Chroma-dependence of CIEDE2000 acceptability thresholds for dentistry.

OBJECTIVES: Determine visual 50:50% color difference acceptability thresholds (AT) for regions of the dental color space with varying chromaticity. METHODS: A 40-observer panel belonging to two different groups (dentists and laypersons) evaluated 144 dental resin composites pairs (divided in three different sets of 48 pairs according to chroma value: Low Chroma (LC), Medium Chroma (MC) and High Chroma (HC) placed 40 cm away and inside of a viewing cabinet (D65 Standard light source; diffuse/0° geometry). A Takagi-Sugeno-Kang (TSK) fuzzy approximation was used for fitting the data points and calculate the 50:50% acceptability thresholds in CIEDE2000. A paired t-test was used to evaluate the statistical significance between thresholds differences and Bonferroni correction was applied. RESULTS: The CIEDE2000 50:50% AT were ∆E00 = 2.84, ∆E00 = 2.31 and ∆E00 = 1.80 for LC, MC and HC sets of sample pairs, respectively. The 50:50% AT values were statistically significant between the different sets of sample pairs, as well as the 50:50% AT values obtained for different observer groups. CONCLUSIONS: 50:50% CIEDE2000 acceptability thresholds for dentistry are significantly different depending on the chromaticity of the samples. Observers show higher acceptability for more achromatic samples (low chroma value) than for more chromatic samples. CLINICAL SIGNIFICANCE: The difference in the AT for distinct regions of the dental color space can assist professionals as a quality control tool to assess clinical performance and interpret visual and instrumental findings in clinical dentistry, dental research, and subsequent standardization processes.

Applications of artificial intelligence in dentistry: A comprehensive review.

OBJECTIVE: To perform a comprehensive review of the use of artificial intelligence (AI) and machine learning (ML) in dentistry, providing the community with a broad insight on the different advances that these technologies and tools have produced, paying special attention to the area of esthetic dentistry and color research. MATERIALS AND METHODS: The comprehensive review was conducted in MEDLINE/PubMed, Web of Science, and Scopus databases, for papers published in English language in the last 20 years. RESULTS: Out of 3871 eligible papers, 120 were included for final appraisal. Study methodologies included deep learning (DL; n = 76), fuzzy logic (FL; n = 12), and other ML techniques (n = 32), which were mainly applied to disease identification, image segmentation, image correction, and biomimetic color analysis and modeling. CONCLUSIONS: The insight provided by the present work has reported outstanding results in the design of high-performance decision support systems for the aforementioned areas. The future of digital dentistry goes through the design of integrated approaches providing personalized treatments to patients. In addition, esthetic dentistry can benefit from those advances by developing models allowing a complete characterization of tooth color, enhancing the accuracy of dental restorations. CLINICAL SIGNIFICANCE: The use of AI and ML has an increasing impact on the dental profession and is complementing the development of digital technologies and tools, with a wide application in treatment planning and esthetic dentistry procedures.

Non-small-cell lung cancer classification via RNA-Seq and histology imaging probability fusion.

BACKGROUND: Adenocarcinoma and squamous cell carcinoma are the two most prevalent lung cancer types, and their distinction requires different screenings, such as the visual inspection of histology slides by an expert pathologist, the analysis of gene expression or computer tomography scans, among others. In recent years, there has been an increasing gathering of biological data for decision support systems in the diagnosis (e.g. histology imaging, next-generation sequencing technologies data, clinical information, etc.). Using all these sources to design integrative classification approaches may improve the final diagnosis of a patient, in the same way that doctors can use multiple types of screenings to reach a final decision on the diagnosis. In this work, we present a late fusion classification model using histology and RNA-Seq data for adenocarcinoma, squamous-cell carcinoma and healthy lung tissue. RESULTS: The classification model improves results over using each source of information separately, being able to reduce the diagnosis error rate up to a 64% over the isolate histology classifier and a 24% over the isolate gene expression classifier, reaching a mean F1-Score of 95.19% and a mean AUC of 0.991. CONCLUSIONS: These findings suggest that a classification model using a late fusion methodology can considerably help clinicians in the diagnosis between the aforementioned lung cancer cancer subtypes over using each source of information separately. This approach can also be applied to any cancer type or disease with heterogeneous sources of information.

A Comparative Analysis of Machine Learning Techniques for Muon Count in UHECR Extensive Air-Showers.

The main goal of this work is to adapt a Physics problem to the Machine Learning (ML) domain and to compare several techniques to solve it. The problem consists of how to perform muon count from the signal registered by particle detectors which record a mix of electromagnetic and muonic signals. Finding a good solution could be a building block on future experiments. After proposing an approach to solve the problem, the experiments show a performance comparison of some popular ML models using two different hadronic models for the test data. The results show that the problem is suitable to be solved using ML as well as how critical the feature selection stage is regarding precision and model complexity.

Composition Classification of Ultra-High Energy Cosmic Rays.

The study of cosmic rays remains as one of the most challenging research fields in Physics. From the many questions still open in this area, knowledge of the type of primary for each event remains as one of the most important issues. All of the cosmic rays observatories have been trying to solve this question for at least six decades, but have not yet succeeded. The main obstacle is the impossibility of directly detecting high energy primary events, being necessary to use Monte Carlo models and simulations to characterize generated particles cascades. This work presents the results attained using a simulated dataset that was provided by the Monte Carlo code CORSIKA, which is a simulator of high energy particles interactions with the atmosphere, resulting in a cascade of secondary particles extending for a few kilometers (in diameter) at ground level. Using this simulated data, a set of machine learning classifiers have been designed and trained, and their computational cost and effectiveness compared, when classifying the type of primary under ideal measuring conditions. Additionally, a feature selection algorithm has allowed for identifying the relevance of the considered features. The results confirm the importance of the electromagnetic-muonic component separation from signal data measured for the problem. The obtained results are quite encouraging and open new work lines for future more restrictive simulations.

Gingival shade guides: Colorimetric and spectral modeling.

OBJECTIVE: To design colorimetric and spectral models of gingival shade guides that adequately represent the color of human gingiva. MATERIALS AND METHODS: A previously compiled database on the spectral reflectance of healthy keratinized gingiva was used for optimization. Coverage Error (CE) and Maximal Error (ME) were optimized using CIELAB and CIEDE2000 color difference formulas. A two-phase process included an FCM algorithm and a nonlinear optimization. A t test was used to compare the performance of the different numbers of clusters/tabs in gingival shade guide models (α = .05). RESULTS: CIELAB CE and ME for shade guide models with 3 to 6 clusters ranged from 3.1 to 3.9 (P = .028 for 3 vs. 4; and P = .033 for 5 vs. 6 cluster/tab comparison), while the corresponding CIEDE2000 range was from 2.1 to 2.8 (P < .001 for 3 vs. 4 tabs; P < .025 for 4 vs. 5; and P = 0.029 for 5 vs. 6 tab comparisons). The percentage of data points exhibiting a CIELAB color difference lower than the acceptability threshold ranged from 48.7% to 71.4%, and from 52.9% to 82.4%. for CIEDE2000. CONCLUSIONS: An increase in the number of clusters in the gingival shade guide models was associated with a decrease in coverage error (better match) to human gingiva. Gingival shade guide models with only 4 tabs provided a CIELAB and CIEDE2000 coverage error lower than the acceptability threshold for gingival color. Spectral clustering of human gingiva was determined to be valid. CIEDE2000 color difference formula outperformed the CIELAB formula in the optimization process. CLINICAL SIGNIFICANCE: Providing a shade guide model with a small number of tabs and a coverage error lower than the 50:50% acceptability threshold would be an optimal solution for shade matching in dentistry. However, no actual gingival or tooth shade guide complies with this. The clustering method, with optimization of both Coverage Error and Maximal Error and spectral clustering that enables more reliable color formulation of cluster representatives of shade guide models, represents an advance when it comes to computer modeling in dentistry.

Color difference thresholds for computer-simulated human Gingiva.

OBJECTIVE: To determine 50:50% perceptibility threshold (PT) and 50:50% acceptability threshold (AT) for computer-simulated samples of human gingiva using CIEDE2000 and CIELAB color difference formulas. MATERIALS AND METHODS: Each of the 60 pairs of simulated human gingiva was displayed on a calibrated monitor, together with three pairs of upper central incisors of different lightness. The color of gingiva left and right from the midline was compared. A total of 30 observers (15 dentists, 15 laypersons) participated in the study. CIEDE2000 and CIELAB formulas were used to calculate the thresholds and a Takagi-Sugeno-Kang Fuzzy Approximation model was used as fitting procedure. Paired t-test (α = 0.05) was used in evaluation of statistical significance of differences. RESULTS: The PT and AT for CIEDE2000 and 95% confidence intervals were 1.1 [0.4, 1.7] and 2.8 [1.8, 4.0], respectively. Corresponding CIELAB values were 1.7 [0.2, 2.6] and 3.7 [2.1, 5.7]. Significant differences (P < .01) were recorded between PT and AT, between the corresponding threshold values in CIEDE2000 and CIELAB formulas as well as between dentists and laypersons. CONCLUSIONS: The difference between the perceptibility and acceptability threshold for gingiva was statistically significant in both CIEDE2000 and CIELAB. The same was true for differences between the corresponding thresholds using two color difference formulas, and between dentists and laypersons. Visual thresholds of human gingiva were not dependent upon lightness of adjacent teeth. Overall, CIEDE2000 color difference formula provided better fit than CIELAB formula in the evaluation of color difference thresholds of human gingiva. CLINICAL SIGNIFICANCE: The data on visual thresholds for healthy human gingiva can be used as quality control tool/guide for selection and evaluation of dental materials, interpretation of color-related findings in clinical dentistry and research, and for standardization in dentistry. It is of particular value that this study was designed based on in-vivo color evaluation of healthy keratinized gingiva of subjects of different ethnicities, age groups, and gender.

Integration of RNA-Seq data with heterogeneous microarray data for breast cancer profiling.

BACKGROUND: Nowadays, many public repositories containing large microarray gene expression datasets are available. However, the problem lies in the fact that microarray technology are less powerful and accurate than more recent Next Generation Sequencing technologies, such as RNA-Seq. In any case, information from microarrays is truthful and robust, thus it can be exploited through the integration of microarray data with RNA-Seq data. Additionally, information extraction and acquisition of large number of samples in RNA-Seq still entails very high costs in terms of time and computational resources.This paper proposes a new model to find the gene signature of breast cancer cell lines through the integration of heterogeneous data from different breast cancer datasets, obtained from microarray and RNA-Seq technologies. Consequently, data integration is expected to provide a more robust statistical significance to the results obtained. Finally, a classification method is proposed in order to test the robustness of the Differentially Expressed Genes when unseen data is presented for diagnosis. RESULTS: The proposed data integration allows analyzing gene expression samples coming from different technologies. The most significant genes of the whole integrated data were obtained through the intersection of the three gene sets, corresponding to the identified expressed genes within the microarray data itself, within the RNA-Seq data itself, and within the integrated data from both technologies. This intersection reveals 98 possible technology-independent biomarkers. Two different heterogeneous datasets were distinguished for the classification tasks: a training dataset for gene expression identification and classifier validation, and a test dataset with unseen data for testing the classifier. Both of them achieved great classification accuracies, therefore confirming the validity of the obtained set of genes as possible biomarkers for breast cancer. Through a feature selection process, a final small subset made up by six genes was considered for breast cancer diagnosis. CONCLUSIONS: This work proposes a novel data integration stage in the traditional gene expression analysis pipeline through the combination of heterogeneous data from microarrays and RNA-Seq technologies. Available samples have been successfully classified using a subset of six genes obtained by a feature selection method. Consequently, a new classification and diagnosis tool was built and its performance was validated using previously unseen samples.

Predictive algorithms for determination of reflectance data from quantity of pigments within experimental dental resin composites.

BACKGROUND: Being able to estimate (predict) the final spectrum of reflectance of a biomaterial, especially when the final color and appearance are fundamental for their clinical success (as is the case of dental resin composites), could be a very useful tool for the industrial development of these type of materials. The main objective of this study was the development of predictive models which enable the determination of the reflectance spectrum of experimental dental resin composites based on type and quantity of pigments used in their chemical formulation. METHODS: 49 types of experimental dental resin composites were formulated as a mixture of organic matrix, inorganic filler, photo activator and other components in minor quantities (accelerator, inhibitor, fluorescent agent and 4 types of pigments). Spectral reflectance of all samples were measured, before and after artificial chromatic aging, using a spectroradiometer. A Multiple Nonlinear Regression Model (MNLR) was used to predict the values of the Reflectance Factors values in the visible range (380 nm-780 nm), before and after aging, from % Pigment (%P1, %P2, %P3 and %P4) within the formulation. RESULTS: The average value of the prediction error of the model was 3.46% (SD: 1.82) across all wavelengths for samples before aging and 3.54% (SD: 1.17) for samples after aging. The differences found between the predicted and measured values of the chromatic coordinates are smaller than the acceptability threshold and, in some cases, are even below the perceptibility threshold. CONCLUSIONS: Within the framework of this pilot study, the nonlinear predictive models developed allow the prediction, with a high degree of accuracy, of the reflectance spectrum of the experimental dental resin composites.

Gene Expression Analysis for Uterine Cervix and Corpus Cancer Characterization.

The analysis of gene expression quantification data is a powerful and widely used approach in cancer research. This work provides new insights into the transcriptomic changes that occur in healthy uterine tissue compared to those in cancerous tissues and explores the differences associated with uterine cancer localizations and histological subtypes. To achieve this, RNA-Seq data from the TCGA database were preprocessed and analyzed using the KnowSeq package. Firstly, a kNN model was applied to classify uterine cervix cancer, uterine corpus cancer, and healthy uterine samples. Through variable selection, a three-gene signature was identified (VWCE, CLDN15, ADCYAP1R1), achieving consistent 100% test accuracy across 20 repetitions of a 5-fold cross-validation. A supplementary similar analysis using miRNA-Seq data from the same samples identified an optimal two-gene miRNA-coding signature potentially regulating the three-gene signature previously mentioned, which attained optimal classification performance with an 82% F1-macro score. Subsequently, a kNN model was implemented for the classification of cervical cancer samples into their two main histological subtypes (adenocarcinoma and squamous cell carcinoma). A uni-gene signature (ICA1L) was identified, achieving 100% test accuracy through 20 repetitions of a 5-fold cross-validation and externally validated through the CGCI program. Finally, an examination of six cervical adenosquamous carcinoma (mixed) samples revealed a pattern where the gene expression value in the mixed class aligned closer to the histological subtype with lower expression, prompting a reconsideration of the diagnosis for these mixed samples. In summary, this study provides valuable insights into the molecular mechanisms of uterine cervix and corpus cancers. The newly identified gene signatures demonstrate robust predictive capabilities, guiding future research in cancer diagnosis and treatment methodologies.

Novel methodology for detecting and localizing cancer area in histopathological images based on overlapping patches.

Cancer disease is one of the most important pathologies in the world, as it causes the death of millions of people, and the cure of this disease is limited in most cases. Rapid spread is one of the most important features of this disease, so many efforts are focused on its early-stage detection and localization. Medicine has made numerous advances in the recent decades with the help of artificial intelligence (AI), reducing costs and saving time. In this paper, deep learning models (DL) are used to present a novel method for detecting and localizing cancerous zones in WSI images, using tissue patch overlay to improve performance results. A novel overlapping methodology is proposed and discussed, together with different alternatives to evaluate the labels of the patches overlapping in the same zone to improve detection performance. The goal is to strengthen the labeling of different areas of an image with multiple overlapping patch testing. The results show that the proposed method improves the traditional framework and provides a different approach to cancer detection. The proposed method, based on applying 3x3 step 2 average pooling filters on overlapping patch labels, provides a better result with a 12.9% correction percentage for misclassified patches on the HUP dataset and 15.8% on the CINIJ dataset. In addition, a filter is implemented to correct isolated patches that were also misclassified. Finally, a CNN decision threshold study is performed to analyze the impact of the threshold value on the accuracy of the model. The alteration of the threshold decision along with the filter for isolated patches and the proposed method for overlapping patches, corrects about 20% of the patches that are mislabeled in the traditional method. As a whole, the proposed method achieves an accuracy rate of 94.6%. The code is available at https://github.com/sergioortiz26/Cancer_overlapping_filter_WSI_images.

Development of Thickness-Dependent Predictive Methods for the Estimation of the CIEL*a*b* Color Coordinates of Monolithic and Layered Dental Resin Composites.

Usually, dentin and enamel shades are layered in dental restorations with the purpose of mimicking the natural appearance of teeth. The main objective of this study was to develop and assess accuracy of a color-prediction method for both monolithic and layered dental resin-based composites with varying shades and under different illuminants. A total of 15 different shades of VITAPAN Excell, VITAPAN Dentine and VITA Physiodens as well as VITA Enamel of five different thicknesses (0.5-2.5 mm range) were used to manufacture monolithic and layered samples. A non-contact spectroradiometer with CIE 45∘/0∘ geometry was used to measure the color of all samples over a standard ceramic black background. Second-degree polynomial regression was used as predictive method for CIE-L*a*b* color coordinates. Performance of predictive models was tested using the CIEDE2000 total color difference formula (ΔE00), while accuracy was evaluated by comparative assessment of ΔE00 with corresponding 50:50% acceptability (AT00) and perceptibly (PT00) thresholds for dentistry. A mean color difference between measured (real) and predicted color of ΔE00=1.71, with 62.86% of the color differences below AT00 and 28.57% below PT00, was registered for monolithic samples. For bi-layered samples, the mean color difference was roughly ΔE00=0.50, with generally 100% and more than 85% of the estimations below AT00 and PT00, respectively. The predictive method allowed highly accurate color estimations for both monolithic and layered dental resin-based composites with varying thicknesses and under different illuminations. These results could be useful to maximize the clinical success of dental restorations.

Performance comparison between multi-center histopathology datasets of a weakly-supervised deep learning model for pancreatic ductal adenocarcinoma detection.

BACKGROUND: Pancreatic ductal carcinoma patients have a really poor prognosis given its difficult early detection and the lack of early symptoms. Digital pathology is routinely used by pathologists to diagnose the disease. However, visually inspecting the tissue is a time-consuming task, which slows down the diagnostic procedure. With the advances occurred in the area of artificial intelligence, specifically with deep learning models, and the growing availability of public histology data, clinical decision support systems are being created. However, the generalization capabilities of these systems are not always tested, nor the integration of publicly available datasets for pancreatic ductal carcinoma detection (PDAC). METHODS: In this work, we explored the performace of two weakly-supervised deep learning models using the two more widely available datasets with pancreatic ductal carcinoma histology images, The Cancer Genome Atlas Project (TCGA) and the Clinical Proteomic Tumor Analysis Consortium (CPTAC). In order to have sufficient training data, the TCGA dataset was integrated with the Genotype-Tissue Expression (GTEx) project dataset, which contains healthy pancreatic samples. RESULTS: We showed how the model trained on CPTAC generalizes better than the one trained on the integrated dataset, obtaining an inter-dataset accuracy of 90.62% ± 2.32 and an outer-dataset accuracy of 92.17% when evaluated on TCGA + GTEx. Furthermore, we tested the performance on another dataset formed by tissue micro-arrays, obtaining an accuracy of 98.59%. We showed how the features learned in an integrated dataset do not differentiate between the classes, but between the datasets, noticing that a stronger normalization might be needed when creating clinical decision support systems with datasets obtained from different sources. To mitigate this effect, we proposed to train on the three available datasets, improving the detection performance and generalization capabilities of a model trained only on TCGA + GTEx and achieving a similar performance to the model trained only on CPTAC. CONCLUSIONS: The integration of datasets where both classes are present can mitigate the batch effect present when integrating datasets, improving the classification performance, and accurately detecting PDAC across different datasets.

Synthetic whole-slide image tile generation with gene expression profile-infused deep generative models.

In this work, we propose an approach to generate whole-slide image (WSI) tiles by using deep generative models infused with matched gene expression profiles. First, we train a variational autoencoder (VAE) that learns a latent, lower-dimensional representation of multi-tissue gene expression profiles. Then, we use this representation to infuse generative adversarial networks (GANs) that generate lung and brain cortex tissue tiles, resulting in a new model that we call RNA-GAN. Tiles generated by RNA-GAN were preferred by expert pathologists compared with tiles generated using traditional GANs, and in addition, RNA-GAN needs fewer training epochs to generate high-quality tiles. Finally, RNA-GAN was able to generalize to gene expression profiles outside of the training set, showing imputation capabilities. A web-based quiz is available for users to play a game distinguishing real and synthetic tiles: https://rna-gan.stanford.edu/, and the code for RNA-GAN is available here: https://github.com/gevaertlab/RNA-GAN.

Comparison of Global Evaluative Assessment of Robotic Surgery with objective performance indicators for the assessment of skill during robotic-assisted thoracic surgery.

BACKGROUND: The Global Evaluative Assessment of Robotic Skills is a popular but ultimately subjective assessment tool in robotic-assisted surgery. An alternative approach is to record system or console events or calculate instrument kinematics to derive objective performance indicators. The aim of this study was to compare these 2 approaches and correlate the Global Evaluative Assessment of Robotic Skills with different types of objective performance indicators during robotic-assisted lobectomy. METHODS: Video, system event, and kinematic data were recorded from the robotic surgical system during left upper lobectomy on a standardized perfused and pulsatile ex vivo porcine heart-lung model. Videos were segmented into steps, and the superior vein dissection was graded independently by 2 blinded expert surgeons with Global Evaluative Assessment of Robotic Skills. Objective performance indicators representing categories for energy use, event data, movement, smoothness, time, and wrist articulation were calculated for the same task and compared to Global Evaluative Assessment of Robotic Skills scores. RESULTS: Video and data from 51 cases were analyzed (44 fellows, 7 attendings). Global Evaluative Assessment of Robotic Skills scores were significantly higher for attendings (P < .05), but there was a significant difference in raters' scores of 31.4% (defined as >20% difference in total score). The interclass correlation was 0.44 for 1 rater and 0.61 for 2 raters. Objective performance indicators correlated with Global Evaluative Assessment of Robotic Skills to varying degrees. The most highly correlated Global Evaluative Assessment of Robotic Skills domain was efficiency. Instrument movement and smoothness were highly correlated among objective performance indicator categories. Of individual objective performance indicators, right-hand median jerk, an objective performance indicator of change of acceleration, had the highest correlation coefficient (0.55). CONCLUSION: There was a relatively poor overall correlation between the Global Evaluative Assessment of Robotic Skills and objective performance indicators. However, both appear strongly correlated for certain metrics such as efficiency and smoothness. Objective performance indicators may be a potentially more quantitative and granular approach to assessing skill, given that they can be calculated mathematically and automatically without subjective interpretation.

Determination of the Severity and Percentage of COVID-19 Infection through a Hierarchical Deep Learning System.

The coronavirus disease 2019 (COVID-19) has caused millions of deaths and one of the greatest health crises of all time. In this disease, one of the most important aspects is the early detection of the infection to avoid the spread. In addition to this, it is essential to know how the disease progresses in patients, to improve patient care. This contribution presents a novel method based on a hierarchical intelligent system, that analyzes the application of deep learning models to detect and classify patients with COVID-19 using both X-ray and chest computed tomography (CT). The methodology was divided into three phases, the first being the detection of whether or not a patient suffers from COVID-19, the second step being the evaluation of the percentage of infection of this disease and the final phase is to classify the patients according to their severity. Stratification of patients suffering from COVID-19 according to their severity using automatic systems based on machine learning on medical images (especially X-ray and CT of the lungs) provides a powerful tool to help medical experts in decision making. In this article, a new contribution is made to a stratification system with three severity levels (mild, moderate and severe) using a novel histogram database (which defines how the infection is in the different CT slices for a patient suffering from COVID-19). The first two phases use CNN Densenet-161 pre-trained models, and the last uses SVM with LDA supervised learning algorithms as classification models. The initial stage detects the presence of COVID-19 through X-ray multi-class (COVID-19 vs. No-Findings vs. Pneumonia) and the results obtained for accuracy, precision, recall, and F1-score values are 88%, 91%, 87%, and 89%, respectively. The following stage manifested the percentage of COVID-19 infection in the slices of the CT-scans for a patient and the results in the metrics evaluation are 0.95 in Pearson Correlation coefficient, 5.14 in MAE and 8.47 in RMSE. The last stage finally classifies a patient in three degrees of severity as a function of global infection of the lungs and the results achieved are 95% accurate.

Machine-Learning-Based Late Fusion on Multi-Omics and Multi-Scale Data for Non-Small-Cell Lung Cancer Diagnosis.

Differentiation between the various non-small-cell lung cancer subtypes is crucial for providing an effective treatment to the patient. For this purpose, machine learning techniques have been used in recent years over the available biological data from patients. However, in most cases this problem has been treated using a single-modality approach, not exploring the potential of the multi-scale and multi-omic nature of cancer data for the classification. In this work, we study the fusion of five multi-scale and multi-omic modalities (RNA-Seq, miRNA-Seq, whole-slide imaging, copy number variation, and DNA methylation) by using a late fusion strategy and machine learning techniques. We train an independent machine learning model for each modality and we explore the interactions and gains that can be obtained by fusing their outputs in an increasing manner, by using a novel optimization approach to compute the parameters of the late fusion. The final classification model, using all modalities, obtains an F1 score of 96.81±1.07, an AUC of 0.993±0.004, and an AUPRC of 0.980±0.016, improving those results that each independent model obtains and those presented in the literature for this problem. These obtained results show that leveraging the multi-scale and multi-omic nature of cancer data can enhance the performance of single-modality clinical decision support systems in personalized medicine, consequently improving the diagnosis of the patient.

CIEDE2000 lightness, chroma and hue human gingiva thresholds.

OBJECTIVE: To determine the visual 50:50% perceptibility and acceptability CIEDE2000 lightness, chroma and hue human gingiva thresholds. METHODS: A psychophysical experiment based on visual assessments of simulated images of human gingiva on a calibrated display was performed. A 20-obsever panel (dentists and laypersons; n=10) evaluated three subsets of simulated human gingiva: lightness subset (|ΔL'/ΔE00|≥ 0.98); chroma subset (|ΔC'/ΔE00|≥ 0.98) and hue subset (|ΔH'/ΔE00|≥ 0.96) using ΔE00< 5 units. A Takagi-Sugeno-Kang (TSK) Fuzzy Approximation model was used as fitting procedure, and 50:50% perceptibility threshold (PT) and acceptability threshold (AT) were calculated. Data was statistically analyzed using t-test (p ≤ 0.05). RESULTS: The 50:50% PT were ΔL' = 0.74 (95% confidence interval (CI) 0.39-1.09); ΔC' = 1.10 (95% CI 0.57-1.46); ΔH' = 2.40 (95% CI 1.66->3.85). The 50:50% AT were ΔL' = 2.57 (95% CI 2.00-3.06); ΔC' = 2.70 (95% CI 2.19-3.38). AT ΔH' may be considered no computable. PT values were statistically significant among the three metric differences (p ≤ 0.05). No difference was found between observers for PT values. CONCLUSIONS: Statistically differences in perceptual limit were found among hue, lightness and chroma for human gingiva. Thus, the observers seem to show lower sensitivity for changes in hue (ΔH') than in chroma (ΔC') and in lightness (ΔL') in the gingiva color space. CLINICAL SIGNIFICANCE: PT and AT for lightness, chroma and hue specific for human gingiva should be used when evaluating natural gingiva, pink gingival shade guides or pink materials, since the thresholds of perception and acceptability for teeth are not suitable.

Reflectance and color prediction of dental material monolithic samples with varying thickness.

OBJECTIVE: To assess accuracy of reflectance reconstruction and color estimation of different dental materials with varying thicknesses using Principal Component Analysis (PCA). METHOD: A1, A2, A3, A3.5, B2, C2 and D2 shades and 5 thicknesses (within 0.5-2.5 mm range) of Vita Suprinity (VS-PC) and Vitapan Dentine (VD), were used. Reflectance measurements were performed over black background using a non-contact spectroradiometer with CIE 45∘∕0∘ geometry. A PCA based algorithm was proposed to reconstruct spectral data and color of samples, using both extrapolation and interpolation approaches. Root Mean Square Error (RMSE), Goodness of Fit (GFC), correlation coefficient (R2) as well as ΔE00 with corresponding 50:50% acceptability and perceptibly thresholds (AT and PT) were used as performance assessment. RESULTS: The interpolation approach provided an average RMSE = 0.01 and GFC > 0.999 when comparing predicted and measured spectral reflectances for both materials, while for the extrapolation approach RMSE = 0.02 and GFC > 0.999. Interpolation approach also resulted in lower overall mean color difference ΔE00 = 0.8 (ΔE00 = 0.9 for VS-PC and ΔE00 = 0.7 for VD), while using extrapolation approach resulted in higher overall mean color difference ΔE00 = 1.6, although below the AT (ΔE00 = 1.8 for VS-PC and ΔE00 = 1.5 for VD). Correlation values between predicted and measured spectral reflectances of R2 = 0.987 and R2 = 0.993 were globally obtained for VS-PC and VD, respectively. SIGNIFICANCE: The proposed PCA-based algorithm is able to efficiently predict reflectance spectrum and color of monolithic samples of different dental materials with varying thickness. It can be used to optimize dental materials manufacturing processes and to improve chromatic accuracy of clinical dental restorations.

Validation of a Hyperspectral Imaging System for Color Measurement of In-Vivo Dental Structures.

A full comprehension of colorimetric relationships within and between teeth is key for aesthetic success of a dental restoration. In this sense, hyperspectral imaging can provide point-wise reliable measurements of the tooth surface, which can serve for this purpose. The aim of this study was to use a hyperspectral imaging system for the colorimetric characterization of 4 in-vivo maxillary anterior teeth and to cross-check the results with similar studies carried out with other measuring systems in order to validate the proposed capturing protocol. Hyperspectral reflectance images (Specim IQ), of the upper central (UCI) and lateral incisors (ULI), were captured on 30 participants. CIE-L*a*b* values were calculated for the incisal (I), middle (M) and cervical (C) third of each target tooth. ΔEab* and ΔE00 total color differences were computed between different tooth areas and adjacent teeth, and evaluated according to the perceptibility (PT) and acceptability (AT) thresholds for dentistry. Non-perceptible color differences were found between UCIs and ULIs. Mean color differences between UCI and ULI exceeded AT (ΔEab* = 7.39-7.42; ΔE00 = 5.71-5.74) in all cases. Large chromatic variations between I, M and C areas of the same tooth were registered (ΔEab* = 5.01-6.07 and ΔE00 = 4.07-5.03; ΔEab* = 5.80-8.16 and ΔE00 = 4.37-5.15; and ΔEab* = 5.42-5.92 and ΔE00 = 3.87-4.16 between C and M, C and I and M and I, respectively). The use of a hyperspectral camera has proven to be a reliable and effective method for color evaluation of in-vivo natural teeth.