective studies containing data on students <20 years of age were included. RESULTS: Schoolchildren are highly motivated to learn basic life support. The CHECK-CALL-COMPRESS algorithm is recommended for all schoolchildren. Regular training in basic life support regardless of age consolidates long-term skills. Young children from 4 years of age are able to assess the first links in the chain of survival. By 10 to 12 years of age, effective chest compression depths and ventilation volumes can be achieved on training manikins. A combination of theoretical and practical training is recommended. Schoolteachers serve as effective basic life support instructors. Schoolchildren also serve as multipliers by passing on basic life support skills to others. The use of age-appropriate social media tools for teaching is a promising approach for schoolchildren of all ages. CONCLUSIONS: Schoolchildren basic life support training has the potential to educate whole generations to respond to cardiac arrest and to increase survival after out-of-hospital cardiac arrest. Comprehensive legislation, curricula, and scientific assessment are crucial to further develop the education of schoolchildren in basic life support.

KIDS SAVE LIVES: Basic Life Support Education for Schoolchildren: A Narrative Review and Scientific Statement From the International Liaison Committee on Resuscitation.

BACKGROUND: Basic life support education for schoolchildren has become a key initiative to increase bystander cardiopulmonary resuscitation rates. Our objective was to review the existing literature on teaching schoolchildren basic life support to identify the best practices to provide basic life support training in schoolchildren. METHODS: After topics and subgroups were defined, a comprehensive literature search was conducted. Systematic reviews and controlled and uncontrolled prospective and retrospective studies containing data on students <20 years of age were included. RESULTS: Schoolchildren are highly motivated to learn basic life support. The CHECK-CALL-COMPRESS algorithm is recommended for all schoolchildren. Regular training in basic life support regardless of age consolidates long-term skills. Young children from 4 years of age are able to assess the first links in the chain of survival. By 10 to 12 years of age, effective chest compression depths and ventilation volumes can be achieved on training manikins. A combination of theoretical and practical training is recommended. Schoolteachers serve as effective basic life support instructors. Schoolchildren also serve as multipliers by passing on basic life support skills to others. The use of age-appropriate social media tools for teaching is a promising approach for schoolchildren of all ages. CONCLUSIONS: Schoolchildren basic life support training has the potential to educate whole generations to respond to cardiac arrest and to increase survival after out-of-hospital cardiac arrest. Comprehensive legislation, curricula, and scientific assessment are crucial to further develop the education of schoolchildren in basic life support.

Deep learning generates synthetic cancer histology for explainability and education.

Artificial intelligence methods including deep neural networks (DNN) can provide rapid molecular classification of tumors from routine histology with accuracy that matches or exceeds human pathologists. Discerning how neural networks make their predictions remains a significant challenge, but explainability tools help provide insights into what models have learned when corresponding histologic features are poorly defined. Here, we present a method for improving explainability of DNN models using synthetic histology generated by a conditional generative adversarial network (cGAN). We show that cGANs generate high-quality synthetic histology images that can be leveraged for explaining DNN models trained to classify molecularly-subtyped tumors, exposing histologic features associated with molecular state. Fine-tuning synthetic histology through class and layer blending illustrates nuanced morphologic differences between tumor subtypes. Finally, we demonstrate the use of synthetic histology for augmenting pathologist-in-training education, showing that these intuitive visualizations can reinforce and improve understanding of histologic manifestations of tumor biology.

Quantifying variation in home spirometry in people with cystic fibrosis during baseline health, and associations with clinical outcomes.

BACKGROUND: Home spirometry is increasingly used to monitor lung function in people with cystic fibrosis (pwCF). Although decreases in lung function in the setting of increased respiratory symptoms are consistent with a pulmonary exacerbation (PEx), the interpretation of home spirometry during asymptomatic periods of baseline health is unclear. The aims of this study were to determine the variation in home spirometry in pwCF during asymptomatic periods of baseline health and to identify associations between this variation and PEx. METHODS: Near-daily home spirometry measurements were obtained from a cohort of pwCF enrolled in a long-term study of the airway microbiome. Associations between the degree of variation in home spirometry and the time to next PEx were evaluated. RESULTS: Thirteen subjects (mean age of 29 years and mean percent predicted forced expiratory volume in one second [ppFEV1] of 60) provided a median of 204 spirometry readings taken during 40 periods of baseline health. The mean week-to-week within-subject level of variation in ppFEV1 was 15.2 ± 6.2%. The degree of variation in ppFEV1 during baseline health was not associated with time to PEx. CONCLUSIONS: Variation in ppFEV1 measured with near-daily home spirometry in pwCF during periods of baseline health exceeded the variation in ppFEV1 expected in clinic spirometry (based on ATS guidelines). The degree of variation in ppFEV1 during baseline health was not associated with time to PEx. These data are relevant for guiding interpretation of home spirometry.

Deep learning architecture for 3D image super-resolution of late gadolinium enhanced cardiac MRI.

Purpose: High-resolution late gadolinium enhanced (LGE) cardiac magnetic resonance imaging (MRI) volumes are difficult to acquire due to the limitations of the maximal breath-hold time achievable by the patient. This results in anisotropic 3D volumes of the heart with high in-plane resolution, but low-through-plane resolution. Thus, we propose a 3D convolutional neural network (CNN) approach to improve the through-plane resolution of the cardiac LGE-MRI volumes. Approach: We present a 3D CNN-based framework with two branches: a super-resolution branch to learn the mapping between low-resolution and high-resolution LGE-MRI volumes, and a gradient branch that learns the mapping between the gradient map of low-resolution LGE-MRI volumes and the gradient map of high-resolution LGE-MRI volumes. The gradient branch provides structural guidance to the CNN-based super-resolution framework. To assess the performance of the proposed CNN-based framework, we train two CNN models with and without gradient guidance, namely, dense deep back-projection network (DBPN) and enhanced deep super-resolution network. We train and evaluate our method on the 2018 atrial segmentation challenge dataset. Additionally, we also evaluate these trained models on the left atrial and scar quantification and segmentation challenge 2022 dataset to assess their generalization ability. Finally, we investigate the effect of the proposed CNN-based super-resolution framework on the 3D segmentation of the left atrium (LA) from these cardiac LGE-MRI image volumes. Results: Experimental results demonstrate that our proposed CNN method with gradient guidance consistently outperforms bicubic interpolation and the CNN models without gradient guidance. Furthermore, the segmentation results, evaluated using Dice score, obtained using the super-resolved images generated by our proposed method are superior to the segmentation results obtained using the images generated by bicubic interpolation (p<0.01) and the CNN models without gradient guidance (p<0.05). Conclusion: The presented CNN-based super-resolution method with gradient guidance improves the through-plane resolution of the LGE-MRI volumes and the structure guidance provided by the gradient branch can be useful to aid the 3D segmentation of cardiac chambers, such as LA, from the 3D LGE-MRI images.

A Deep Learning Framework to Estimate Elastic Modulus from Ultrasound Measured Displacement Fields.

Ultrasound (US) elastography is a technique that enables non-invasive quantification of material properties, such as stiffness, from ultrasound images of deforming tissue. The displacement field is measured from the US images using image matching algorithms, and then a parameter, often the elastic modulus, is inferred or subsequently measured to identify potential tissue pathologies, such as cancerous tissues. Several traditional inverse problem approaches, loosely grouped as either direct or iterative, have been explored to estimate the elastic modulus. Nevertheless, the iterative techniques are typically slow and computationally intensive, while the direct techniques, although more computationally efficient, are very sensitive to measurement noise and require the full displacement field data (i.e., both vector components). In this work, we propose a deep learning approach to solve the inverse problem and recover the spatial distribution of the elastic modulus from one component of the US measured displacement field. The neural network used here is trained using only simulated data obtained via a forward finite element (FE) model with known variations in the modulus field, thus avoiding the reliance on large measurement data sets that may be challenging to acquire. A U-net based neural network is then used to predict the modulus distribution (i.e., solve the inverse problem) using the simulated forward data as input. We quantitatively evaluated our trained model with a simulated test dataset and observed a 0.0018 mean squared error (MSE) and a 1.14% mean absolute percent error (MAPE) between the reconstructed and ground truth elastic modulus. Moreover, we also qualitatively compared the output of our U-net model to experimentally measured displacement data acquired using a US elastography tissue-mimicking calibration phantom.

Developing and Evaluating the Fidelity of Patient Specific Kidney Emulating Phantoms for Image-guided Intervention Applications.

Patient specific organ and tissue mimicking phantoms are used routinely to develop and assess new image-guided intervention tools and techniques in laboratory settings, enabling scientists to maintain acceptable anatomical relevance, while avoiding animal studies when the developed technology is still in its infancy. Gelatin phantoms, specifically, offer a cost-effective and readily available alternative to the traditional manufacturing of anatomical phantoms, and also provide the necessary versatility to mimic various stiffness properties specific to various organs or tissues. In this study, we describe the protocol to develop patient specific anthropomorphic gelatin kidney phantoms and we also assess the faithfulness of the developed phantoms against the patient specific CT images and corresponding virtual anatomical models used to generate the phantoms. We built the gelatin phantoms by first using additive manufacturing to generate a kidney mold based on patient specific CT images, into which the gelatin was poured. We then evaluated the fidelity of the phantoms (i.e., children) against the virtual kidney model generated from the patient specific CT image (i.e., parent) by comparing it to the surface model of the mold and gelatin phantoms (i.e., children) following their CT imaging post-manufacturing using various registration metrics. Our experiments showed a 0.58 ± 0.48 mm surface-to-surface distance between the phantoms and mold models following landmark-based registration, and 0.52 ± 0.40 mm surface-to-surface distance between the phantoms and the mold model following iterative closest point (ICP) registration. These experiments confirm that the described protocol provides a reliable, fast, and cost-effective method for manufacturing faithful patient specific organ emulating gelatin phantoms and can be applied or extended to other image-guided intervention applications.

A Disparity Refinement Framework for Learning-based Stereo Matching Methods in Cross-domain Setting for Laparoscopic Images.

Stereo matching methods that enable depth estimation are crucial for visualization enhancement applications in computer-assisted surgery (CAS). Learning-based stereo matching methods are promising to predict accurate results for applications involving video images. However, they require a large amount of training data, and their performance may be degraded due to domain shifts. Maintaining robustness and improving performance of learning-based methods are still open problems. To overcome the limitations of learning-based methods, we propose a disparity refinement framework consisting of a local disparity refinement method and a global disparity refinement method to improve the results of learning-based stereo matching methods in a cross-domain setting. Those learning-based stereo matching methods are pre-trained on a large public dataset of natural images and are tested on a dataset of laparoscopic images. Results from the SERV-CT dataset showed that our proposed framework can effectively refine disparity maps on an unseen dataset even when they are corrupted by noise, and without compromising correct prediction, provided the network can generalize well on unseen datasets. As such, our proposed disparity refinement framework has the potential to work with learning-based methods to achieve robust and accurate disparity prediction. Yet, as a large laparoscopic dataset for training learning-based methods does not exist and the generalization ability of networks remains to be improved, it will be beneficial to incorporate the proposed disparity refinement framework into existing networks for more accurate and robust depth estimation.