BABianca Marques de Mattos de Araujo

Bianca Marques de Mattos de Araujo

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Bianca Marques de Mattos de Araujo has a total of 67 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Limitations of off-the-shelf multimodal foundation models in secondary caries detection on interproximal radiographs

    by Camila Guimarães da Costa Campos, Cristiano Miranda de Araujo, Luana Beatriz das Portas Luiz, Diulia Pereira Bubna, Angela Graciela Deliga Schroder, Flares Baratto-Filho, Camila Paiva Perin, Bianca Marques de Mattos de Araujo, Thalita de Paris Matos

    To evaluate whether off-the-shelf general-purpose multimodal foundation models can achieve clinically meaningful diagnostic performance in detecting secondary caries on interproximal radiographs without domain-specific training, using GPT-5o (OpenAI) and Gemini 2.5 Flash (Google Inc.) as representative publicly accessible models. A comparative exploratory study used 50 interproximal radiographs of posterior teeth obtained from a dental imaging center, yielding 208 teeth. Each tooth was evaluated twice by each model in independent sessions, totaling 416 observations per model. A specialist in restorative dentistry established the reference standard. Models were assessed in a zero-shot setting using a standardized prompt. Diagnostic performance was evaluated at the tooth level using accuracy, sensitivity, specificity, precision, F1-score, and Cohen’s kappa, with 95% confidence intervals estimated via bootstrap resampling (1000 iterations). Both models demonstrated limited diagnostic performance. GPT-5o showed moderate sensitivity (64.7%) but low specificity (32.3%), resulting in an accuracy of 41.3% and negligible agreement with the reference standard (κ = − 0.021). Gemini 2.5 Flash exhibited higher sensitivity (82.8%) but extremely low specificity (13.0%), with lower accuracy (32.5%) and similarly negligible agreement (κ = − 0.026). Both models showed a strong tendency toward false-positive classifications. Tooth localization performance was poor, with low identification rates in both models. GPT-5o and Gemini 2.5 Flash, when used in their off-the-shelf configurations, demonstrated inadequate diagnostic performance for detecting secondary caries on interproximal radiographs. Their limited accuracy, low specificity, and poor localization ability do not support their direct use for secondary caries detection on interproximal radiographs without domain-specific training and rigorous independent validation.

  • Open Access

    Machine learning–based screening tool for predicting the risk of oropharyngeal dysphagia in patients with ischemic stroke

    by Suzanne Bettega Almeida, Bianca Marques de Mattos de Araujo, Maria Cristina de Alencar Nunes, Luana Beatriz das Portas Luiz, Rayane Délcia da Silva, Bianca Simone Zeigelboim, Karinna Veríssimo Meira Taveira, Elisa Souza Camargo, Rosane Sampaio Santos, Cristiano Miranda de Araujo

    Purpose: Dysphagia frequently complicates recovery after stroke, necessitating effective predictive tools. This study aimed to develop supervised machine learning models to predict dysphagia risk in ischemic stroke patients. Method: We retrospectively analyzed data from 103 ischemic stroke patients, aged over 18 years, using fiberoptic endoscopic evaluation of swallowing (FEES) as the diagnostic standard for dysphagia. Clinical variables included age, sex, EAT-10 score, Functional Oral Intake Scale (FOIS ® ), and Glasgow Coma Scale. Eleven ML algorithms were evaluated: Logistic Regression, SVM, Gradient Boosting, Random Forest, Decision Tree, K-Nearest Neighbors, AdaBoost, Multilayer Perceptron, XGBoost, LightGBM, and CatBoost. Data was split into 70% training and 30% testing sets with stratified sampling. Hyperparameter optimization was via grid search and 5-fold stratified cross-validation. Model performance was measured by AUC, accuracy, precision, recall, and F1-score; 95% confidence intervals were estimated by bootstrapping. Results: Predictive models effectively distinguished dysphagia cases. Gradient Boosting and CatBoost achieved the highest AUC values (AUC = 0.99) on the test set. Logistic Regression performed consistently across test and cross-validation sets, achieving an AUC of 0.95 [95% CI: 0.86–1.00] and precision of 0.91 [95% CI: 0.80–1.00] in the test set, and high values in cross-validation (AUC = 0.93 [95% CI: 0.84–1.00]; precision = 0.92 [95% CI: 0.81–0.99]). Key predictors were FOIS ® , EAT-10, and age. Conclusion: Machine Learning models showed promise for dysphagia screening post-stroke, with Gradient Boosting, CatBoost, and Logistic Regression showing strong clinical potential for decision support and early referral.

  • Open Access

    Use of the mental foramen as a marker for sex estimation using supervised machine learning

    by Ana Laura Borkovski, Milena Sampaio Kuczera, Ana Julia Borkovski, Arthur Felipe Golin Silveira, Isabela Bittencourt Basso, Flares Baratto-Filho, Erika Calvano Küchler, Odilon Guariza-Filho, Bianca Marques de Mattos de Araujo, Angela Graciela Deliga Schroder, et al.

    Background: This study aimed to develop and evaluate supervised machine learning models for sex prediction using morphometric measurements of the mental foramen obtained from digital panoramic radiographs. A total of 535 individuals (45.5% male and 54.5% female) were included. Seven linear measurements related to the position of the mental foramen were collected and initially screened through univariate analysis. Variables showing statistically significant differences between sexes were used as input features to train several classification algorithms. Model optimization was performed using grid search combined with five-fold cross-validation, and performance was assessed using accuracy, precision, recall, F1-score, and Area Under the ROC Curve, with 95% confidence intervals estimated via bootstrapping. Results: All mental foramen–related measurements showed statistically significant differences between sexes (p < 0.001), with moderate to high effect sizes, particularly for distances relative to the inferior border of the mandible. Among the evaluated algorithms, the K-Nearest Neighbors and Support Vector Machine models achieved the best performance. On test data, the K-Nearest Neighbors model reached an accuracy of 0.75 (95% CI: 0.68–0.81), precision of 0.75 (95% CI: 0.68–0.81), and an AUC of 0.78 (95% CI: 0.70–0.85). Similarly, the Support Vector Machine achieved an accuracy of 0.71 (95% CI: 0.63–0.78), precision of 0.71 (95% CI: 0.63–0.78), and an AUC of 0.78 (95% CI: 0.71–0.86). Model performance remained consistent during cross-validation, suggesting stable performance across partitions. Conclusion: The machine learning models developed were able to identify morphometric patterns of the mental foramen associated with sexual dimorphism, showing consistent performance across testing and cross-validation phases. These findings suggest that machine learning can be applied to sex prediction based on mental foramen measurements, serving as an auxiliary tool in forensic dentistry, particularly in scenarios where dental references are unavailable.

  • Open Access

    Prevalence of persistente pain after endodontic treatment

    by Edmundo Duarte Martins, Allya Francisca Marques Borges, Lidiane Oliveira Leão, Bianca Marques de Mattos de Araujo, José Stechman Neto, Tatiana Carvalho Kowaltschuk, Camila de Castro Corrêa Corrêa, Cristiano Miranda de Araújo, Karinna Veríssimo Meira Taveira

    Aim: To determine the prevalence of persistent pain in endodontic patients. Methods: Six electronic databases (EMBASE, LILACS, LIVIVO, PubMed/MEDLINE, Scopus, and Web of Science) were searched, along with the gray literature (Google Scholar, ProQuest Dissertations and Theses). The review used a random-effects model for meta-analysis of proportions and assessed the risk of bias based on the Joanna Briggs Institute Critical Appraisal tool. Results: The review included 16 articles (seven cohort studies and nine cross-sectional studies) published in English from 1994 to 2022. Fifteen of them were selected for meta-analysis. The combined prevalence estimate was 5% (95% CI = 0.01 – 0.10). The risk of bias was moderate in four studies and low in the others. Conclusion: The estimated prevalence of persistent pain after endodontic treatment was approximately one in every 25 patients.

  • Top Cited Papers
    Open Access

    Machine Learning Model for Predicting Postoperative Pain in Cases of Irreversible Pulpitis

    by Pedro Felipe de Jesus Freitas, Ingrid Luiza Mendonça Cunha, Thalita de Paris Matos, Flares Baratto‐Filho, Bianca Marques de Mattos de Araujo, Manoel Damião Sousa‐Neto, Leandro de Moura Martins, Fernando José Herkrath, Emílio Carlos Sponchiado‐Júnior, Cristiano Miranda de Araujo

    Aim: Postoperative pain is a frequent clinical concern following endodontic treatment. This study aimed to develop and validate supervised machine learning models to predict the occurrence of postoperative pain in cases of irreversible pulpitis. Methodology: A prospective sample of 354 patients aged 18 to 60 years undergoing standardised endodontic treatment was analysed. In the original randomised clinical trials from which the data were derived, each patient had only one eligible tooth included. Clinical variables included postoperative pain at 24 and 72 h, treated tooth group, occlusal reduction, photobiomodulation therapy, use of non‐steroidal anti‐inflammatory drugs (NSAIDs), sex and age. Eight supervised machine learning algorithms were trained to predict pain occurrence, including Logistic Regression, Support Vector Machine, Gradient Boosting, Random Forest, Decision Tree, K‐Nearest Neighbours, AdaBoost and Multilayer Perceptron. The dataset was divided into training (70%) and testing (30%) sets using stratified sampling. Class imbalance in the training set, characterised by a lower proportion of cases with moderate or severe pain, was addressed using the Synthetic Minority Oversampling Technique. Hyperparameters were optimised through grid search combined with stratified five‐fold cross validation. Model performance was evaluated using the area under the curve (AUC), accuracy, precision, recall and F1‐score, with 95% confidence intervals estimated by bootstrapping. Results: The predictive models achieved good discrimination of pain outcomes. Logistic Regression showed the best test performance at 24 h (AUC 0.74 [95% CI: 0.61 to 0.85], precision 0.81 [95% CI: 0.73 to 0.88]). At 72 h, the Support Vector Machine achieved the highest performance (AUC 0.81 [95% CI: 0.69 to 0.92], precision 0.88 [95% CI: 0.79 to 0.94]). Age and sex emerged as the most influential predictors across models. Conclusions: Supervised machine learning models demonstrated promising performance for predicting postoperative pain following endodontic treatment. Logistic Regression and Support Vector Machine algorithms presented the most consistent results, supporting their potential clinical application for personalised pain management.


  • Open Access

    Machine learning model for sex determination in adults based on two-dimensional cephalometric measurements of the mandibular symphysis

    by Aline Xavier Ferraz, Pedro Felipe Jesus Freitas, Bianca Marques de Mattos de Araujo, Erika Calvano Küchler, Odilon Guariza-Filho, Karinna Verissimo Meira Taveira, Flares Baratto-Filho, Rosane Sampaio Santos, Angela Graciela Deliga Schroder, Cristiano Miranda de Araujo

    Background: This study aimed to develop machine learning-based predictive models for sex determination using the height and width of the mandibular symphysis from two-dimensional lateral cephalometric images. Data from 495 adult patients who underwent lateral cephalometric radiography at a private clinic in southern Brazil were analyzed. Mandibular symphysis measurements were collected and used to train eight supervised machine-learning algorithms. The models were optimized using Grid Search and evaluated through a 5-fold cross-validation. Metrics such as Area Under the Curve, precision, sensitivity, and F1-score were calculated, with confidence intervals estimated via bootstrapping. Results: The analyses showed that both the height and width of the mandibular symphysis exhibited statistically significant differences between the sexes ( p < 0.05). The models achieved Area Under the Curve values ranging from 0.77 [95% CI: 0.69–0.85] to 0.60 [95% CI: 0.46–0.62] in testing, and from 0.78 [95% CI: 0.76–0.82] to 0.68 [95% CI: 0.64–0.74] in the cross-validation. Among the evaluated algorithms, the SVM, Logistic Regression, and K-Nearest Neighbors models demonstrated the highest predictive performance. Conclusion: Machine learning enabled sex determination based on mandibular symphysis measurements. The results suggest that this approach may serve as a complementary tool in forensic anthropology. Future studies should validate these models in different populations to ensure their generalizability. These findings suggest that the application of artificial intelligence methods can enhance the accuracy of sex determination, contributing to significant advancements in the field.


  • Open Access

    PostOPbot: an artificial intelligence chatbot for postoperative care following tooth extraction

    by Aline Xavier Ferraz, Pedro Felipe de Jesus Freitas, Flavio Magno Gonçalves, Michelle Nascimento Meger, Guilherme Protzek Guariza, Rodrigo Nunes Rached, Julia da Silva Germiniani, Allan Abuabara, Angela Graciela Deliga Schroder, Flares Baratto-Filho, et al.

    Postoperative care following dental extractions requires clear and continuous guidance; however, patient adherence to postoperative instructions remains suboptimal, and access to timely professional support may be limited. This study aimed to develop and evaluate PostOPbot, an artificial intelligence (AI)-based virtual assistant designed to provide standardized postoperative guidance through natural language processing. The system was implemented in Python with a web-based interface and integrated with the ChatGPT-4o application programming interface through a conversational endpoint. Postoperative recommendations were identified through a structured PubMed search and incorporated into a static prompt that defined the system’s operational rules, restricted its scope to postoperative counseling, and guided the interpretation of user input. The rule set was embedded into the code so that each user query was combined with predefined instructions before being processed by the language model. The chatbot was evaluated using three simulated postoperative scenarios representing different time intervals after tooth extraction. For each scenario, the system was expected to provide a predefined set of recommendations, and its performance was assessed by determining whether each item was delivered spontaneously or only after explicit prompting. PostOPbot provided most expected recommendations with few omissions, and missing recommendations were consistently generated upon direct request. Although large language model-based systems have inherent limitations, the findings suggest that PostOPbot may serve as a complementary AI tool to reinforce postoperative guidance without replacing professional evaluation.

  • Open Access

    Association between polymorphisms in the gene encoding parathyroid hormone and orthodontically induced external apical root resorption

    by Veridiana Freitas, Luiz Fernando Fariniuk, Michelle Nascimento Meger, Angela Graciela Deliga Schroder, Camila Guimarães da Costa Campos, Rafaela Scariot, Livia Azeredo A. Antunes, Bianca Marques de Araujo, Christian Kirschneck, Erika Calvano Küchler, et al.

    Introduction The mechanism of action of parathyroid hormone (PTH) in bone resorption supports the idea that polymorphisms in this gene may be associated with root resorption. Therefore, this study aimed to investigate the association between polymorphisms in the gene encoding PTH and the occurrence of severe orthodontically induced external apical root resorption (OIEARR) in patients with severe skeletal malocclusion. Methodology This cross-sectional study involved the evaluation of panoramic radiographs from orthodontic patients and DNA extracted from saliva. Post-treatment external apical root resorption was evaluated according to the Levander and Malmgren classification. Patients exhibiting Grade 4 resorption in at least one permanent tooth were included in the case group.Three polymorphisms in the PTH gene (rs694, rs6256, and rs307247) were investigated. DNA concentration and purity were quantified using a NanoDrop spectrophotometer, and genomic DNA was subsequently genotyped for allelic discrimination using TaqMan-based real-time PCR assays. Allele and genotype frequencies were calculated. Comparisons between groups with and without severe OIEARR were performed using the chi-square test and binary logistic regression analysis with a significance level of 5%. Results A total of 159 patients (53 males and 106 females) were included. Most patients presented a Class III profile. Sex, age, and facial profile were not associated with severe OIEARR (p>0.05). A borderline association was observed in the allele distribution for the polymorphism rs6256 (p=0.057; OR=1.7, 95% CI=1.0–3.2). Binary logistic regression analysis did not show statistical significance (p>0.05). Conclusion The G allele in the rs6256 polymorphism in the PTH gene showed borderline values for severe OIEARR. The multivariable model adjusted for sex and facial profile type did not show statistical associations between the studied PTH genetic polymorphisms and OIEARR.

  • Open Access

    Association of Pulp Stones With Single Nucleotide Polymorphisms of Mineralization‐Related Genes in Orthodontically Treated Patients: A Cross‐Sectional Study

    by Daniel Hemming, Bianca Marques de Mattos de Araujo, Christian Kirschneck, Rafaela Scariot, Camila Paiva Perin, Manoel D. Sousa-Neto, Gabriela Fonseca-Souza, Natanael Henrique Ribeiro Mattos, Michelle Nascimento Meger, Flares Baratto-Filho, et al.

    Aims: The current evidence suggests that the formation of pulp stones (PS) has a genetic background. Therefore, the present study aimed to evaluate the association between PS in orthodontically treated patients and single nucleotide polymorphisms (SNPs) in the mineralization‐related genes ( bone morphogenetic protein 2 [ BMP2] , BMP4 , SMAD6 , and runt-related transcription factor 2 [ RUNX2] ). Materials and Methods: This cross‐sectional study analyzed genomic DNA and panoramic radiographs from patients with a history of orthodontic treatment. Panoramic radiographs were analyzed and PS were identified as well‐ or poorly defined radiopaque opacities located within the pulp chamber of the maxillary and mandibular first and second permanent molars. Genomic DNA obtained from all participants was analyzed. The genetic analysis focused on seven specific SNPs located in genes involved in mineralization pathways: BMP2 (rs235768 and rs1005464), BMP4 (rs17563), SMAD6 (rs3934908 and rs2119261), and RUNX2 (rs59983488 and rs1200425). Comparisons of allele and genotype (codominant, dominant, and recessive models) distributions between patients with and without PS were performed using the chi‐square test. Logistic regression analyses were performed to control for potential confounders such as age and sex for the SNPs located in the RUNX2 gene. One‐way ANOVA was applied to evaluate differences in the mean number of teeth with PS across genotypes. The significance level was established at α = 0.05. Results: The final sample comprised 235 patients: 120 presented with PS, and 115 served as controls. For the SNP rs59983488 in RUNX2 , a statistical association was observed for genotype distribution in the dominant model ( p = 0.033) and for allele distribution ( p = 0.022). For the SNP rs1200425 in RUNX2 , a statistical association was observed for genotype distribution in the recessive model ( p = 0.046; odds ratio [OR] = 2.0, confidence interval [CI] 95% 1.0–3.8). No statistically significant association was observed in the logistic regression analysis. Conclusion: Our study suggests that RUNX2 could have a small effect on the susceptibility of orthodontic patients to develop PS.

  • Open Access

    SOS Dental Trauma—An Artificial Intelligence Chatbot via WhatsApp for Guiding Patients After Dental Trauma

    by Luana Beatriz das Portas Luiz, Diulia Pereira Bubna, Suzanne Bettega Almeida, Flares Baratto‐Filho, Natanael Henrique Ribeiro Mattos, Manoel Damião Sousa‐Neto, Erika Calvano Küchler, Liliane Roskamp, Cristiano Miranda de Araujo, Bianca Marques de Mattos de Araujo

    Background: Traumatic dental injuries (TDIs) are highly prevalent worldwide and require immediate and appropriate management to ensure favorable outcomes. The aim of the study was to develop and validate SOS Dental Trauma, an interactive artificial intelligence‐based chatbot integrated with WhatsApp. SOS Dental Trauma was designed to provide first‐aid guidance based on the International Association of Dental Traumatology (IADT) recommendations until professional care is available. Methods: The chatbot was developed using natural language generation techniques, programmed with a structured clinical script strictly limited to dental trauma management in primary and permanent dentitions. The chatbot was implemented with the FastAPI framework in Python, hosted on a cloud server, and integrated into WhatsApp through Twilio. Validation was performed in two stages: (1) assessment by six independent experts in the field of dental traumatology (three general dentists and three endodontists) who evaluated clarity, appropriateness, and coherence of the interaction flow; and (2) performance testing with simulated clinical cases corresponding to trauma types included in the IADT ToothSOS application. Outcomes were defined as accuracy (correspondence to IADT guidelines) and completeness (including all clinically relevant steps). A minimum sample of 384 interactions was calculated, and a performance threshold of 90% accuracy was established. Cases not reaching this threshold were revised and retested. Results: Experts highlighted the need for more explicit question formulation and adjustments to interaction flow, which were incorporated into the final version. In the initial tests, the chatbot achieved 100% accuracy and completeness for permanent dentition. Nevertheless, it fell below the threshold in cases of primary dentition, particularly luxation and displacement, due to oversimplification of management options. After prompt revisions, the system achieved 100% accuracy and completeness for all trauma types. The chatbot provided structured, accessible, and user‐friendly instructions in alignment with IADT recommendations, demonstrating consistency and reliability. Conclusion: The SOS Dental Trauma chatbot showed high accuracy and completeness in simulated scenarios, offering immediate, evidence‐based guidance for TDIs in both primary and permanent dentitions. Delivering structured recommendations through a widely used platform such as WhatsApp expands access to reliable first‐aid information, supports patients and caregivers in dental trauma management, and has potential as a complementary tool in dental emergencies and health education.