Similar Accuracy Research Articles

Line-field confocal optical coherence tomography, a novel non-invasive tool for the diagnosis of onychomycosis.

Onychomycosis is common and important to distinguish from other nail diseases. Rapid and accurate diagnosis is necessary for optimal patient treatment and outcome. Non-invasive diagnostic tools have increasing potential for nail diseases including onychomycosis. This study evaluated line-field confocal optical coherence tomography (LC-OCT) as a rapid non-invasive tool for diagnosing onychomycosis as compared to confocal laser scanning microscopy (CLSM), optical coherence tomography (OCT), and conventional methods. In this prospective study 86 patients with clinically suspected onychomycosis and 14 controls were examined using LC-OCT, OCT, and CLSM. KOH-preparation, fungal culture, PCR, and histopathology were used as comparative conventional methods. LC-OCT had the highest sensitivity and negative predictive value of all methods used, closely followed by PCR and OCT. Specificity and positive predictive value of LC-OCT were as high as with CLSM, while OCT scored much lower. The gold standard technique, fungal culture, showed the lowest sensitivity and negative predictive value. Only PCR and culture allowed species differentiation. LC-OCT enables quick and non-invasive detection of onychomycosis, with advantages over CLSM and OCT, and similar diagnostic accuracy to PCR but lacking species differentiation. For accurate nail examination, LC-OCT requires well-trained and experienced operators.

Predicting Cardiovascular Disease Using Simple Machine Learning Techniques

Cardiovascular diseases (CVD) become a major health concern, which needs improved prediction models for intervention. In this study, this research explores the possibilities of using simple machine learning methods, including Logistic Regression, Decision Tree, and Random Forest, for predicting cardiovascular diseases effectively. Using a large dataset (308854 instances) containing demographic, binary, and numerical information, the research applied the above machine learning algorithms to develop predictive models. This study involved data preprocessing, including feature selection and handling duplicate values. This study then divides the dataset into two subsets: training set and testing set by 7:3 and 8:2. The Logistic Regression algorithm demonstrated good predictive performance, with an accuracy rate of 92%. Decision Tree exhibited a similar accuracy of 92%, while Random Forest underperformed both slightly with an accuracy of 91%.

Effects of SNP marker density and training population size on prediction accuracy in alfalfa (Medicago sativa L.) genomic selection.

Effects of individual single-nucleotide polymorphism (SNP) markers and the size of "training" and "test" populations affect prediction accuracy in genomic selection (GS). This study evaluated 11 subsets of 4932 SNPs using six genetic additive methods to understand marker density in GS prediction in alfalfa (Medicago sativa L.). In the GS methods, the effect of "training" to "test" population size was also evaluated. Fourteen alfalfa populations sampled from long-term grazing sites were genotyped using genotyping by sequencing for the identification of SNPs. These populations were also phenotyped for six agromorphological and three nutritive traits from 2018 to 2020. The accuracy of GS prediction improved across six GS methods when the ratio of "training" to "test" population size increased. However, the prediction accuracy of the six GS methods reduced to a range of -0.27 to 0.11 when random, uninformative SNPs were used. In this study, five Bayesian methods and ridge-regression best linear unbiased prediction (rrBLUP) method had similar GS accuracies for "training" sets, but rrBLUP tended to outperform Bayesian methods in independent "test" sets when SNP subsets with high mean-squared-estimated-marker effect were used. These findings can enhance the application of GS in alfalfa genetic improvement.

Determination of growth and developmental stages in hand-wrist radiographs : Can fractal analysis in combination with artificial intelligence be used?

The goal of this work was to assess the classification of maturation stage using artificial intelligence (AI) classifiers. Hand-wrist radiographs (HWRs) from 1067 individuals aged between 7and 18years were included. Fifteen regions of interest were selected for fractal dimension (FD) analysis. Five predictive models with different inputs were created (model1: only FD; model2: FD and Chapman sesamoid stage; model3: FD, age, and sex; model4: FD, Chapman sesamoid stage, age, and sex; model5: Chapman sesamoid stage, age, and sex). The target diagnoses were accelerating growth velocity, very high growth velocity, and decreasing growth velocity. Four AI algorithms were applied: multilayer perceptron (MLP), support vector machine (SVM), gradient boosting machine (GBM) and C5.0 decision tree classifier. All AI algorithms except for C5.0 yielded similar overall predictive accuracies for the five models. In order from lowest to highest, the predictive accuracies of the models were as follows: model1 < model3 < model2 < model5 < model4. The highest overall F1 score, which was used instead of accuracy especially for models with unbalanced data, was obtained for models1,2, and3 based on SVM, for model4 based on MLP, and for model5 based on C5.0. Adding Chapman sesamoid stage, chronologic age, and sex as additional inputs to the FD values significantly increased the F1 score. Applying FD analysis to HWRs is not sufficient to predict maturation stage in growing patients but can be considered agrowth rate prediction method if combined with the Chapman sesamoid stage, age, and sex.

Deep learning models for groundwater level prediction based on delay penalty

Abstract In irrigation agriculture, predicting groundwater level (GWL) using deep learning models can help decision-makers coordinate surface water and groundwater usage, thus aiding in the sustainable development and utilization of groundwater. However, when making a long sequence prediction, prediction sequences often have severe delays affecting the availability of prediction results. In this paper, a new loss function is proposed to minimize the lag and oversmoothing on the prediction of GWLs. GWL, meteorology, and pumping data are collected via an irrigation Internet of Things system in Hutubi County, Xinjiang. Through Pearson's correlation analysis, historical potential evapotranspiration (ET0), groundwater extraction, and GWL were chosen to predict GWLs. Datasets were constructed through the proposed spatiotemporal data fusion method; then, the best model from the six deep learning models was selected by comparing the prediction capability of the datasets. Finally, the mean-squared error (MSE) loss function is replaced by the proposed loss function. Compared to the mean absolute error, MSE, and predicted sequence graphs, the new loss function significantly depresses the time delay with similar prediction accuracy.

Assessment of online information on robotic cardiac and thoracic surgery.

Online health resources are important for patients seeking perioperative information on robotic cardiac and thoracic surgery. The value of the resources depends on their readability, accuracy, content, quality, and suitability for patient use. We systematically assess current online health information on robotic cardiac and thoracic surgery. Systematic online searches were performed to identify websites discussing robotic cardiac and thoracic surgery. For each website, readability was measured by nine standardized tests, and accuracy and content were assessed by an independent panel of two robotic cardiothoracic surgeons. Quality and suitability of websites were evaluated using the DISCERN and Suitability Assessment of Materials tools, respectively. A total of 220 websites (120 cardiac, and 100 thoracic) were evaluated. Both robotic cardiac and thoracic surgery websites were very difficult to read with mean readability scores of 13.8 and 14.0 (p = 0.97), respectively, requiring at least 13years of education to be comprehended. Both robotic cardiac and thoracic surgery websites had similar accuracy, amount of content, quality, and suitability (p > 0.05). On multivariable regression, academic websites [Exp (B)], 2.25; 95% confidence interval [CI], 1.60-3.16; P < 0.001), and websites with higher amount of content [Exp (B)],1.73; 95% CI, 1.24-2.41; P < 0.001) were associated with higher accuracy. There was no association between readability of websites and accuracy [Exp (B)], 1.04; 95% CI, 0.90-1.21; P = 0.57). Online information on robotic cardiac and thoracic surgery websites overestimate patients' understanding and require at least 13years of education to be comprehended. As website accuracy is not associated with ease of reading, the readability of online resources can be improved without compromising accuracy.

Multimodal imaging measures in the prediction of clinical response to deep brain stimulation for refractory depression: A machine learning approach

Objectives This study compared machine learning models using unimodal imaging measures and combined multi-modal imaging measures for deep brain stimulation (DBS) outcome prediction in treatment resistant depression (TRD). Methods Regional brain glucose metabolism (CMRGlu), cerebral blood flow (CBF), and grey matter volume (GMV) were measured at baseline using 18F-fluorodeoxy glucose (18F-FDG) positron emission tomography (PET), arterial spin labelling (ASL) magnetic resonance imaging (MRI), and T1-weighted MRI, respectively, in 19 patients with TRD receiving subcallosal cingulate (SCC)-DBS. Responders (n = 9) were defined by a 50% reduction in HAMD-17 at 6 months from the baseline. Using an atlas-based approach, values of each measure were determined for pre-selected brain regions. OneR feature selection algorithm and the naïve Bayes model was used for classification. Leave-out-one cross validation was used for classifier evaluation. Results The performance accuracy of the CMRGlu classification model (84%) was greater than CBF (74%) or GMV (74%) models. The classification model using the three image modalities together led to a similar accuracy (84%0 compared to the CMRGlu classification model. Conclusions CMRGlu imaging measures may be useful for the development of multivariate prediction models for SCC-DBS studies for TRD. The future of multivariate methods for multimodal imaging may rest on the selection of complementing features and the developing better models.Clinical Trial Registration: ClinicalTrials.gov (#NCT 01983904)

A comparative assessment of time-consuming and laborious diatom analysis：Brief experimentation with suggestion of automatic identification

Diatom testing is considered a useful method for providing supportive evidence for the diagnosis of drowning in forensic pathology. However, various factors remain controversial for recognizing diatoms, such as being time-consuming and laborious and influencing the consistency of the results. Given the absence of precise and well-defined studies on this subject, this study aimed to determine the relationship between the ability to identify diatoms and researchers with different technical backgrounds. A total of 55 samples from 18 cases, including water, lungs, liver, and kidneys, were treated using the microwave digestion-vacuum filtration-automated scanning electron microscopy (MD-VF-Auto SEM), which was used to compare diatom analyses among three groups of well-trained forensic pathologists (FPs), trained junior employees (JEs), and new trainees (TEs). In addition to achieving similar accuracy of positive findings from drowning cases, counting efficiency was evaluated based on taxonomy records and counting time after viewing more than 5500 diatom images. In contrast to the higher counting efficiency of the JE group than that of the TE group, we observed a statistically significant difference (p < 0.05) in the diatom classification between these two groups. Based on our experiments, an efficient analysis for automatically identifying and classifying diatoms is urgently required.

A stable model for maximizing the number of significant features

AbstractIdentifying significant features (SFs) is important because they are driving factors of a target outcome. However, it is difficult when they have much more features than observations. The problem becomes more challenging when there are multicollinearity and infrequent common features. In such case, standard explainable methods such as OLS and Lasso often fail to identify many SF. To tackle these problems, we propose a stable model to maximize the number of SFs using selective inference called SFLasso-SI. First, in each point in the regularization path, SFLasso-SI conducts selective inference for conservative significance test. Then, it chooses the optimum value of regularization that maximizes the number of SFs. Our extensive experiments across different types of data - text, image, and video show that our SFLasso-SI can find the biggest number of SFs while maintaining similar prediction accuracy as the benchmarking methods.

Coastal bathymetry in central Dronning Maud Land controls ice shelf stability

Knowledge of the bathymetry of Antarctica’s margins is crucial for models and interpretations of ice-ocean interactions and their influence on ongoing and future sea level change, but remains patchy where ice shelves and multi-year sea ice block measurements. Here, we present a bathymetric model for the central Dronning Maud Land margin, based on a constrained inversion of airborne gravity data. It shows the cavities beneath the region’s two ice shelves to be much deeper than in existing bathymetric compilations, but to be shielded from Warm Deep Water ingress and basal melting by the presence of shallow bathymetric sills along the continental shelf. Over areas of multi-year sea ice, the model returns bathymetric estimates of similar accuracy to gravity interpolation-based methods over open water. Airborne gravity thus presents an opportunity to bathymetrically map hundreds of thousands of square kilometres of the most inaccessible margins of Antarctica at resolutions adequate for regional and global oceanographic and glaciological modelling and interpretation.

Can adverse childhood experiences predict chronic health conditions? Development of trauma-informed, explainable machine learning models.

Decades of research have established the association between adverse childhood experiences (ACEs) and adult onset of chronic diseases, influenced by health behaviors and social determinants of health (SDoH). Machine Learning (ML) is a powerful tool for computing these complex associations and accurately predicting chronic health conditions. Using the 2021 Behavioral Risk Factor Surveillance Survey, we developed several ML models-random forest, logistic regression, support vector machine, Naïve Bayes, and K-Nearest Neighbor-over data from a sample of 52,268 respondents. We predicted 13 chronic health conditions based on ACE history, health behaviors, SDoH, and demographics. We further assessed each variable's importance in outcome prediction for model interpretability. We evaluated model performance via the Area Under the Curve (AUC) score. With the inclusion of data on ACEs, our models outperformed or demonstrated similar accuracies to existing models in the literature that used SDoH to predict health outcomes. The most accurate models predicted diabetes, pulmonary diseases, and heart attacks. The random forest model was the most effective for diabetes (AUC = 0.784) and heart attacks (AUC = 0.732), and the logistic regression model most accurately predicted pulmonary diseases (AUC = 0.753). The strongest predictors across models were age, ever monitored blood sugar or blood pressure, count of the monitoring behaviors for blood sugar or blood pressure, BMI, time of last cholesterol check, employment status, income, count of vaccines received, health insurance status, and total ACEs. A cumulative measure of ACEs was a stronger predictor than individual ACEs. Our models can provide an interpretable, trauma-informed framework to identify and intervene with at-risk individuals early to prevent chronic health conditions and address their inequalities in the U.S.

Comparative analysis of metagenomic classifiers for long-read sequencing datasets

BackgroundLong reads have gained popularity in the analysis of metagenomics data. Therefore, we comprehensively assessed metagenomics classification tools on the species taxonomic level. We analysed kmer-based tools, mapping-based tools and two general-purpose long reads mappers. We evaluated more than 20 pipelines which use either nucleotide or protein databases and selected 13 for an extensive benchmark. We prepared seven synthetic datasets to test various scenarios, including the presence of a host, unknown species and related species. Moreover, we used available sequencing data from three well-defined mock communities, including a dataset with abundance varying from 0.0001 to 20% and six real gut microbiomes.ResultsGeneral-purpose mappers Minimap2 and Ram achieved similar or better accuracy on most testing metrics than best-performing classification tools. They were up to ten times slower than the fastest kmer-based tools requiring up to four times less RAM. All tested tools were prone to report organisms not present in datasets, except CLARK-S, and they underperformed in the case of the high presence of the host’s genetic material. Tools which use a protein database performed worse than those based on a nucleotide database. Longer read lengths made classification easier, but due to the difference in read length distributions among species, the usage of only the longest reads reduced the accuracy. The comparison of real gut microbiome datasets shows a similar abundance profiles for the same type of tools but discordance in the number of reported organisms and abundances between types. Most assessments showed the influence of database completeness on the reports.ConclusionThe findings indicate that kmer-based tools are well-suited for rapid analysis of long reads data. However, when heightened accuracy is essential, mappers demonstrate slightly superior performance, albeit at a considerably slower pace. Nevertheless, a combination of diverse categories of tools and databases will likely be necessary to analyse complex samples. Discrepancies observed among tools when applied to real gut datasets, as well as a reduced performance in cases where unknown species or a significant proportion of the host genome is present in the sample, highlight the need for continuous improvement of existing tools. Additionally, regular updates and curation of databases are important to ensure their effectiveness.

Classifying types of gully changes with unoccupied aircraft vehicles 3D multitemporal point clouds for training of satellite data analysis in Northwest Namibia

AbstractThe development of standardised data acquisition strategies and analytical workflows is crucial to quantify gully changes. In this study, we explore synergies between unoccupied aircraft vehicles (UAV) and satellite remote sensing in order to classify gully morphodynamics. Using Time Series Forest (TSF) and the Sentinel‐1 radar backscatter coefficient (σo), gully scenarios can be classified into four categories: gully topographical change, no change outside gully, no change inside gully, and non‐topographical change. In addition, a Random Forest (RF) classification is performed employing individual features obtained from elevation models and temporally aggregated datasets. Training data are generated from multitemporal UAV‐borne photogrammetric point clouds through a manual segmentation of different gullies in Namibia. This information is transferred from point clouds (sub‐m) to satellite imagery (10 m) generating training data at Sentinel‐1 pixel level. Results indicate that the TSF (on the σo Vertical‐Vertical polarisation) and RF (on temporally aggregated features) perform best when training and testing areas are located within the same geographical extent. Both approaches yield similar Total Accuracy (TA ≈ 79%–80%) and Cohen's Kappa value (Kappa ≈ 0.7), but TSF achieves superior Producer Accuracy (PA = 78.5%) and User Accuracy (UA = 84.6%) for the gully topographical change class. Additionally, the utilisation of TSF in Vertical‐Vertical polarisation is the most effective method if the testing and training areas are in different geographical locations, allowing gully identification with TA &gt; 80% and Kappa = 0.49. However, this method presents limitations to precisely delineate the change types, as dynamics are rain‐driven and therefore are geographically related. In summary, by combining the complementary benefits of UAV‐based and satellite‐based solutions, this study opens a line of research for the study and classification of surface land dynamics and geomorphological feature extraction in regional extents.

Probabilistic analysis of tunnel convergence in spatially variable soil based on Gaussian process regression

Traditional probabilistic analysis of tunnel convergence in spatially variable soil typically involves using the random finite difference method, which is computationally intensive. In this study, we propose an efficient machine learning method based on the Gaussian Process Regression (GPR) model for predicting the horizontal and vertical convergences of a high-speed railway tunnel. In the proposed approach, the soil Young's modulus random field is simplified from a 2D matrix to a 1D vector, which serves as the input of the GPR model. Considering the prior knowledge about the probability distribution of the tunnel convergence, the horizontal and vertical convergences are transformed into two random variables that can be predicted effectively by the GPR model. Notably, the developed GPR model requires only 1.97% of the computational cost to achieve similar prediction accuracy as the 2D-convolutional neural network model. A training dataset comprising 1400 samples can deliver satisfactory prediction performance of the surrogate model under a certain scale of fluctuation (SOF). Moreover, selecting 700 samples from each random field optimally composes a mixed dataset for training the GPR model. The findings of this study suggest that the trained GPR model provides excellent performance for the probabilistic analysis of the tunnel convergences considering the spatially variable soil Young's modulus.

Avatar embodiment prior to motor imagery training in VR does not affect the induced event-related desynchronization: a pilot study

Motor-imagery brain-computer interfaces (MI-BCIs) have the potential to improve motor function in individuals with neurological disorders. Their effectiveness relies on patients’ ability to generate reliable MI-related electroencephalography (EEG) patterns, which can be influenced by the quality of neurofeedback. Virtual Reality (VR) has emerged as a promising tool for enhancing proprioceptive feedback due to its ability to induce a sense of embodiment (SoE), where individuals perceive a virtual body as their own. Although prior research has highlighted the importance of SoE in enhancing MI skills and BCI performance, to date, no study has successfully isolated nor manipulated the SoE in VR before MI training, creating a gap in our understanding of the precise role of the priming effect of embodiment in MI-BCIs. In this study, we aimed to examine whether the virtual SoE when induced, as priming of avatar embodiment, and assessed before MI training, could enhance MI-induced EEG patterns. To achieve this, we divided 26 healthy participants into two groups: the embodied group, which experienced SoE with an avatar before undergoing VR-based MI training, and the non-embodied group, which underwent the same MI training without a prior embodiment phase, serving as a control. We analyzed subjective measures of embodiment, the event-related desynchronization (ERD) power of the sensorimotor rhythms, lateralization of ERD, and offline classification BCI accuracy. Although the embodiment phase effectively induced SoE in the embodied group, both groups exhibited similar MI-induced ERD patterns and BCI classification accuracy. This suggests that the induction of SoE prior to MI training may not significantly influence the training outcomes. Instead, it appears that the integration of embodied VR feedback during MI training itself is sufficient to induce appropriate ERD, as evidenced by previous research.

A Potential Radiomics-Clinical Model for Predicting Failure of Lymph Node Control after Definite Radiotherapy in Locally Advanced Head and Neck Cancer.

Background and Objectives: To optimally predict lymph node (LN) failure after definite radiotherapy (RT) in head and neck cancer (HNC) with LN metastases, this study examined radiomics models extracted from CT images of different periods during RT. Materials and Methods: This study retrospectively collected radiologic and clinical information from patients undergoing definite RT over 60 Gy for HNC with LN metastases from January 2010 to August 2021. The same largest LNs in each patient from the initial simulation CT (CTpre) and the following simulation CT (CTmid) at approximately 40 Gy were indicated as regions of interest. LN failure was defined as residual or recurrent LN within 3 years after the end of RT. After the radiomics features were extracted, the radiomics alone model and the radiomics plus clinical parameters model from the set of CTpre and CTmid were compared. The LASSO method was applied to select features associated with LN failure. Results: Among 66 patients, 17 LN failures were observed. In the radiomics alone model, CTpre and CTmid had similar mean accuracies (0.681 and 0.697, respectively) and mean areas under the curve (AUC) (0.521 and 0.568, respectively). Radiomics features of spherical disproportion, size zone variance, and log minimum 2 were selected for CTpre plus clinical parameters. Volume, energy, homogeneity, and log minimum 1 were selected for CTmid plus clinical parameters. Clinical parameters including smoking, T-stage, ECE, and regression rate of LN were important for both CTpre and CTmid. In the radiomics plus clinical parameters models, the mean accuracy and mean AUC of CTmid (0.790 and 0.662, respectively) were more improved than those of CTpre (0.731 and 0.582, respectively). Conclusions: Both models using CTpre and CTmid were improved by adding clinical parameters. The radiomics model using CTmid plus clinical parameters was the best in predicting LN failure in our preliminary analyses.

Usefulness of the preoperative thoracic inlet angle in comparison to the T1 slope for predicting cervical kyphosis after laminoplasty.

Thoracic inlet angle (TIA) is a sagittal radiographic parameter with a constant value regardless of posture and is significantly correlated with the sagittal balance of the cervical spine. However, the practical use of TIA has not been studied. This study aimed to investigate the usefulness of the preoperative TIA for predicting the development of kyphotic deformity after cervical laminoplasty in comparison to the preoperative T1 slope (T1S). A total of 98 patients who underwent cervical laminoplasty without preoperative kyphotic alignment were included (mean age, 73.7years; 41.8% female). Radiography was evaluated before surgery and at the 2-year follow-up examination. The cervical sagittal parameters were measured on standing radiographs, and the TIA was measured on T2-weighted MRI in a supine position. Cervical alignment with a C2-C7 angle of ≥ 0° was defined as lordosis, and that with an angle of < 0° was defined as kyphosis. Postoperative kyphosis occurred in 11 patients (11.2%). Preoperatively, the kyphosis group showed significantly lower values in the T1S (23.5° vs. 30.3°, p = 0.034) and TIA (76.1° vs. 81.8°, p = 0.042). We performed ROC curve analysis to clarify the impact of the preoperative TIA and T1S on kyphotic deformity after laminoplasty. The optimal cutoff angles for TIA and T1S were 68° and 19°, respectively, with similar diagnostic accuracy. This study demonstrated the clinical utility of the preoperative TIA for predicting the risk of postoperative kyphotic deformity after cervical laminoplasty. These findings suggest the importance of the preoperative assessment of thoracic inlet alignment in cervical spine surgery.

Mice with an autism-associated R451C mutation in neuroligin-3 show intact attention orienting but atypical responses to methylphenidate and atomoxetine in the mouse-Posner task.

Atypical attention orienting has been associated with some autistic symptoms, but the neural mechanisms remain unclear. The human Posner task, a classic attention orienting paradigm, was recently adapted for use with mice, supporting the investigation of the neurobiological underpinnings of atypical attention orienting in preclinical mouse models. The current study tested mice expressing the autism-associated R451C gene mutation in neuroligin-3 (NL3) on the mouse-Posner (mPosner) task. NL3R451C and wild-type (WT) mice were trained to respond to a validly or invalidly cued target on a touchscreen. The cue was a peripheral non-predictive flash in the exogenous task and a central spatially predictive image in the endogenous task. The effects of dopaminergic- and noradrenergic-modulating drugs, methylphenidate and atomoxetine, on task performance were assessed. In both tasks, mice were quicker and more accurate in the validly versus invalidly cued trials, consistent with results in the human Posner task. NL3R451C and WT mice showed similar response times and accuracy but responded differently when treated with methylphenidate and atomoxetine. Methylphenidate impaired exogenous attention disengagement in NL3R451C mice but did not significantly affect WT mice. Atomoxetine impaired endogenous orienting in WT mice but did not significantly affect NL3R451C mice. NL3R451C mice demonstrated intact attention orienting but altered responses to the pharmacological manipulation of the dopaminergic and noradrenergic networks. These findings expand our understanding of the NL3R451C mutation by suggesting that this mutation may lead to selective alterations in attentional processes.

From Staining Techniques to Artificial Intelligence: A Review of Colorectal Polyps Characterization.

This review article provides a comprehensive overview of the evolving techniques in image-enhanced endoscopy (IEE) for the characterization of colorectal polyps, and the potential of artificial intelligence (AI) in revolutionizing the diagnostic accuracy of endoscopy. We discuss the historical use of dye-spray and virtual chromoendoscopy for the characterization of colorectal polyps, which are now being replaced with more advanced technologies. Specifically, we focus on the application of AI to create a "virtual biopsy" for the detection and characterization of colorectal polyps, with potential for replacing histopathological diagnosis. The incorporation of AI has the potential to provide an evolutionary learning system that aids in the diagnosis and management of patients with the best possible outcomes. A detailed analysis of the literature supporting AI-assisted diagnostic techniques for the detection and characterization of colorectal polyps, with a particular emphasis on AI's characterization mechanism, is provided. The benefits of AI over traditional IEE techniques, including the reduction in human error in diagnosis, and its potential to provide an accurate diagnosis with similar accuracy to the gold standard are presented. However, the need for large-scale testing of AI in clinical practice and the importance of integrating patient data into the diagnostic process are acknowledged. In conclusion, the constant evolution of IEE technology and the potential for AI to revolutionize the field of endoscopy in the future are presented.

Descriptor generation from Morgan fingerprint using persistent homology

ABSTRACT In cheminformatics, molecular fingerprints (FPs) are used in various tasks such as regression and classification. However, predictive models often underutilize Morgan FP for regression and related tasks in machine learning. This study introduced descriptors derived from reshaped Morgan FPs using persistent homology for the predictive accuracy improvement. In the solvation free energy (FreeSolv) and water solubility (ESOL) datasets, persistent homology was found to enhance predictive accuracy compared to the use of only Morgan FPs. Notably, using the first-order persistence diagram (PD1) for descriptor generation resulted in more significant improvements than using the zeroth-order persistence diagram (PD0). Combining 4096 bits Morgan FPs with PD1-generated descriptors increased the average coefficient of determination in the Gaussian process regression from 0.597 to 0.667 for FreeSolv and from 0.629 to 0.654 for ESOL. Adjusting the grid size parameter during PD-based descriptor generation is crucial, as finer grids, especially with PD0, generate more descriptors but reduce predictive accuracy. Coarsening the grid or applying principal component analysis (PCA) mitigates overfitting and enhances accuracy. When descriptors were generated from Morgan FPs with randomly shuffled bit positions, coarsening the grid and/or applying PCA achieved similar accuracy improvements as when the persistent homology of the original Morgan FPs was used.