Otso Tirkkonen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Otso Tirkkonen has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Gil Guilherme Gasparello
Research Unit of Population Health University of Oulu Oulu Finland
4 shared publications - Orlando Motohiro Tanaka
Pontifícia Universidade Católica do Paraná, Curitiba, Brazil
3 shared publications - Elina Väyrynen
Research Unit of Population Health, Faculty of Medicine, University of Oulu, Aapistie 3, Oulu, FI
2 shared publications - Henna Tiensuu
University of Oulu, Oulu, Finland
2 shared publications - Jaakko Suutala
Faculty of Information and Electrical Engineering, University of Oulu, Oulu, Finland
2 shared publications - Katri Kukkola
University of Oulu
2 shared publications
Latest Publications
- Open Access
- 19 September 2026
- Published by Wiley in European Journal of Dental Education
Functional and Existential AI Dependence in Dental Students and Professionals: A Hybrid Network Analysis
Introduction: Generative artificial intelligence (AI) tools are now embedded in dental teaching and clinical learning, yet little is known about how reliance on them is psychologically organised in dental students and professionals. This study mapped how two forms of AI dependence relate to problematic social media use, anxiety, and attentional control. Materials and Methods: Cross‐sectional online survey of 342 Brazilian dental students and professionals. Latent factor scores from confirmatory factor analysis were combined with observed variables in a hybrid network. Network structure, centrality, and stability were estimated with regularised Gaussian graphical modelling and bootstrapping. Results: Functional Dependence (FD) and Existential Dependence (ED) were positively associated with each other (partial r = 0.447). FD was negatively and ED positively associated with anxiety ( r = −0.146 and 0.264); both were positively associated with problematic social media use, ED more strongly ( r = 0.237 vs. 0.168). Attentional focusing was negatively associated with anxiety and problematic social media use, but not with either form of AI dependence. Centrality estimates were not stable (CS‐coefficient = 0.05, below the recommended 0.25) and are not used to rank nodes. Discussion: Instrumental and emotional reliance on AI did not behave as one construct. Only ED was positively associated with anxiety. Conclusions: Dental education programmes introducing AI should address not only technical competence but also the emotional and epistemic relationship students develop with these tools. The design is cross‐sectional, so all relationships are associations, not directional effects.
- 19 September 2026
- Published by Wiley in European Journal of Dental Education
Open AccessFunctional and Existential AI Dependence in Dental Students and Professionals: A Hybrid Network Analysis
Introduction: Generative artificial intelligence (AI) tools are now embedded in dental teaching and clinical learning, yet little is known about how reliance on them is psychologically organised in dental students and professionals. This study mapped how two forms of AI dependence relate to problematic social media use, anxiety, and attentional control. Materials and Methods: Cross‐sectional online survey of 342 Brazilian dental students and professionals. Latent factor scores from confirmatory factor analysis were combined with observed variables in a hybrid network. Network structure, centrality, and stability were estimated with regularised Gaussian graphical modelling and bootstrapping. Results: Functional Dependence (FD) and Existential Dependence (ED) were positively associated with each other (partial r = 0.447). FD was negatively and ED positively associated with anxiety ( r = −0.146 and 0.264); both were positively associated with problematic social media use, ED more strongly ( r = 0.237 vs. 0.168). Attentional focusing was negatively associated with anxiety and problematic social media use, but not with either form of AI dependence. Centrality estimates were not stable (CS‐coefficient = 0.05, below the recommended 0.25) and are not used to rank nodes. Discussion: Instrumental and emotional reliance on AI did not behave as one construct. Only ED was positively associated with anxiety. Conclusions: Dental education programmes introducing AI should address not only technical competence but also the emotional and epistemic relationship students develop with these tools. The design is cross‐sectional, so all relationships are associations, not directional effects.
- Open Access
- 18 May 2026
- Published by Wiley in Orthodontics & Craniofacial Research
- Vol. 29 (5) , 963-971
Comparing Facial Attractiveness and Professional Acceptance Between Real and AI ‐Generated Orthodontic Treatment Outcome Images
Background: Facial attractiveness influences social and professional perceptions, including competence and employability. With the increasing use of artificial intelligence (AI) to generate photorealistic faces, it is important to understand how these images compare to real orthodontic treatment outcomes. Objective: To compare perceptions of real versus AI‐generated orthodontic treatment results in terms of attractiveness, employability, and distinguishability. Methods: A cross‐sectional online survey included 288 Brazilian adults. Participants: Dentists ( n = 76), dental students ( n = 63), and laypeople ( n = 149). Each evaluated three sets of images (pre‐treatment, post‐treatment, and AI‐generated post‐treatment) for three patients. Participants rated facial attractiveness (VAS 0–100), assessed employability, and identified whether the image was AI‐generated. Results: AI‐generated images were rated significantly more attractive than real post‐treatment ones (mean VAS: AI = 68.8; Real = 54.4; p < 0.001). Dentists, students, and laypeople all rated AI images higher. Employability was also greater for AI images (50.4%) than real ones (39.1%) ( p = 0.007). Misclassification was common, with dental students achieving the highest accuracy (66.8%). Conclusion: Participants showed limited ability to accurately distinguish between real and AI‐generated images. AI‐generated images were perceived as more attractive and had a greater positive impact on employability judgements compared to real post‐treatment images.
- 18 May 2026
- Published by Wiley in Orthodontics & Craniofacial Research
- Vol. 29 (5) , 963-971
Open AccessComparing Facial Attractiveness and Professional Acceptance Between Real and AI ‐Generated Orthodontic Treatment Outcome Images
Background: Facial attractiveness influences social and professional perceptions, including competence and employability. With the increasing use of artificial intelligence (AI) to generate photorealistic faces, it is important to understand how these images compare to real orthodontic treatment outcomes. Objective: To compare perceptions of real versus AI‐generated orthodontic treatment results in terms of attractiveness, employability, and distinguishability. Methods: A cross‐sectional online survey included 288 Brazilian adults. Participants: Dentists ( n = 76), dental students ( n = 63), and laypeople ( n = 149). Each evaluated three sets of images (pre‐treatment, post‐treatment, and AI‐generated post‐treatment) for three patients. Participants rated facial attractiveness (VAS 0–100), assessed employability, and identified whether the image was AI‐generated. Results: AI‐generated images were rated significantly more attractive than real post‐treatment ones (mean VAS: AI = 68.8; Real = 54.4; p < 0.001). Dentists, students, and laypeople all rated AI images higher. Employability was also greater for AI images (50.4%) than real ones (39.1%) ( p = 0.007). Misclassification was common, with dental students achieving the highest accuracy (66.8%). Conclusion: Participants showed limited ability to accurately distinguish between real and AI‐generated images. AI‐generated images were perceived as more attractive and had a greater positive impact on employability judgements compared to real post‐treatment images.
- Open Access
- 21 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Perception of AI-generated smile versus real orthodontic treatment outcomes among dentists, students, and laypeople
In orthodontics, the increasing use of AI-generated smile images in patient communication raises ethical and practical concerns about user perception and misinterpretation of these visuals. This cross-sectional, non-probabilistic sample study evaluated the ability of dentists, dental students, and laypeople to distinguish between real and AI-generated orthodontic smile images, and their perceived attractiveness. The final sample consisted of 288 participants, (63.4% female; mean age = 32.4 years) including 76 dentists, 63 dental students, and 149 laypeople. Each participant was presented with three clinical scenarios, mild dental misalignment, midline diastema, and moderate anterior crowding, and viewed randomized sets of images depicting pre-treatment, real post-treatment, and AI-generated smiles. For each image, participants indicated whether they believed it was AI-generated or real and rated its aesthetic appeal using a visual analog scale ranging from 0 to 100. Data were analyzed using descriptive statistics and diagnostic performance metrics (accuracy, sensitivity, specificity, PPV, and NPV), with attractiveness ratings compared between AI-generated and real images. Sensitivity for identifying AI-generated images was low across all groups (< 50%), while specificity for recognizing real images was high (> 87%). Dental students achieved the highest overall accuracy (72.7%), followed by laypeople (66.3%) and dentists (62.6%). AI-generated smiles were consistently rated as significantly more attractive than real outcomes by all groups (mean VAS: 78.8 vs. 37.9; p < 0.001). AI-generated smile images were less accurately identified and more aesthetically pleasing than real clinical outcomes compared to real post-treatment outcomes, which were more consistently recognized across all participant groups, regardless of treatment type.
- 21 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessPerception of AI-generated smile versus real orthodontic treatment outcomes among dentists, students, and laypeople
In orthodontics, the increasing use of AI-generated smile images in patient communication raises ethical and practical concerns about user perception and misinterpretation of these visuals. This cross-sectional, non-probabilistic sample study evaluated the ability of dentists, dental students, and laypeople to distinguish between real and AI-generated orthodontic smile images, and their perceived attractiveness. The final sample consisted of 288 participants, (63.4% female; mean age = 32.4 years) including 76 dentists, 63 dental students, and 149 laypeople. Each participant was presented with three clinical scenarios, mild dental misalignment, midline diastema, and moderate anterior crowding, and viewed randomized sets of images depicting pre-treatment, real post-treatment, and AI-generated smiles. For each image, participants indicated whether they believed it was AI-generated or real and rated its aesthetic appeal using a visual analog scale ranging from 0 to 100. Data were analyzed using descriptive statistics and diagnostic performance metrics (accuracy, sensitivity, specificity, PPV, and NPV), with attractiveness ratings compared between AI-generated and real images. Sensitivity for identifying AI-generated images was low across all groups (< 50%), while specificity for recognizing real images was high (> 87%). Dental students achieved the highest overall accuracy (72.7%), followed by laypeople (66.3%) and dentists (62.6%). AI-generated smiles were consistently rated as significantly more attractive than real outcomes by all groups (mean VAS: 78.8 vs. 37.9; p < 0.001). AI-generated smile images were less accurately identified and more aesthetically pleasing than real clinical outcomes compared to real post-treatment outcomes, which were more consistently recognized across all participant groups, regardless of treatment type.
- Open Access
- 1 January 2026
- Published by FapUNIFESP (SciELO) in Dental Press Journal of Orthodontics
- Vol. 31 (2) , e2625314
When does Orthodontics make a difference? Threshold effects in attractiveness perception and hiring intentions
Objective: To evaluate the impact of orthodontic treatment on perceived smile attractiveness and hiring decisions in a Brazilian adult population. Methods: A cross-sectional observational study was conducted with 288 adult participants recruited via an online survey. Three clinical cases were selected, and standardized pre- and post-treatment smile photographs were shown in randomized order. Participants rated attractiveness on a 0-100 Visual Analog Scale (VAS) and indicated hiring intention (“yes,” “no,” or “can’t tell”). Descriptive statistics, Mann-Whitney U tests, chi-square tests, and logistic mixed-effects regression models were used for data analysis. Results: Post-treatment images showed significantly higher attractiveness ratings, compared with pre-treatment images (mean VAS: 54.2 vs. 36.4; p < 0.001). Hiring likelihood also increased after orthodontic treatment (44.0% vs. 27.9%; p < 0.001). Affirmative hiring decisions were consistently associated with higher attractiveness scores. In multivariable analysis, attractiveness was a significant independent predictor of hiring intention, while the interaction between VAS and treatment status was not significant (p = 0.201). Individual-level assessments demonstrated that many participants initially classified as non-hireable were reevaluated positively after treatment. Conclusions: Orthodontic treatment significantly improved smile attractiveness and increased hiring likelihood. Dental alignment alone was sufficient to shift social judgments, although overall facial esthetics also contributed. The identified attractiveness threshold marked the point at which orthodontic improvement began to influence employability perceptions. These findings highlight the psychosocial relevance of orthodontics beyond functional benefits.
- 1 January 2026
- Published by FapUNIFESP (SciELO) in Dental Press Journal of Orthodontics
- Vol. 31 (2) , e2625314
Open AccessWhen does Orthodontics make a difference? Threshold effects in attractiveness perception and hiring intentions
Objective: To evaluate the impact of orthodontic treatment on perceived smile attractiveness and hiring decisions in a Brazilian adult population. Methods: A cross-sectional observational study was conducted with 288 adult participants recruited via an online survey. Three clinical cases were selected, and standardized pre- and post-treatment smile photographs were shown in randomized order. Participants rated attractiveness on a 0-100 Visual Analog Scale (VAS) and indicated hiring intention (“yes,” “no,” or “can’t tell”). Descriptive statistics, Mann-Whitney U tests, chi-square tests, and logistic mixed-effects regression models were used for data analysis. Results: Post-treatment images showed significantly higher attractiveness ratings, compared with pre-treatment images (mean VAS: 54.2 vs. 36.4; p < 0.001). Hiring likelihood also increased after orthodontic treatment (44.0% vs. 27.9%; p < 0.001). Affirmative hiring decisions were consistently associated with higher attractiveness scores. In multivariable analysis, attractiveness was a significant independent predictor of hiring intention, while the interaction between VAS and treatment status was not significant (p = 0.201). Individual-level assessments demonstrated that many participants initially classified as non-hireable were reevaluated positively after treatment. Conclusions: Orthodontic treatment significantly improved smile attractiveness and increased hiring likelihood. Dental alignment alone was sufficient to shift social judgments, although overall facial esthetics also contributed. The identified attractiveness threshold marked the point at which orthodontic improvement began to influence employability perceptions. These findings highlight the psychosocial relevance of orthodontics beyond functional benefits.
- Open Access
- 28 August 2025
- Published by JMIR Publications Inc. in JMIR Medical Informatics
- Vol. 13, e75117
A Machine Learning Algorithm With an Oversampling Technique in Limited Data Scenarios for the Prediction of Present and Future Restorative Treatment Need: Development and Validation Study
Background: Untreated dental caries is the most common health condition worldwide. Therefore, new strategies need to be developed to reduce the manifestations of dental caries. Objective: This study aimed to develop and test a machine learning (ML) algorithm for detecting present and predicting future carious lesions in the adolescent population using a set of easy-to-collect predictive variables. In addition, this study aimed to deal with an imbalanced and small dataset using an oversampling method. Methods: This population-based study was conducted among secondary schoolchildren, aged between 13 and 17 years, from the northern parts of Finland in 2022. After meeting the inclusion criteria, a total of 218 participants were included in this study. The inclusion criteria consisted of participants having completed a web-based risk assessment questionnaire and having undergone a clinical examination at public health care services. Dental caries (International Caries Detection and Assessment System [ICDAS] scores of 4, 5, and 6; ie, ICDAS 4-6) and active initial caries (ICDAS 2+, 3+) were considered as outcomes. Several predictors, such as behavioral and dietary habits, were included. An extreme gradient boosting model was developed, tested, and assessed for its predictive performance. A 4-fold cross-validation was performed using the nested resampling technique. The random oversampling examples method and the k-nearest neighbors classifiers were used for all 4 folds. The mean (SD) performance of all the folds was computed. Results: Dental caries (ICDAS 2+,3+,4-6) were prevalent in 65.6% (143/218) of the participants. The mean area under the curve was 0.77 (SD 0.04) and the mean F1-score was 0.82 (SD 0.06) for the extreme gradient boosting model. Similarly, the mean area under the curve and mean F1-scores after oversampling were 0.74 (SD 0.05) and 0.79 (SD 0.04), respectively. The Shapley additive explanation values were calculated for all 4 folds to assess feature importance, revealing that previous dental fillings were the feature most strongly associated with the need for restorative treatment. Conclusions: On the basis of the performance metrics, the ML algorithm developed and tested in this study can be considered good. The ML algorithm could serve as a cost-effective screening tool for dental professionals to identify the risk of future restorative treatment needs. However, future studies with longitudinal cohorts and longitudinal data, along with external validation for generalizability, are needed to validate our model.
- 28 August 2025
- Published by JMIR Publications Inc. in JMIR Medical Informatics
- Vol. 13, e75117
Open AccessA Machine Learning Algorithm With an Oversampling Technique in Limited Data Scenarios for the Prediction of Present and Future Restorative Treatment Need: Development and Validation Study
Background: Untreated dental caries is the most common health condition worldwide. Therefore, new strategies need to be developed to reduce the manifestations of dental caries. Objective: This study aimed to develop and test a machine learning (ML) algorithm for detecting present and predicting future carious lesions in the adolescent population using a set of easy-to-collect predictive variables. In addition, this study aimed to deal with an imbalanced and small dataset using an oversampling method. Methods: This population-based study was conducted among secondary schoolchildren, aged between 13 and 17 years, from the northern parts of Finland in 2022. After meeting the inclusion criteria, a total of 218 participants were included in this study. The inclusion criteria consisted of participants having completed a web-based risk assessment questionnaire and having undergone a clinical examination at public health care services. Dental caries (International Caries Detection and Assessment System [ICDAS] scores of 4, 5, and 6; ie, ICDAS 4-6) and active initial caries (ICDAS 2+, 3+) were considered as outcomes. Several predictors, such as behavioral and dietary habits, were included. An extreme gradient boosting model was developed, tested, and assessed for its predictive performance. A 4-fold cross-validation was performed using the nested resampling technique. The random oversampling examples method and the k-nearest neighbors classifiers were used for all 4 folds. The mean (SD) performance of all the folds was computed. Results: Dental caries (ICDAS 2+,3+,4-6) were prevalent in 65.6% (143/218) of the participants. The mean area under the curve was 0.77 (SD 0.04) and the mean F1-score was 0.82 (SD 0.06) for the extreme gradient boosting model. Similarly, the mean area under the curve and mean F1-scores after oversampling were 0.74 (SD 0.05) and 0.79 (SD 0.04), respectively. The Shapley additive explanation values were calculated for all 4 folds to assess feature importance, revealing that previous dental fillings were the feature most strongly associated with the need for restorative treatment. Conclusions: On the basis of the performance metrics, the ML algorithm developed and tested in this study can be considered good. The ML algorithm could serve as a cost-effective screening tool for dental professionals to identify the risk of future restorative treatment needs. However, future studies with longitudinal cohorts and longitudinal data, along with external validation for generalizability, are needed to validate our model.
- Open Access
- 4 April 2025
- Published by JMIR Publications Inc. in JMIR Preprints
A Machine Learning Algorithm With an Oversampling Technique in Limited Data Scenarios for the Prediction of Present and Future Restorative Treatment Need: Development and Validation Study (Preprint)
BACKGROUND Untreated dental caries is the most common health condition worldwide. Therefore, new strategies need to be developed to reduce the manifestations of dental caries. OBJECTIVE This study aimed to develop and test a machine learning (ML) algorithm for detecting present and predicting future carious lesions in the adolescent population using a set of easy-to-collect predictive variables. In addition, this study aimed to deal with an imbalanced and small dataset using an oversampling method. METHODS This population-based study was conducted among secondary schoolchildren, aged between 13 and 17 years, from the northern parts of Finland in 2022. After meeting the inclusion criteria, a total of 218 participants were included in this study. The inclusion criteria consisted of participants having completed a web-based risk assessment questionnaire and having undergone a clinical examination at public health care services. Dental caries (International Caries Detection and Assessment System [ICDAS] scores of 4, 5, and 6; ie, ICDAS 4-6) and active initial caries (ICDAS 2+, 3+) were considered as outcomes. Several predictors, such as behavioral and dietary habits, were included. An extreme gradient boosting model was developed, tested, and assessed for its predictive performance. A 4-fold cross-validation was performed using the nested resampling technique. The random oversampling examples method and the k-nearest neighbors classifiers were used for all 4 folds. The mean (SD) performance of all the folds was computed. RESULTS Dental caries (ICDAS 2+,3+,4-6) were prevalent in 65.6% (143/218) of the participants. The mean area under the curve was 0.77 (SD 0.04) and the mean F1-score was 0.82 (SD 0.06) for the extreme gradient boosting model. Similarly, the mean area under the curve and mean F1-scores after oversampling were 0.74 (SD 0.05) and 0.79 (SD 0.04), respectively. The Shapley additive explanation values were calculated for all 4 folds to assess feature importance, revealing that previous dental fillings were the feature most strongly associated with the need for restorative treatment. CONCLUSIONS On the basis of the performance metrics, the ML algorithm developed and tested in this study can be considered good. The ML algorithm could serve as a cost-effective screening tool for dental professionals to identify the risk of future restorative treatment needs. However, future studies with longitudinal cohorts and longitudinal data, along with external validation for generalizability, are needed to validate our model.
- 4 April 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessA Machine Learning Algorithm With an Oversampling Technique in Limited Data Scenarios for the Prediction of Present and Future Restorative Treatment Need: Development and Validation Study (Preprint)
BACKGROUND Untreated dental caries is the most common health condition worldwide. Therefore, new strategies need to be developed to reduce the manifestations of dental caries. OBJECTIVE This study aimed to develop and test a machine learning (ML) algorithm for detecting present and predicting future carious lesions in the adolescent population using a set of easy-to-collect predictive variables. In addition, this study aimed to deal with an imbalanced and small dataset using an oversampling method. METHODS This population-based study was conducted among secondary schoolchildren, aged between 13 and 17 years, from the northern parts of Finland in 2022. After meeting the inclusion criteria, a total of 218 participants were included in this study. The inclusion criteria consisted of participants having completed a web-based risk assessment questionnaire and having undergone a clinical examination at public health care services. Dental caries (International Caries Detection and Assessment System [ICDAS] scores of 4, 5, and 6; ie, ICDAS 4-6) and active initial caries (ICDAS 2+, 3+) were considered as outcomes. Several predictors, such as behavioral and dietary habits, were included. An extreme gradient boosting model was developed, tested, and assessed for its predictive performance. A 4-fold cross-validation was performed using the nested resampling technique. The random oversampling examples method and the k-nearest neighbors classifiers were used for all 4 folds. The mean (SD) performance of all the folds was computed. RESULTS Dental caries (ICDAS 2+,3+,4-6) were prevalent in 65.6% (143/218) of the participants. The mean area under the curve was 0.77 (SD 0.04) and the mean F1-score was 0.82 (SD 0.06) for the extreme gradient boosting model. Similarly, the mean area under the curve and mean F1-scores after oversampling were 0.74 (SD 0.05) and 0.79 (SD 0.04), respectively. The Shapley additive explanation values were calculated for all 4 folds to assess feature importance, revealing that previous dental fillings were the feature most strongly associated with the need for restorative treatment. CONCLUSIONS On the basis of the performance metrics, the ML algorithm developed and tested in this study can be considered good. The ML algorithm could serve as a cost-effective screening tool for dental professionals to identify the risk of future restorative treatment needs. However, future studies with longitudinal cohorts and longitudinal data, along with external validation for generalizability, are needed to validate our model.