Artificial Intelligence-based Algorithms Research Articles

Intravital microscopy visualizes innate immune crosstalk and function in tissue microenvironment.

Significant advances have been made in the field of intravital microscopy (IVM) on myeloid cells due to the growing number of validated fluorescent probes and reporter mice. IVM provides a visualization platform to directly observe cell behavior and deepen our understanding of cellular dynamics, heterogeneity, plasticity, and cell-cell communication in native tissue environments. This review outlines the current studies on the dynamic interaction and function of innate immune cells with a focus on those that are studied with IVM and covers the advances in data analysis with emerging artificial intelligence-based algorithms. Finally, the prospects of IVM on innate immune cells are discussed.

Performance of artificial intelligence-based algorithms to predict prolonged length of stay after head and neck cancer surgery

BackgroundMedical resource management can be improved by assessing the likelihood of prolonged length of stay (LOS) for head and neck cancer surgery patients. The objective of this study was to develop predictive models that could be used to determine whether a patient's LOS after cancer surgery falls within the normal range of the cohort. MethodsWe conducted a retrospective analysis of a dataset consisting of 300 consecutive patients who underwent head and neck cancer surgery between 2017 and 2022 at a single university medical center. Prolonged LOS was defined as LOS exceeding the 75th percentile of the cohort. Feature importance analysis was performed to evaluate the most important predictors for prolonged LOS. We then constructed 7 machine learning and deep learning algorithms for the prediction modeling of prolonged LOS. ResultsThe algorithms reached accuracy values of 75.40 (radial basis function neural network) to 97.92 (Random Trees) for the training set and 64.90 (multilayer perceptron neural network) to 84.14 (Random Trees) for the testing set. The leading parameters predicting prolonged LOS were operation time, ischemia time, the graft used, the ASA score, the intensive care stay, and the pathological stages. The results revealed that patients who had a higher number of harvested lymph nodes (LN) had a lower probability of recurrence but also a greater LOS. However, patients with prolonged LOS were also at greater risk of recurrence, particularly when fewer (LN) were extracted. Further, LOS was more strongly correlated with the overall number of extracted lymph nodes than with the number of positive lymph nodes or the ratio of positive to overall extracted lymph nodes, indicating that particularly unnecessary lymph node extraction might be associated with prolonged LOS. ConclusionsThe results emphasize the need for a closer follow-up of patients who experience prolonged LOS. Prospective trials are warranted to validate the present results.

The neglected emotional drawbacks of the prioritization of embryos to transfer

In recent years, increasing efforts have been made to develop advanced techniques that could predict the potential of implantation of each single embryo and prioritize the transfer of those at higher chance. The most promising include non-invasive preimplantation genetic testing for aneuploidy and artificial intelligence-based algorithms using time lapse images. The psychological effect of these add-ons is neglected. One could speculate that embarking on another transfer after one or more failures with the prospect of receiving an embryo of lower potential may be distressing for the couple. In addition, the symbolic and mental representation of an embryo with ‘lower capacity to implant’ is currently unknown but could affect couples’ choices and wellbeing. These emotional responses may also undermine adherence to the programme and, ultimately, its real effectiveness. Future trials aimed at evaluating the validity of prioritization procedures must also consider the emotional burden on the couples.

Helicopter radio system for low altitudes and flight speed measuring with pulsed ultra-wideband stochastic sounding signals and artificial intelligence elements

The subject matter of this study is algorithms for measuring the components of an aircraft speed vector and altitude. The goal of this study is to improve algorithms for processing wideband stochastic pulse signals in helicopter low-altitude and flight-speed radio systems by introducing secondary signal processing based on artificial intelligence elements. The tasks to be solved are as follows: to develop an optimal algorithm for determining the speed and altitude of flight for a helicopter radio complex; to supplement the signal processing algorithm with an artificial intelligence-based processor to determine the "safety" of the current trajectory; provide the pilot with relevant information about possible options for further actions based on an analysis of the current position of the helicopter and flight parameters; and to analyse the efficiency of the proposed complex when using various artificial intelligence-based algorithms. The methods used are as follows: methods of mathematical statistics and optimal solutions for solving problems of statistical synthesis of active radio complex structure; methods of machine learning; and methods of computer simulation. The following results were obtained. The algorithms for signal processing in a helicopter radio complex are obtained by the method of maximum likelihood, and the use of three radio channels to calculate the full vector of speed and altitude is argued. The structure of a secondary information processing system, using algorithms based on artificial intelligence is proposed. The effectiveness of determining the safety of the current landing trajectory using various algorithms based on artificial intelligence (LinearSVC, GaussianNB, DecisionTreeClassifier, RandomForestClassifier, KNeighborsClassifier, MLPClassifier and RidgeClassifier) was analysed. Conclusions. The simulation results show that in the presence of accurate (noise-free) information on the current location of the helicopter, its axial velocities, and a map of the terrain with defined areas dangerous for landing, the DecisionTreeClassifier and RandomForestClassifier algorithms can provide a high probability of correctly determining the safety of the current landing trajectory. At the same time, in the presence of instability in the measurements of helicopter movement parameters, only the RandomForestClassifier algorithm maintains high accuracy.

Automatic analysis of bronchus-artery dimensions to diagnose and monitor airways disease in cystic fibrosis.

Cystic fibrosis (CF) lung disease is characterised by progressive airway wall thickening and widening. We aimed to validate an artificial intelligence-based algorithm to assess dimensions of all visible bronchus-artery (BA) pairs on chest CT scans from patients with CF. The algorithm fully automatically segments the bronchial tree; identifies bronchial generations; matches bronchi with the adjacent arteries; measures for each BA-pair bronchial outer diameter (Bout), bronchial lumen diameter (Bin), bronchial wall thickness (Bwt) and adjacent artery diameter (A); and computes Bout/A, Bin/A and Bwt/A for each BA pair from the segmental bronchi to the last visible generation. Three datasets were used to validate the automatic BA analysis. First BA analysis was executed on 23 manually annotated CT scans (11 CF, 12 control subjects) to compare automatic with manual BA-analysis outcomes. Furthermore, the BA analysis was executed on two longitudinal datasets (Copenhagen 111 CTs, ataluren 347 CTs) to assess longitudinal BA changes and compare them with manual scoring results. The automatic and manual BA analysis showed no significant differences in quantifying bronchi. For the longitudinal datasets the automatic BA analysis detected 247 and 347 BA pairs/CT in the Copenhagen and ataluren dataset, respectively. A significant increase of 0.02 of Bout/A and Bin/A was detected for Copenhagen dataset over an interval of 2 years, and 0.03 of Bout/A and 0.02 of Bin/A for ataluren dataset over an interval of 48 weeks (all p<0.001). The progression of 0.01 of Bwt/A was detected only in the ataluren dataset (p<0.001). BA-analysis outcomes showed weak to strong correlations (correlation coefficient from 0.29 to 0.84) with manual scoring results for airway disease. The BA analysis can fully automatically analyse a large number of BA pairs on chest CTs to detect and monitor progression of bronchial wall thickening and bronchial widening in patients with CF.

Automated diagnosis of pancreatic mucinous and serous cystic neoplasms with modality-fusion deep neural network using multi-modality MRIs.

Pancreatic cystic neoplasms are increasingly diagnosed with the development of medical imaging technology and people's self-care awareness. However, two of their sub-types, serous cystic neoplasms (SCN) and mucinous cystic neoplasms (MCN), are often misclassified from each other. Because SCN is primarily benign and MCN has a high rate of malignant transformation. Distinguishing SCN and MCN is challenging and essential. MRIs have many different modalities, complete with SCN and MCN diagnosis information. With the help of an artificial intelligence-based algorithm, we aimed to propose a multi-modal hybrid deep learning network that can efficiently diagnose SCN and MCN using multi-modality MRIs. A cross-modal feature fusion structure was innovatively designed, combining features of seven modalities to realize the classification of SCN and MCN. 69 Patients with multi-modalities of MRIs were included, and experiments showed performances of every modality. The proposed method with the optimized settings outperformed all other techniques and human radiologists with high accuracy of 75.07% and an AUC of 82.77%. Besides, the proposed disentanglement method outperformed other fusion methods, and delayed contrast-enhanced T1-weighted MRIs proved most valuable in diagnosing SCN and MCN. Through the use of a contemporary artificial intelligence algorithm, physicians can attain high performance in the complex challenge of diagnosing SCN and MCN, surpassing human radiologists to a significant degree.

Using artificial intelligence-based algorithms to identify critical fouling factors and predict fouling behavior in anaerobic membrane bioreactors

Membrane fouling is one of the major obstacles that hinder the widespread applications of anerobic membrane bioreactors (AnMBRs) for wastewater treatment. Due to the intrinsic complexity of membrane fouling, identifying critical fouling factors and predicting fouling behavior are of great significance in membrane fouling control. Herein, artificial intelligence (AI) algorithms and their modeling framework were developed to predict membrane fouling behaviors in AnMBRs. Operating parameters, biomass properties, and membrane characteristics were considered as input variables for membrane fouling prediction. The results of hyper-parameter optimization showed that the optimal architecture of artificial neural network (ANN) model for membrane fouling prediction was “14-9-6-1” while the best hyper-parameters of random forest (RF) model were n_trees = 1200 and n_features = 14. After hyper-parameter adjusting, RF had more robustness of predictive capabilities (R2=0.906, mean squared error (MSE) = 0.061) for membrane fouling than the ANN model (R2=0.800, MSE = 0.118). More importantly, the feature importance and Shapley additive explanations analysis indicated SMPp/SMPc (0.281) > EPSp/EPSc (0.110) > organic loading rate (0.106), which were the most critical factors affecting membrane fouling. Partial dependence plot analysis further verified the marginal and interaction effects of various input features on membrane fouling. The revealed partial dependence relationships of critical variables can provide theoretical reference for optimization of practical operation. Overall, this study established a novel AI-based approach for predicting membrane fouling and comprehensively understanding the complicated effects of various influencing factors on membrane fouling while overcoming the “black-box” nature of conventional models.

The potential role of 3D-printed phantoms in quality control of artificial intelligence-based algorithms in medical imaging

Abstract To ensure accurate and consistent imaging of patients, medical imaging systems are controlled and tested using phantoms. Despite the availability of commercial standard phantoms for decades, 3D printing technology has gained special attention as a tool for producing accurate and costeffective tissue-mimicking phantoms. As artificial intelligence (AI) becomes increasingly prevalent in medical imaging, dedicated phantoms are needed for testing their reliability, robustness, and quality before they are implemented in clinical settings. In this context, 3D-printed imaging phantoms, which have specific requirements relevant to AI models, can play a crucial role. Due to its unique ability to create phantoms of almost any complexity, 3D printing technology seems a suitable approach for the quality control of AI models in medical imaging. The following reviews some of the works that used 3D-printed technology to create custom-built phantoms for use in Computed Tomography (CT), nuclear imaging, Magnetic Resonance Imaging (MRI), and ultrasound imaging. The focus of this short review is on the accuracy of 3D-printed technology in creating imaging phantoms. In the end, the potential of the 3D-printed phantoms in testing and quality control of AI-based algorithms in radiology is discussed.

New Approach for Generating Synthetic Medical Data to Predict Type 2 Diabetes.

The lack of medical databases is currently the main barrier to the development of artificial intelligence-based algorithms in medicine. This issue can be partially resolved by developing a reliable high-quality synthetic database. In this study, an easy and reliable method for developing a synthetic medical database based only on statistical data is proposed. This method changes the primary database developed based on statistical data using a special shuffle algorithm to achieve a satisfactory result and evaluates the resulting dataset using a neural network. Using the proposed method, a database was developed to predict the risk of developing type 2 diabetes 5 years in advance. This dataset consisted of data from 172,290 patients. The prediction accuracy reached 94.45% during neural network training of the dataset.

Enhancing Pest Detection: Assessing Tuta absoluta (Lepidoptera: Gelechiidae) Damage Intensity in Field Images through Advanced Machine Learning

Tomato (Solanum lycopersicum (Solanaceae)), susceptible to significant yield loss due to pests, is particularly impacted by Tuta absoluta (Meyrick) (Lepidoptera: Gelechiidae), a pest that directly and profoundly influences tomato yields. Consequently, early detection of T. absoluta damage intensity on leaves using machine learning or artificial intelligence-based algorithms is crucial for effective pest control. In this ground-breaking study, the galleries generated by T. absoluta were examined via field images using the Decision Trees (DTs) algorithm, a machine learning method. The unique advantage of DTs over other algorithms is their inherent capacity to identify complex and vague shapes without the necessity of feature extraction, providing a more streamlined and effective approach. The DTs algorithm was meticulously trained using pixel values from the leaf images, leading to the classification of pixels within regions with and without galleries on the leaves. Accordingly, the gallery intensity was determined to be 9.09% and 35.77% in the test pictures. The performance of the DTs algorithm, as evidenced by a high coefficient of determination (R²) and an accuracy rate of 0.98 and 0.99 respectively, testifies to its robust predictive and classification abilities. This pioneering study has far-reaching implications for the future of precision agriculture, potentially informing the development of advanced algorithms that can be integrated into autonomous vehicles. The integration of DTs in such applications, thanks to their distinctive ability to handle complex and indistinct shapes without the need for feature extraction, sets the stage for a new era of efficient and effective pest control strategies.

Void fraction prediction using prompt gamma neutron activation analysis and artificial intelligence

This paper presents the simulation of a water-gas biphasic system in annular and stratified flow regimes using a neutron beam from a241Am–Be radiation source. The objective is to evaluate the potential of prompt gamma neutron activation analysis (PGNAA) and artificial intelligence-based algorithms in flow measurements. A measurement geometry consisted of a neutron flat source with parallel emission and a spherical detector was developed using the MCNP6 code. Prompt gamma rays spectra were used as input data of a feed-forward multilayer perceptron (MLP) artificial neural network aiming to predict void fraction and to identify the investigated flow regimes. The results showed that the MLP was able to predict void fraction, for both flow regimes, with mean absolute percentage error within 1.58%. In addition, the same MLP achieved 100% of success in flow regime identification. Therefore, it is possible to combine PGNAA with MLP in order to investigate flow characteristics.

Current Technologies for Managing Type 1 Diabetes Mellitus and Their Impact on Quality of Life-A Narrative Review.

Type 1 diabetes mellitus is a chronic autoimmune disease that affects millions of people and generates high healthcare costs due to frequent complications when inappropriately managed. Our paper aimed to review the latest technologies used in T1DM management for better glycemic control and their impact on daily life for people with diabetes. Continuous glucose monitoring systems provide a better understanding of daily glycemic variations for children and adults and can be easily used. These systems diminish diabetes distress and improve diabetes control by decreasing hypoglycemia. Continuous subcutaneous insulin infusions have proven their benefits in selected patients. There is a tendency to use more complex systems, such as hybrid closed-loop systems that can modulate insulin infusion based on glycemic readings and artificial intelligence-based algorithms. It can help people manage the burdens associated with T1DM management, such as fear of hypoglycemia, exercising, and long-term complications. The future is promising and aims to develop more complex ways of automated control of glycemic levels to diminish the distress of individuals living with diabetes.

Optical Coherence Tomography and Optical Coherence Tomography Angiography: Essential Tools for Detecting Glaucoma and Disease Progression.

Early diagnosis and detection of disease progression are critical to successful therapeutic intervention in glaucoma, the leading cause of irreversible blindness worldwide. Optical coherence tomography (OCT) is a non-invasive imaging technique that allows objective quantification in vivo of key glaucomatous structural changes in the retina and the optic nerve head (ONH). Advances in OCT technology have increased the scan speed and enhanced image quality, contributing to early glaucoma diagnosis and monitoring, as well as the visualization of critically important structures deep within the ONH, such as the lamina cribrosa. OCT angiography (OCTA) is a dye-free technique for noninvasively assessing ocular microvasculature, including capillaries within each plexus serving the macula, peripapillary retina and ONH regions, as well as the deeper vessels of the choroid. This layer-specific assessment of the microvasculature has provided evidence that retinal and choroidal vascular impairments can occur during early stages of glaucoma, suggesting that OCTA-derived measurements could be used as biomarkers for enhancing detection of glaucoma and its progression, as well as to reveal novel insights about pathophysiology. Moreover, these innovations have demonstrated that damage to the macula, a critical region for the vision-related quality of life, can be observed in the early stages of glaucomatous eyes, leading to a paradigm shift in glaucoma monitoring. Other advances in software and hardware, such as artificial intelligence-based algorithms, adaptive optics, and visible-light OCT, may further benefit clinical management of glaucoma in the future. This article reviews the utility of OCT and OCTA for glaucoma diagnosis and disease progression detection, emphasizes the importance of detecting macula damage in glaucoma, and highlights the future perspective of OCT and OCTA. We conclude that the OCT and OCTA are essential glaucoma detection and monitoring tools, leading to clinical and economic benefits for patients and society.

IMAGE RETRIEVAL FOR 3D MODELLING OF ARCHITECTURE USING AI AND PHOTOGRAMMETRY

Abstract. This research is intended to provide an initial solution to the problem of finding images for processing by photogrammetry in special cases where these do not exist. An overview of existing artificial intelligence-based algorithms that enable the extension of source image dataset is reported. In particular, this research focused on the use of prompt-to-image systems for obtaining images to be used in reconstruction and then in the next step of 3D modelling. Thus, the combined use of these three techniques, AI, photogrammetry, and modelling allowed the creation of a model of a building that never existed except in the collective imagination, which is the tower of Babel. In particular, the case study chosen is the illustration in Kircher book present in the library of the Brixen seminary that is closed to the public and for which it was necessary to create a tool to enhance the value and knowledge of this heritage for external users. Therefore, the creation of an augmented reality app enabled the visualization of the model created by offering possibilities for immersive experiences and dissemination of the research to a wide audience.

Evaluation of Changes in Facial Attractiveness and Estimated Facial Age After Blepharoplasty with an Artificial Intelligence Algorithm

Objective: The aim of this study is to evaluate the effects of blepharoplasty operation on facial attractiveness and estimated facial age with an artificial intelligence-based algorithm over pre- and post-treatment facial photographs. In addition, it is aimed to make a comparison by reviewing the observable changes according to gender and operation type (upper, lower, combined).&#x0D; Methods: Preoperative and postoperative photos of patients who underwent open access and copyright-free blepharoplasty operation on social media platforms (instagram and youtube) were collected. The photos were evaluated by an artificial intelligence algorithm trained to estimate facial age and evaluate facial attractiveness.&#x0D; Results: A total of 541 patients, of which 454 (83.92%) were female and 87 (16.08%) were male. When all patients were evaluated without subgrouping, there was a -1.91±3.35 years younger face age and 0.43±0.64 point increase in facial attractiveness (p&lt;0.005).&#x0D; Conclusion: In this study, the effects of blepharoplasty on facial attractiveness and apparent age are presented with quantitative data. In addition, it has been concluded that artificial intelligence can be used in scoring the apparent age and facial attractiveness after blepharoplasty.

The formal rationality of artificial intelligence-based algorithms and the problem of bias

This paper presents a new perspective on the problem of bias in artificial intelligence (AI)-driven decision-making by examining the fundamental difference between AI and human rationality in making sense of data. Current research has focused primarily on software engineers’ bounded rationality and bias in the data fed to algorithms but has neglected the crucial role of algorithmic rationality in producing bias. Using a Weberian distinction between formal and substantive rationality, we inquire why AI-based algorithms lack the ability to display common sense in data interpretation, leading to flawed decisions. We first conduct a rigorous text analysis to uncover and exemplify contextual nuances within the sampled data. We then combine unsupervised and supervised learning, revealing that algorithmic decision-making characterizes and judges data categories mechanically as it operates through the formal rationality of mathematical optimization procedures. Next, using an AI tool, we demonstrate how formal rationality embedded in AI-based algorithms limits its capacity to perform adequately in complex contexts, thus leading to bias and poor decisions. Finally, we delineate the boundary conditions and limitations of leveraging formal rationality to automatize algorithmic decision-making. Our study provides a deeper understanding of the rationality-based causes of AI’s role in bias and poor decisions, even when data is generated in a largely bias-free context.

Local conduction velocities determined by non-invasive electrocardiographic imaging predict arrhythmia-free survival after pulmonary vein isolation

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – EU funding. Main funding source(s): This project has received funding from the European Union's Horizon Research and Innovation Programme under the Marie-Sklodowska Curie Grant Agreement No. 860974 Background Recurrence rates after atrial fibrillation (AF) ablation are still unsatisfactory. As catheter ablation is primarily targeting pulmonary vein (PV) ectopic activity, it is not surprising that extra-PV arrythmogenic substrate is a key determinant of arrhythmia recurrence. Against this background, several studies have proposed assessment of extra-PV substrate in terms of atrial fibrosis or locally reduced conduction velocities (CV) to guide treatment. However, to date no non-invasive method directly assessing electrical arrhythmogenic substrate has been established in clinical practice, and treatment decisions are commonly based on crude surrogates like AF type or left atrial size. Here we establish and validate a novel non-invasive method based on electrocardiographic imaging (ECGi) to determine atrial arrhythmogenic substrate in terms of reduced local CVs and its predictive value regarding arrhythmia recurrence after PVI. Methods and results 52 consecutive patients scheduled for AF ablation (PVI-only) and 19 healthy controls were prospectively included and received ECGi to assess left and right atrial arrhythmogenic substrate. This ECGi-based method uses 64 electrodes placed on the torso. Subsequently, a 3D model of the torso is acquired as an anatomical reference using a 3D reconstruction camera. A personalised 3D atrial geometry is then derived from a database of human atria using an artificial intelligence-based algorithm. Finally, unipolar surface electrograms are projected onto the cardiac geometry and local CVs are estimated. Mean ECGi-determined atrial CVs were significantly lower in AF patients than in healthy controls, both in a global analysis (1.45±0.15 m/s vs. 1.64±0.15 m/s; p&amp;lt;0.0001) and a regional analysis of 19 predefined left and right atrial segments (Fig. 1). Considering only the segments with the lowest average CV in each patient, differences in CVs were more pronounced (0.80±0.22 vs. 1.08±0.26 m/s; p&amp;lt;0.0001). Multivariate logistic regression analyses combined with c-statistics including other previously proposed predictors found this mean CV of the "slowest" segment to be the strongest predictor of arrhythmia recurrence. A ROC analysis revealed that a cut-off for this variable of 0.72 m/s best discriminates PVI responders from non-responders: patients with a mean CV &amp;gt;0.72 m/s in all atrial regions showed a 6-months arrhythmia-free survival of 90.9%, whereas patients with one or more atrial regions with a mean CV &amp;lt;0.72 m/s had a poor outcome with an arrhythmia-free survival rate of only 57.9% (Fig. 2). Conclusion This was the first study to investigate local atrial CV non-invasively and to validate their predictive value regarding outcome after PVI. The absence of ECGi-determined areas of slow conduction well discriminated PVI responders from non-responders. Such non-invasive assessment of electrical arrhythmogenic substrate may guide treatment strategies and be an important step towards personalised AF therapy.

Cardiogenetics and Artificial Intelligence: the Mutscore Algorithm

Abstract Funding Acknowledgements Type of funding sources: None. Backgroud The analysis of whole genome, exomes or deputed genes in the context of multigene panels, while being extensively available in routine practice, has raised new challenges in result interpretation. Novel sequence variants have become an increasingly frequent finding which, differing from the sequence of the referenced general population, remain of uncertain significance. Purpose We report our experience on a cohort of consecutive patients (pts) referred for suspicion of cardiac hereditary disease to the cardiogenetic outpatient clinic of our center. After the identification of likely pathogenic/pathogenic variants (LP/PV) and of variants of uncertain significance (VUS), we applied the new computational predictive algorithm Mutscore, based on a machine-learning approach, to test its performance and ability in reclassifying VUS. Methods We retrospectively reviewed DNA sequencing results of pts tested for suspected cardiac hereditary disease and addressed at our center from July 1st 2007 through December 31st 2021. For each LP/PV and for VUS we applied the Mutscore algorithm using a random forest approach. The Mutscore, which integrates already existing predictive algorithms to data concerning variant topographic localization, has been already validated. Of note, VUS Mutscore distribution has been analyzed according to two cut-off values (i.e. 0.14 and 0.73) identified in our previous study. In our precedent validation dataset, a Mutscore of 0.14 represented the threshold below which the 95% of variants were likely benign/benign and 0.73 the threshold above which the 95% of variants were likely pathogenic/pathogenic. VUS were then reclassified as likely benign variants when their Mutscore was &amp;lt;0.14 and as likely pathogenic when their Mutscore was &amp;gt;0.73. Results Among the 488 tested pts, a missense LP/PV was found in 126 (25.8%) pts, while 57 missense VUS were identified in 41 (8%) pts. Panel A represents Mutscore distribution among the 115 pts with missense LP/PV and a computable Mutscore. We observed an excellent predictive performance of the Mutscore in our dataset, as attested by an AUC (Area Under the Curve) of 0.895 (Panel B) and by the statistically significant positive correlation (r = 0.65, p-value = 0.000) between the Mutscore and the variant interpretation reported in ClinVar. Among the 55 pts with missense VUS and a computable Mutscore, Panel C represents Mutscore distribution according to two cut-off values (i.e. 0.14 and 0.73) identified in our previous study. The Mutscore supported the reclassification for the 47.2% of VUS into likely pathogenic/benign variants. Conclusions We showed the excellent predictive performance of the Mutscore algorithm and its accuracy in reclassifying VUS. Although further validation studies are needed, the Mutscore appears as a valuable algorithm contributing to VUS disambiguation. Artificial intelligence-based algorithms are promising tools towards precision medicine.

Zio XT for ambulatory heart rhythm monitoring: a cost-benefit analysis

Abstract Funding Acknowledgements Type of funding sources: None. Introduction Zio XT is a recent technology that has enabled automated heart rhythm analysis using an artificial intelligence-based algorithm. It can be fitted by patients and record the heart rhythm for 14 days, offering potential convenience, accuracy and efficiency compared to Holter monitors. However, data on its cost-effectiveness and impact on healthcare resources is lacking. Purpose We performed a cost-benefit analysis of Zio XT for electrocardiogram (ECG) monitoring, using Holter monitors as the standard. Methods 200 patients that had Holters and 204 that had Zio XT at our centre were evaluated. Primary outcomes included impact of Zio on the number of outpatient appointments (OPAs) per year and impact on the cardiac physiologist (CP) workforce, depicted by whole time equivalent (WTE), calculated as the number of hours per year spent on monitor fitting or rhythm analysis ÷ total working hours per year. Secondary outcome was the impact on implantable loop recorders (ILR) insertions per year and the resultant effect on cost and time. Data from the 200 Holter and 204 Zio cases was extrapolated to model outcomes for 1 year, assuming 2000 patients undergo ECG monitoring per year (based on annual figures for our centre). Results 96% of Holter patients had OPAs for monitor fitting and 87% for follow-up. Corresponding figures for Zio patients were 0.5% and 70.1% respectively. Extrapolating this to our annual cohort estimated a net reduction of 1910 OPAs for monitor fitting and 338 for follow-up, saving a total of 2248 OPAs/year with Zio use. 65%, 12% and 23% of Holter patients underwent 24 hour (hr)-, 48hr- and 7 day-tapes respectively. Using a projected time of 20 minutes to fit a device and 30, 45 and 60 minutes to analyse a 24hr-, 48hr- and 7 day-tape respectively, we derived a total WTE of 0.41 for monitor fitting and WTE of 0.79 for rhythm analysis (Table 1), i.e. 41% of a band 4 CP’s time is spent on device fitting and 79% of a band 6 CP’s time is spent on analysis. For Zio, the WTE for device fitting was 0.002 and that for analysis was 0, suggesting that a WTE of 0.41 can be saved on monitor fitting and 0.788 on rhythm analysis. Using our data, we extrapolated that 10 patients in the Zio group and 60 in the Holter group would require ILRs in a year. Assuming an ILR insertion takes 15 minutes and removal takes 30 minutes, we estimated that Zio use could save 38 hrs of a CP’s time/year. Further, a saving of £90,000 was projected from the 50 ILR insertions that could be avoided (using a cost of £1800 per ILR insertion). Conclusion Our findings indicate that Zio XT is associated with a significant reduction in OPAs, ILR insertions and time spent on monitor fitting and rhythm analysis, and thus, is potentially cost- and time-saving. This could reduce outpatient waiting lists and liberate valuable physiologist time that can be redirected towards other services. Further reduction in OPAs may be observed as we emerge from the Zio learning path.

A Deep Learning Approach for Meibomian Gland Appearance Evaluation

To develop and evaluate a deep learning algorithm for Meibomian gland characteristics calculation. Evaluation of diagnostic technology. A total of 1616 meibography images of both the upper (697) and lower (919) eyelids from a total of 282 individuals. Images were collected using the LipiView II device. All the provided data were split into 3 sets: the training, validation, and test sets. Data partitions used proportions of 70/10/20% and included data from 2 optometry settings. Each set was separately partitioned with these proportions, resulting in a balanced distribution of data from both settings. The images were divided based on patient identifiers, such that all images collected for one participant could end up only in one set. The labeled images were used to train a deep learning model, which was subsequently used for Meibomian gland segmentation. The model was then applied to calculate individual Meibomian gland metrics. Interreader agreement and agreement between manual and automated methods for Meibomian gland segmentation were also carried out to assess the accuracy of the automated approach. Meibomian gland metrics, including length ratio, area, tortuosity, intensity, and width, were measured. Additionally, the performance of the automated algorithms was evaluated using the aggregated Jaccard index. The proposed semantic segmentation-based approach achieved average aggregated Jaccard index of mean 0.4718 (95% confidence interval [CI], 0.4680-0.4771) for the 'gland' class and a mean of 0.8470 (95% CI, 0.8432-0.8508) for the 'eyelid' class. The result for object detection-based approach was a mean of 0.4476 (95% CI, 0.4426-0.4533). Both artificial intelligence-based algorithms underestimated area, length ratio, tortuosity, widthmean, widthmedian, width10th, and width90th. Meibomian gland intensity was overestimated by both algorithms compared with the manual approach. The object detection-based algorithm seems to be as reliable as the manual approach only for Meibomian gland width10th calculation. The proposed approach can successfully segment Meibomian glands; however, to overcome problems with gland overlap and lack of image sharpness, the proposed method requires further development. The study presents another approach to utilizing automated, artificial intelligence-based methods in Meibomian gland health assessment that may assist clinicians in the diagnosis, treatment, and management of Meibomian gland dysfunction. The authors have no proprietary or commercial interest in any materials discussed in this article.