Gerald Torgersen

• Torgersen, Gerald; Winsnes, Karianne Ervik; Handal, Trude; Diep, My Tien; Ørstavik, Dag Solmund & Sunde, Pia Titterud (2026). Open‐Weight Deep Learning for Periapical Index Scoring: Training and Comparison of Convolutional Neural Network Models Based on Three Distinct Architectures and an Ensemble Approach. International Endodontic Journal. ISSN 0143-2885. doi: 10.1111/iej.70268. Full text in Research Archive Show summary

  Aim This study aimed to (i) train and compare three convolutional neural network (CNN) architectures for Periapical Index (PAI) scoring on intraoral radiographs, and (ii) assess whether an ensemble of these models improves performance compared with any single architecture. Methods A dataset of 17,549 apex‐centred image clips extracted from intraoral radiographs was annotated by a calibrated observer (KW) with PAI scores (1–5) and used to train three CNN architectures (ResNet50, EfficientNet‐B3 and ConvNeXt‐Tiny) for PAI scoring. The image clips were split 80/20 into training and validation sets, and model selection was based on validation performance of quadratic weighted kappa (QWK). The three CNN models were combined into an equal‐weight soft‐voting ensemble model, and all four models were evaluated on an independent balanced test set of 200 images (40 per PAI score). QWK was used as the primary outcome metric, and bootstrapping was used to estimate the confidence intervals (CIs). The PAI scale was dichotomized into healthy (PAI 1–2) and diseased (PAI 3–5), and accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and F1 score were calculated. Results The ensemble model achieved the highest performance with a QWK of 0.79 on the test set, followed by EfficientNet‐B3 (QWK = 0.77), ResNet50 (QWK = 0.76) and ConvNeXt‐Tiny (QWK = 0.70), with overlapping confidence intervals (CIs). For the binary PAI classification, the performance of the three CNN models and the ensemble model was comparable. The ensemble model achieved the following results: accuracy 86%, sensitivity 83%, specificity 90%, PPV 93%, NPV 78% and F1 score 0.88. Conclusion Optimization and evaluation of three CNN models and an ensemble model for PAI scoring showed comparable performance on both the five‐class PAI scale and the binary PAI scale on an independent test set. The ensemble model did not show statistically significant improvement over the three individual CNN architectures. The AI models trained and tested in this study are made open‐source under the open‐source MIT licence.

• Katle, Erik; Zandi, Homan; Pedersen, Deniz; Sunde, Pia Titterud; Torgersen, Gerald R. & Ørstavik, Dag Solmund (2024). Radiographic outcome of endodontic treatment and retreatment of teeth with apical periodontitis using two different root canal irrigants. A prospective cohort study. International Endodontic Journal. ISSN 0143-2885. 57(3), p. 297–304. doi: 10.1111/iej.14019. Full text in Research Archive Show summary

  Aim The aim of this prospective cohort study was to compare the radiographic outcome of endodontic treatment and retreatment of teeth with apical periodontitis using either 1% sodium hypochlorite (NaOCl) or 2% chlorhexidine digluconate (CHX) for root canal irrigation. Materials and Methods In the years from 2013 to 2015 standard irrigation varied by semester between NaOCl and CHX at the Department of Endodontics at the Faculty of Dentistry. During that time, 912 teeth received treatment for apical periodontitis in 744 patients, of whom 532 responded to the request for a 1-year follow-up. Only one tooth per person (the most distally located) were included; 285 teeth treated with NaOCl and 247 with CHX. One hundred cases were then randomly sampled from each irrigation group and analysed for outcome by periapical index (PAI) scoring using criteria for success, uncertain and failure. Clinical and other radiographic parameters were scored or recorded and analysed for associations with radiographic outcome using chi-square, ANOVA and regression analyses. Results Success rates (PAI score 1 or 2 at control) were nearly identical for the two irrigation liquids. The use of irrigating solution also did not significantly influence the outcome in chi-square analyses of subgroups of teeth or regression analyses with other variables included. Ordinal regression analysis established that preoperative lesion size or preoperative PAI score were significantly associated with outcome, and teeth in the mandible had significantly better outcomes than in the maxilla. Conclusions No significant differences in the radiographic outcome using either 1% NaOCl or 2% CHX as irrigants were found. The outcome was better for teeth with small lesions or lower PAI scores at completion of treatment and for mandibular teeth.

• Kubberød, Jon Olav; Torgersen, Gerald R.; Gjermo, Per E; Baelum, Vibeke & Preus, Hans Ragnar (2023). Five-year radiological findings from a randomized controlled trial of four periodontitis treatment strategies. European Journal of Oral Sciences. ISSN 0909-8836. doi: 10.1111/eos.12949. Full text in Research Archive Show summary

  Radiographic findings from long-term studies of periodontitis treatment have rarely been reported. Although bone destruction is a prominent feature of periodontitis, the long-term effect on alveolar bone levels of different treatment strategies, with or without adjunctive metronidazole (MTZ), has not been reported. We investigated the 5-year radiographic outcome of therapy in patient groups treated with conventional scaling and root planing (SRP) or same-day full-mouth disinfection (FDIS), with or without adjunctive MTZ. Following a 3-month oral hygiene phase, 184 periodontitis patients were randomly allocated to one of four treatment regimens: (i) FDIS+MTZ; (ii) FDIS+placebo; (iii) SRP+MTZ; or (iv) SRP+placebo. Following active treatment, patients received biannual maintenance. In total, 161 patients (87.5%) completed the 5-year follow-up examination, at which the radiographic bone level (RBL), clinical attachment level, probing pocket depth, presence of plaque, and bleeding were recorded again. At the 5-year follow up examination, minor radiological bone loss was observed in the intervention groups FDIS+placebo, SRP+MTZ, and SRP+placebo; by contrast, the FDIS+MTZ group did not show any change in RBL. Full-mouth disinfection did not generally perform better than conventional SRP performed over a period of 2 to 4 weeks.

• Jäghagen, Eva Levring; Lange, Thomas de; Torgersen, Gerald; Møystad, Anne & Ahlqvist, Jan (2023). The use of post-exam feedback in oral radiology: A survey study of dental and dental hygienist schools in Europe. European journal of dental education. ISSN 1396-5883. p. 1048–1052. doi: 10.1111/eje.12897. Full text in Research Archive Show summary

  Introduction: A combination of summative assessments and post-exam feedback can create a formative environment to support student learning. It remains unknown to what extent feedback sessions are conducted after exams in oral radiology courses for dental and dental hygienist students in Europe. This survey study aimed to explore whether post-exam feedback sessions were conducted, and if so, in what format. Materials and Methods: A questionnaire was distributed to 76 dental schools in 26 countries and 34 dental hygienist schools in 14 countries in Europe. The questions addressed whether feedback was provided after exams in oral radiology, the feedback format, and whether feedback sessions were formalised or performed for other reasons. Results: Questionnaire responses were received from 33 dental schools in 18 countries, and 20 dental hygienist schools in 11 countries. Information about grades was provided in 23–30% of the schools. Post-exam feedback sessions including opportunities for discussion was provided at 39% of the schools, all with guidelines stating that these sessions should be carried out. The remaining schools provided feedback either on request by individual students or as written examples of correct answers. Conclusion: Almost one third of the dental schools announced grades without providing any post-exam feedback that could support learning. Two fifth utilised feedback sessions to assist students in gaining knowledge and sorting out misconceptions, all according to guidelines. Given the possible learning potential in post-exam feedback with opportunity to discuss, further studies of this format would be valuable, including the role and use of feedback guidelines.

• Lange, Thomas de; Møystad, Anne; Torgersen, Gerald R.; Ahlqvist, Jan & Jäghagen, Eva Levring (2023). Students' perceptions of post-exam feedback in oral radiology—A comparative study from two dental hygienist educational settings. European journal of dental education. ISSN 1396-5883. 28(2), p. 1–11. doi: 10.1111/eje.12959. Full text in Research Archive Show summary

  Introduction: The aim of this study was to investigate how students perceive the benefit of participating in a teacher-organised session providing feedback on exams, termed post-exam feedback, in two dental hygienist programmes. Methods: The study was based on interviews with 22 participants, including 18 students and 4 faculty teachers. The data were approached on the basis of thematic analysis, allowing us to generate insights on how the participants reflected on their participation in the post-exam feedback sessions and how they perceived this arrangement as learners. Results: The findings from the study suggest that motivated students consider post-exam feedback to be beneficial in clearing up uncertainties and deepening their understanding of issues not fully understood during the exam, as well as supporting their further learning. Less motivated students mainly consider post-exam feedback to be relevant for students who do not pass the exams. Conclusions: Organised in a student-centred way and with attentiveness to student learning preferences, the results suggest that post-exam feedback can be valuable for enhancing assessment and supporting student learning related to exams.

• Young, Alix; Skudutyte-Rysstad, Rasa; Torgersen, Gerald R. & Giertsen, Elin (2022). Teaching radiographic caries detection and treatment planning: A seminar using an audience response system. Caries Research. ISSN 0008-6568. 56(3), p. 226–233. doi: 10.1159/000526109. Full text in Research Archive Show summary

  This study presents a seminar model for teaching radiographic caries detection and treatment planning at the Faculty of Dentistry, University of Oslo. The seminar is based partly on an audience response system (ARS) and uses patient cases to focus on caries risk assessment and treatment planning. This paper describes the seminar design, implementation, learning outcomes, and observational study of variability in caries registrations and students’ attitudes to use of ARS. Dental and dental hygiene students participate in two seminar modules. Module 1 aims to develop and increase individual student skills in radiographic caries lesion detection, scoring, and differential diagnosis. Students perform trial registrations on bitewings using an ARS with anonymous live polling, and scorings are discussed in plenum. Students then perform individual registrations on 12 bitewing pairs. Using digital scoring, students detect and grade caries lesions on all approximal and occlusal tooth surfaces. After the session, students use the ARS to repeat scorings on selected tooth surfaces, and results are again discussed in plenum. Module 2 involves group exercises on 4 patient cases that are later presented with plenary discussions. In total, 1,624 caries registrations performed by 150 students attending the seminar between 2016 and 2018 were assessed for variability between students. As expected, variations in caries registrations were observed between students, mostly related to restored surfaces or tooth surfaces that were otherwise difficult to register. In 2022, 63 dental and dental hygiene students attending the seminar answered a questionnaire about use of ARS. The responses were scored using a five-point Likert scale. Overall, no significant difference in satisfaction with the ARS-based module was observed between dental and dental hygiene students (χ<sup>2</sup> test, <i>p</i> > 0.05). The majority of the students were positive toward the use of ARS (94%), but some disagreed on the role of ARS in usefulness for understanding the seminar content (3.2%), and in increasing their confidence in radiographic registration of caries (3.2%). The ARS-based module provides a positive learning environment that ensures student anonymity, interactivity, and engagement, and combined with the other seminar module gives students basic skills in caries detection and treatment planning.

• Jordal, Kristin; Skudutyte-Rysstad, Rasa; Sen, Abhijit; Torgersen, Gerald; Ørstavik, Dag & Sunde, Pia Titterud (2021). Effects of an individualized training course on technical quality and periapical status of teeth treated endodontically by dentists in the Public Dental Service in Norway. An observational intervention study. International Endodontic Journal. ISSN 0143-2885. doi: 10.1111/iej.13669. Full text in Research Archive Show summary

  Aim: To investigate the effect of a continuing education course on technical quality and treatment outcome for root filled permanent teeth in Møre and Romsdal County, Norway. Methodology: Fifty-two dentists employed in the Public Dental Service in Møre and Romsdal county, Norway, completed a two-day continuing education course in root canal treatment. Periapical radiographs of root filled teeth treated before and after the course, with at least one-year follow-up were identified and scored for technical quality and periapical status. Technical root filling quality was assessed by density and length and the treatment outcome by PAI scores. Treatment information was achieved from the county's electronic dental record system. Information regarding the dental practitioners' background and treatment procedure routines was collected by a questionnaire. Descriptive statistics analyses and mixed-effect logistic regression analyses were performed to evaluate the effect of the course. Results: Radiographs were available for 224 teeth root filled before and for 221 teeth after the course. The proportion of teeth with adequate root filling quality was significantly lower after the course (p =0.006), associated mainly with short root fillings (p 18 years. Conclusions: A two-day continuing education course in root canal treatment attended by Public Dental Service dentists in Norway did not improve the technical quality of root fillings or periapical status associated with root filled teeth. Keywords: apical periodontitis; continuing education; endodontics; general dental practitioner; reciprocating technique; root filling quality.

• Knivsberg, Ina Charlotte; Kopperud, Simen E.; Bjørk, Mai-Britt; Torgersen, Gerald; Skramstad, Katarzyna Ewa Kocent & Kvaal, Sigrid I. (2021). Digitalised exercise material in forensic odontology. International journal of legal medicine (Print). ISSN 0937-9827. doi: 10.1007/s00414-021-02740-7. Full text in Research Archive Show summary

  Abstract Introduction This paper presents digital educational material in forensic odontology, including dental identification after multiple fatalities and dental age estimation from different age groups. Material and method Electronic patient records consisting of intraoral scans of the dentition, digital radiographs, photographs and written dental records were collected. Exercises in age estimations contained digital radiographs and photographs of ground tooth sections, with digital measuring tools and tables according to age groups. The teaching material was organised as a module in an electronic Learning Management System with external links to all relevant teaching material. Results For the identification exercises, intraoral scans and the latest digital radiographs simulated the postmortem examination of the deceased. For comparison, all other radiographs, photographs and dental records were available as antemortem material. The exercise was to match postmortem findings with the antemortem records using the Interpol standard and reconciliation. Age assessment of children used designated tables to grade tooth development on digital radiographs. For adults, non-destructive methods, digital radiographs, photographs and measuring tools were used. Discussion The teaching concept was hybrid, but it can easily be adapted as a fully digital exercise. The instructions and written material can be translated into different languages. The level of difficulty in the exercises can be adjusted according to the participant’s level of knowledge. Conclusion The educational material embraces the new possibilities for digitalisation and intraoral scanning. This might be a valuable tool for motivating and engaging the students in their participation and understanding of the subject.

• Lange, Thomas de; Møystad, Anne & Torgersen, Gerald R. (2020). How can video‐based assignments integrate practical and conceptual knowledge in summative assessment? Student experiences from a longitudinal experiment. British Educational Research Journal (BERJ). ISSN 0141-1926. doi: 10.1002/berj.3632. Full text in Research Archive Show summary

  This article focuses on how video assignments presenting clinical situations can be implemented in digital summative assessment to enhance the integration of practical and conceptual knowledge. The underlying perspective is that sustainable assessment should comprise evaluative practices that equip students for the challenges they will face in their future professional lives. Drawing on the literature and on empirical results of a qualitative longitudinal research study at a Norwegian university, the integrative potential and challenges of video assignments are examined. The study confirms that video‐based assignments are well suited for assessing integrated competencies. However, the complexities of videos’ visual displays have clear consequences that indicate how these resources should be produced. To optimise the potential of video use in enhancing sustainable summative assessment, we must strike a balance between technical accommodation, task clarity and integrative potential.

• Ketharanthan, Vimalan; Torgersen, Gerald R.; Petrovski, Beata & Preus, Hans Ragnar (2019). Radiographic alveolar bone level and levels of serum 25-OH-Vitamin D3 in ethnic Norwegian and Tamil periodontitis patients and their periodontally healthy controls. BMC Oral Health. 19(1). doi: 10.1186/s12903-019-0769-6. Full text in Research Archive Show summary

  Background Studies suggest association between low serum 25-OH-Vitamin D3 (VitD) and chronic destructive periodontal diseases. The main sources of VitD is sun exposure and fat fish. Subjects with dark skin will therefore generate less VitD as response to sun exposure. The aim of the study was to assess the radiographic bone level and levels of serum VitD in ethnic Norwegian and Tamil periodontitis patients and their respective healthy controls. Methods Twenty-seven Tamil periodontitis patients living in Norway were compared to 21 Tamil controls as well as to 21 Norwegian periodontitis patients and 23 Norwegian controls. Marginal bone level was diagnosed on radiographs. VitD levels were diagnosed in blood samples by high-performance liquid chromatography-mass spectrometry. Results VitD levels were lower in Norwegian periodontitis patients than in controls, while no significant differences were observed between Tamil periodontitis patients and controls despite the significant difference between RBL between the periodontitis patients and controls in both groups. When calculating the odds ratio for having periodontal disease in both populations together, it appeared that one unit increased serum VitD (i.e. 1 nmol/L) decreased the odds of having radiographic bone loss by 4%. Conclusion According to logistic regression, and after correcting for confounding factors, VitD levels showed significant association with the presence of periodontitis, as expressed by radiographic bone loss, in all patients combined.

• Lange, Thomas de; Møystad, Anne & Torgersen, Gerald R. (2018). Increasing clinical relevance in oral radiology: Benefits and challenges when implementing digital assessment. European journal of dental education. ISSN 1396-5883. 22(3), p. 198–208. doi: 10.1111/eje.12326. Full text in Research Archive Show summary

  Abstract Aims The aims of the study were to investigate benefits and challenges in implementing a digital examination and study the clinical relevance of the digital examination in relation to clinical training and practice. Material and method The study was based on semi‐structured focus‐group interviews from two distinct student populations (2016 and 2017) in a bachelor programme in dental hygiene. In addition, conversational data from a plenary discussion from the whole second student population (2017) were collected and analysed. The data were approached on basis of content analysis. Results A benefit experienced in the digital examination was the ease in typing and editing answers on the computer. This suggests an increased effectiveness in computer‐based compared to analogue examinations. An additional advantage was the experienced relevance of the examination related to the clinic. This finding refers not only to the digital presentations of images, but also to the entire setting in the clinic and dental practice. The limitations reported by the students were non‐optimal viewing conditions for presenting radiographic images and difficulties in obtaining an overview of the assignments compared to paper‐based examinations due to the linear digital examination format. The last finding on lacking overview revealed an influence on student performances which should be taken seriously in designing digital examinations. Conclusion In conclusion, the digital layout increases efficiency and clinical relevance of examinations to a certain extent. Obstacles were found in limitations related to image presentation and lack of overview of the examination. The latter challenge raises questions related to developing suitable assessment software.

• Elkhateeb, Sara M.; Torgersen, Gerald R. & Arnout, Eman A. (2016). Image quality assessment of clinically-applied CBCT protocols using a QAT phantom. Dentomaxillofacial Radiology. ISSN 0250-832X. 45(5). doi: 10.1259/dmfr.20160075. Full text in Research Archive Show summary

  Objectives: This study has two main purposes: (1) To evaluate the effect of various clinically-applied protocols (changing FOV, voxel size and tube voltage) on CS 9300 CBCT image quality using its QAT phantom; and (2) to evaluate the efficacy of a newly-designed prototype plugin on ImageJ which was introduced to measure MTF using a modified slanted-edge method. Methods: Using a specifically-constructed QAT phantom, five-image quality parameters were assessed (image uniformity, signal to noise ratio [SNR], contrast, position checking, and the modulation transfer function [MTF] for ten variables, clinically-applied protocols on a Care- Stream (CS) 9300 PREMIUM 3D CBCT device. Results: With FOV and voxel size fixed, high-dose protocols resulted in less image noise and less uniformity. On the other hand, enhanced uniformity was achieved with less kVp. Moreover, high-dose protocols resulted in a high contrast-to-noise ratio. There were significant differences in CNR between 60 and 70 kVp as well as between 70 and 90 kVp. Our study identified no difference between MTF and either voxel size or FOV. Conclusions: The image quality of the CS 9300 CBCT varied with variability in exposure protocols and different voxel sizes. The negative correlation between greater spatial resolution (smaller voxel size) and the MTF shows that the slanted edge method not ideal for MTF evaluation of CBCT, and, therefore, do not recommend using it for this purpose.

• Hol, Caroline; Hellén-Halme, Kristina; Torgersen, Gerald R.; Nilsson, Mats & Møystad, Anne (2015). How do dentists use CBCT in dental clinics? A Norwegian nationwide survey. Acta Odontologica Scandinavica. ISSN 0001-6357. 73(3), p. 195–201. doi: 10.3109/00016357.2014.979866. Full text in Research Archive
• Strindberg, Jerker Edèn; Hol, Caroline; Torgersen, Gerald R.; Møystad, Anne; Nilsson, Mats & Näsström, Karin [Show all 7 contributors for this article] (2015). Comparison of Swedish and Norwegian Use of Cone-Beam Computed Tomography: a Questionnaire Study. eJournal of Oral And Maxillofacial Research - JOMR. 6(4), p. e 2–e 2. doi: 10.5037/jomr.2015.6402. Full text in Research Archive
• Preus, Hans Ragnar; Torgersen, Gerald R.; Koldsland, Odd Carsten; Hansen, Bjørn Frode; Aass, Anne Merete & Larheim, Tore Arne [Show all 7 contributors for this article] (2015). A new digital tool for radiographic bone level measurements in longitudinal studies. BMC Oral Health. 15(1), p. 1–7. doi: 10.1186/s12903-015-0092-9. Full text in Research Archive Show summary

  Background The reproducibility of measurements on radiographs is influenced by the techniques by which the images as well as the measurements are obtained. Thus, bias resulting from errors in the image and/or image examinations at two points in time may result in wrongful registrations of true biological or pathological changes. The aim of the present study was to propose and evaluate an indirect radiological examination technique, by which bias, when measuring radiographic bone level, could be substantially reduced as compared to the technique using direct mm measurements. Methods A plugin to ImageJ was designed to reduce bias when measuring bone loss on radiographic images. In human dry mandibles, radiographic images of 20 teeth were obtained parallel with the tooth axis (alpha = 0) and at an angle of 30° deviation. The direct technique of measuring radiographic bone level (RBL) and the indirect, length-adjusted RBL were registered by four researchers in a double blinded fashion. Results When mean RBL measured at 0° angle was 7.0 mm, the corresponding mean RBL measured at 30° angle was 7.8 mm, signifying an 11.4 % increase (p = 0.032), whereas the mean length-adjusted RBL increased by 0.6 % (p = 0.9). Conclusions This study showed that the use of the original, direct technique (ImageJ) resulted in markedly biased radiographic bone level at 30° angle, while the proposed indirect length-adjusted technique (ImageJ plugin) did not.

• Torgersen, Gerald R.; Hol, Caroline; Møystad, Anne; Hellén-Halme, Kristina & Nilsson, Mats (2014). A phantom for simplified image quality control of dental cone beam computed tomography units. Oral surgery, oral medicine, oral pathology and oral radiology. ISSN 2212-4403. 118(5), p. 603–611. doi: 10.1016/j.oooo.2014.08.003. Full text in Research Archive

View all works in NVA

• Torgersen, Gerald R.; Winsnes, Karianne Ervik; Handal, Trude & Sunde, Pia Titterud (2025). Visualizing AI Decisions in Automated Periapical Index Assessment. Full text in Research Archive Show summary

  Objective: Explainable artificial intelligence (XAI) is critical for clinical acceptance of AI models in dental radiology. This study explores the use of XAI—primarily Grad-CAM and LayerCAM—to interpret deep learning-based Periapical Index (PAI) scoring from intraoral radiographs, with a focus on exploring attentions patterns, meaning the areas contributing to the models decision of PAI score. Methods: We trained an open-source deep learning model for PAI classification on 300×300 pixel radiographic clips, using expert-labeled data. Grad-CAM and LayerCAM were applied to visualize model attention, generating class-specific average heatmaps and analyzing individual cases for anatomical plausibility and potential artifacts. Additional XAI methods were explored for comparison. Code and weights will be published together with the main publication. Results: XAI visualizations revealed that model attention patterns systematically varied with PAI class: lower scores showed focused periapical activation, while higher scores exhibited more diffuse and multi-focal attention. Individual heatmaps highlighted both expected anatomical focus and occasional peripheral activations, suggesting areas for further model refinement. LayerCAM and Grad-CAM provided complementary insights into model interpretability. Conclusions: Grad-CAM and related XAI methods offer valuable insights into the decision-making of AI models for PAI scoring. Ongoing work aims to refine XAI techniques and analyze individual attention maps to improve both model performance and interpretability.

• Engebråten, Christian Andersson; Torgersen, Gerald R. & Arvidsson, Linda Zamoline (2025). AI-Derived Metrics for Assessing Panoramic Radiograph Quality. Full text in Research Archive Show summary

  Purpose Panoramic radiographs are widely used in dental diagnostics, yet image quality remains highly variable due to frequent patient positioning errors, impacting diagnostic reliability and workflow efficiency. While artificial intelligence (AI) has shown promise in automating quality assessment, current evaluation methods often rely on subjective or limited landmark-based criteria, which may not fully leverage the capabilities of modern AI systems. This project aims to develop and explore novel, AI-adapted metrics for the objective assessment of panoramic image quality, focusing on patient positioning parameters such as head tilt and rotation. Positioning errors compromise diagnostic accuracy and necessitate image retakes. The resulting framework will serve both clinical practice and educational settings, providing more objective quality assessment tools that may help improve diagnostic reliability and reduce unnecessary radiation exposure. Methods Using deep learning models architectures, we automatically detect and segment relevant anatomical structures to enable geometric and symmetry-based analyses. These AI-derived metrics will be compared with established evaluation methods and validated against assessments done by a human expert to determine clinical relevance. Results We have fine-tuned deep learning models for segmentation of the mandible, maxilla, and teeth, enabling automated anatomical identification and classification. We compared the AI-derived metrics with traditional landmark-based methods to evaluate their agreement and practical application. Conclusions Our preliminary work demonstrates the feasibility of using AI-based segmentation to support objective image quality assessment in panoramic radiography. We discuss the potential of these new geometric metrics and their similarities and differences compared to established evaluation approaches.

• Maroufi, Sada; Milanovic, Blagica; Hove, Lene Hystad; Høyvik, Ann Catrin; Torgersen, Gerald R. & Fjeld, Katrine Gahre (2025). SimSamBot – Development of an AI-Based Training Robot Assisting Students in Communication with Dementia Patients. Full text in Research Archive Show summary

  Background The increasing prevalence of dementia presents unique challenges for dental professionals, particularly in communication and patient management. Effective communication is critical to providing compassionate, patient-centered care, yet dental students often lack the necessary training to interact confidently with dementia patients. Aims and Objectives This project aims to equip dental students with the skills and knowledge required to communicate effectively with dementia patients. Specific objectives include: 1 Developing a communication training module tailored to dementia care; 2 Enhancing students’ ability to recognize and adapt to the needs of patients with cognitive decline; and 3 Promoting empathy and understanding through experiential learning approaches. Materials and Methods A large language model (LLM) acting as communication robot is supplied with data on the patient groups it simulates. Information is collected through literature reviews and audio recordings of clinical consultations with dementia patients, with consent from the healthcare provider, patient and relatives if needed. Recordings are anonymized and transcribed immediately after consultations, with the original recordings subsequently deleted. Only anonymized data is used in the robot. Results While full implementation is forthcoming, anticipated outcomes include improved student confidence in communicating with dementia patients, enhanced empathy, and better preparedness for clinical practice. Discussion and Conclusion This initiative addresses a critical need in dental education by preparing students for the complexities of dementia care. By fostering effective communication skills, this project aligns with ADEEʼs vision of reimagining education to meet evolving societal needs. The outcomes will contribute to improved oral health care for dementia patients while advancing the pedagogical standards of dental education globally.

• Torgersen, Gerald R.; Torp, Monica Jeanette Wicksell & Mavroudi, Anna (2025). Exploring Generative AI in Dental Education: Student Experiences in Radiology-Focused Curriculum. doi: https:/api.ltb.io/show/ACJTX. Full text in Research Archive Show summary

  Background Generative AI tools are increasingly being integrated into education to enhance learning and engagement. This study focuses on the use of an AI assistant in teaching dental hygiene students within a curriculum covering radiation physics, radiation biology, radiation protection, and radiological technology. These topics were chosen based on the teaching expertise of a project team member, ensuring subject matter competency and enabling effective integration of AI to support student learning within a focused and manageable framework. Aims and Objectives The study aims to evaluate how generative AI can support student learning in radiology-related topics for dental hygiene students. Objectives include: - Assessing the impact of AI-assisted learning on student comprehension and engagement. - Collecting qualitative insights into student experiences and perceptions of AI as a learning tool. - Developing pedagogical strategies for integrating AI effectively into radiology education. Materials and methods A mixed-methods approach was employed: - AI Tool Integration: Students used an AI assistant to complete structured tasks related to radiology topics. - Focus Group Interviews: semi-structured interviews were conducted with student participants to collect qualitative data on their experiences. -Iterative Refinement: Feedback from initial activities informed adjustments to both the system prompts and teaching strategies. Results Formal analysis of interview data is ongoing. Preliminary findings, including insights from student and faculty experiences, will be presented and discussed at the ADEE meeting. Discussion and conclusion This study establishes a foundation for exploring how generative AI can be integrated into a focused radiology curriculum for dental hygiene students. While the analysis of interview data is still underway, the findings are anticipated to provide valuable insights into how AI tools influence student engagement, comprehension, and learning processes. These insights will inform future strategies for integrating AI into dental education, with an emphasis on fostering critical thinking and active learning.

• Torgersen, Gerald & Winsnes, Karianne Ervik (2024). Hot Insights: Explainable AI and Heatmaps for Accurate PAI-Scoring in Endodontic Treatments. Full text in Research Archive Show summary

  Introduction The application of Convolutional Neural Networks (CNN) for pathology detection within dental imaging is rapidly expanding(1). Evaluating the outcome of root canal treatments, particularly using the periapical index (PAI)(2), is crucial in endodontics. Traditionally, PAI scoring involves qualitative judgment following a calibration process. However, as Pythagoras famously noted, there is no such thing as quality in the world; everything is based on numbers. Previous studies have demonstrated the potential of CNNs in automating PAI scoring(3,4). Building on this work, our project, in collaboration with endodontic specialist residents, aims to develop a CNN model for accurate PAI scoring on dental radiographs. This year, our focus shifts to enhancing interpretability by visualizing regions of interest using a Gradient-weighted Class Activation Map (Grad-CAM) (5). By doing so, we aim to uncover the underlying quantitative factors that influence PAI classification. This method not only enhances model transparency but also aligns with the principles of explainable AI, providing valuable insights for clinical decision-making. Methods A set of 300 x 300 pixel clips, centered on the apex and scored by an experienced general dental practitioner working at the Department of endodontics, were used for transfer learning on an EfficientNet B3 model(6) pretrained on the ImageNet dataset(7). When inferencing test images we used the Grad-Cam framework to visualize impact of image areas on choosing different classes. Results The heatmaps show attention localized to the apex on the inferred class. Conclusion Heatmaps provide insight into the areas the model focuses on when determining the class. Further analysis will be conducted to utilize this information to enhance the machine learning approach for PAI scoring. We also plan to use the heatmaps as feedback to the human scorers. References 1. Pauwels R. A brief introduction to concepts and applications of artificial intelligence in dental imaging. Oral Radiol. 2021 Jan;37(1):153–60. 2. Ørstavik D, Kerekes K, Eriksen HM. The periapical index: A scoring system for radiographic assessment of apical periodontitis. Dent Traumatol. 1986 Feb;2(1):20–34. 3. Moidu NP, Sharma S, Chawla A, Kumar V, Logani A. Deep learning for categorization of endodontic lesion based on radiographic periapical index scoring system. Clin Oral Investig. 2022 Jan;26(1):651–8. 4. Pauwels R, Brasil DM, Yamasaki MC, Jacobs R, Bosmans H, Freitas DQ, et al. Artificial intelligence for detection of periapical lesions on intraoral radiographs: Comparison between convolutional neural networks and human observers. Oral Surg Oral Med Oral Pathol Oral Radiol. 2021 May;131(5):610–6. 5. Selvaraju RR, Cogswell M, Das A, Vedantam R, Parikh D, Batra D. Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization. Int J Comput Vis. 2019 Oct;128(2):336–59. 6. Tan M, Le QV. EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks. 2019 [cited 2024 Aug 1]; Available from: https://arxiv.org/abs/1905.11946 7. Yang K, Qinami K, Fei-Fei L, Deng J, Russakovsky O. Towards Fairer Datasets: Filtering and Balancing the Distribution of the People Subtree in the ImageNet Hierarchy. In: Conference on Fairness, Accountability, and Transparency. 2020.

• Hol, Caroline; Skoglund, Maria Karoline; Torgersen, Gerald; Møystad, Anne & Hellén-Halme, Kristina (2024). CBCT – klinikkens beste venn? Si meg – hvordan bruker du den? Den norske tannlegeforenings Tidende. ISSN 0029-2303. 134(8), p. 669–669. doi: 10.56373/2024-8-33. Full text in Research Archive
• Torgersen, Gerald (2024). Fra Kunstig Intelligens til ekte strålevernkompetanse: AI som Veileder i strålevern og strålefysikk. Full text in Research Archive Show summary

  Mål: I dette pilotprosjektet utforsker vi anvendelsen av en chatbot basert på avanserte språkmodeller (Large Language Model, LLM) som en pedagogisk assistent for tannhelsestudenter. Disse studentene står overfor utfordrende temaer som stråling, strålevern og radiologisk teknologi, som ofte oppleves som teoretisk komplekse. Med tanke på at eksamensresultatene ikke speiler det ønskede læringsnivået, sikter vi mot å implementere chatboten som en konstant tilgjengelig personlig læringsassistent (24/7/365) for å forbedre forståelsen og læringsutbyttet. Forskning har vist at personlig veiledning, som en-til-en undervisning, kan ha en betydelig positiv effekt på læring. Materiale og metode: Vi benytter oss av en rekke metoder, inkludert anvendelse av kommersielle LLMs (med databehandleravtaler som tilfredsstiller GDPR) tilpasset med spesifikke instruksjoner og matet med fagrelatert innhold samt studentenes forespørsler. På lengre sikt planlegger vi også å finjustere (fine-tune) åpen kildekode LLMs med fagrelevant materiale. Vi har utviklet en spørsmål-og-svar-database ved å bruke dokumenter fra Direktoratet for strålevern og atomtrygghet (DSA), myndighetenes retningslinjer, våre egne undervisningstekster, samt andre åpne kilder på norsk innen dette feltet. Kommersielt tilgjengelige LLMs (som GPT-4, Claude 2 og Bard) har blitt brukt for å generere eksempler til databasen. Gjennomføring: Vi testet bruken av LLM som læringsassistent på en studentgruppe som arbeidet med spesifikke oppgaver. Deres resultater og opplevelser vil bli analysert for å evaluere effektiviteten Presentasjon: Jeg vil gi en kort, men grundig innføring i språkmodeller, hvor jeg utforsker deres funksjonalitet og ulike metoder for å finjustere dem for spesifikke bruksområder. Videre vil jeg diskutere prosessen rundt produksjonen og anvendelsen av treningsdata for disse modellene. Avslutningsvis presenterer jeg de utfordringene jeg har møtt, mine erfaringer så langt, og de oppnådde resultatene.

• Torgersen, Gerald R. & Winsnes, Karianne Ervik (2024). PAI meets AI – Can Artificial Intelligence Perform Classification of Endodontic Treatment Outcomes? Full text in Research Archive Show summary

  Aim: To identify the most effective convolutional neural network (CNN) model for fine-tuning on a dataset of intraoral radiographs and to evaluate the accuracy of the fine-tuned model in classifying the periapical index (PAI), thereby determining the model‘s capability in accurately reflecting endodontic treatment outcomes. Material and Methods: The model employed anonymized intraoral radiographs with landmarks and PAI scores from endodontic treatments performed at the University of Oslo. Scores were done by master students, residents in endodontics and PHD-candidates. A plugin to ImageJ was used to manually mark landmarks and set the PAI as an indicator for treatment outcome. The PyTorch library in the Python programming language was used to fine-tune and test several pretrained opensource CNN models. Several models including RESNET and VGG19 were fine-tuned on our dataset and tested for accuracy. Results: A fine-tuned VGG19 based model performed best, achieving an accuracy rate of 64%. Conclusion: While applying machine learning to image analysis in standard datasets, such as handwritten digits (MNIST) or flower classification, appears straightforward, the development of a model tailored to endodontic treatment evaluation poses significant challenges. Further results from refinement of training data and experimenting with different CNN models will be presented.

• Torgersen, Gerald R. (2024). Teach me some Radiation Physics - Large Language Models as Teaching Assistants in Oral Radiology. Full text in Research Archive Show summary

  Aim: To develop Large language models (LLM) that can serve as a personal teaching assistant fluent in the Norwegian language within the fields of radiation physics, biology, protection, and radiological technology. The model would give up-to-date knowledge on local radiation protection regulations and relevant clinical and technological aspects while enhancing students’ comprehension and engagement with the material. Materials and methods: We created a database of more than 6000 question/answer pairs relevant to the curriculum for dentist and dental hygienist students in Norway. Specially instructed LLMs were used for creating the data. We fine-tuned the Meta Llama 2 model, which was one of the most advanced open-source models with acceptable knowledge of the Norwegian language on this data. We also used the database as source for a retrieval-augmented generation (RAG) approach to align foundation models with the learning outcomes. Results: None of the approaches has yet resulted in an acceptable outcome. Different approaches, challenges and shortcomings will be presented and discussed. Conclusion: Further research on how to align models with the pedagogical and professional requirements is needed. The project is ongoing and efforts on further enhancement of fine-tuning and improvement of the RAG approach to align the models will be reported.

• Torgersen, Gerald (2023). Happy Teacher, Happier Students - Boosting Learning Outcomes in Theoretical Subjects with a Dash of Entertainment and a Pinch of AI Magic. Full text in Research Archive Show summary

  Teaching radiation physics, radiation protection, and radiological technology can be challenging, as students may not always share the same passion for these subjects as their educators. This workshop aims to share and explore methods that make these subjects more engaging and accessible for students, with a particular focus on orthodontic applications. Creativity, technology, and the incorporation of AI tools will be central to our approach.

• Torgersen, Gerald; Mavroudi, Anna & Torp, Monica Jeanette Wicksell (2023). KI-assistert Læring: Språkmodeller som undervisningsassistent. Full text in Research Archive Show summary

  Tannleger og tannpleiere trenger kompetanse i stråling, strålevern og radiologisk teknologi i sin yrkesutøvelse. Derfor er det en del av pensum i deres studium i oral radiologi. Mange studenter opplever dette emnet som krevende. For å støtte læringsprosessen, utforsker vi bruk av en KI-verktøy, drevet av avanserte språkmodeller, for å forstå disse komplekse temaene. Språkmodeller kan noen ganger generere misvisende informasjon, en utfordring kjent som "hallusinering". Dessuten mangler de ofte spesialisert fagkunnskap. For å møte disse utfordringene, har vi utviklet et norskspråklig datasett spesifikt for vårt fagfelt. Dette datasettet hjelper oss å skreddersy språkmodellene til våre behov. Vi bruker en tilpasset versjon av OpenAI GPT-4, driftet av UiO og kompatibel med GDPR, som integrerer vårt datasett for å gi mer presise svar på studentenes spørsmål. Dette sikrer at studentene får bedre veiledning enn hva standard språkmodeller kan tilby. Vi skal prøve dette verktøyet med første års bachelorstudenter i tannpleie, hvor de vil benytte det i arbeidet med korte essayoppgaver. Verktøyet vil være tilgjengelig for utvalgte oppgaver. Vi vil evaluere studentenes erfaringer med verktøyet gjennom intervjuer og eventuelt spørreskjemaer. Vi vil også analysere oppgavesvarene for å identifisere eventuelle forskjeller som kan tilskrives bruken av KI. På Utdanningskonferansen vil vi ha med oss en laptop der vi kan demonstrere systemet vårt og deltakere kan også selv prøve det. Vi håper å få interessante diskusjoner med kolleger rundt både de tekniske og de pedagogiske sidene av prosjektet.

• Torgersen, Gerald; Javed, Ayesha & Gustavsson, Miriam (2023). Drill, fill, chat …Kunstig intelligens – noen erfaringer fra odontologisk fakultet. Full text in Research Archive Show summary

  Erfaringer fra lærer og to studenter i bruk av ChatGPT til undervisnings- og læringsrelaterte aktiviteter.

• Torgersen, Gerald R. (2022). Når fotonene danser - strålevern i praksis. Full text in Research Archive Show summary

  Hva er egentlig røntgenstråling? • Hvordan påvirker den mennesker? • Og hvordan bruker vi den til å oppnå optimale røntgenbilder med lavest mulig stråledose? Disse spørsmålene krever kompliserte svar og kan ikke dekkes på en halvtime. Dette blir en kort intro- duksjon som forhåpentligvis både er lærerik og litt underholdende. Målet med foredraget er at du tar med deg noen nye impulser som hjelper deg å jobbe sikrere og bedre når du utfører røntgenundersøkelser. MÅLGRUPPE: Tannhelsekretærer

• Torgersen, Gerald; Lange, Thomas de; Ahlqvist, Jan; Jäghagen, Eva Levring & Møystad, Anne (2022). A comparative study of students perceptions of post-exam feedback in oral radiology - A teachers experience. Full text in Research Archive Show summary

  Introduction Post-exam feedback aims to provide students with insight into their performances and enable them to identify knowledge gaps to guide further learning. According to a survey, giving post-exam feedback is mandatory in half of the responding universities in Europe. The aim of this study was to assess dental hygienist students’ perceptions of their learning gains from post-exam sessions in a university with (Umea) and without (Oslo) mandatory post-exam feedback and to compare differences in arrangements with the purpose of improving the quality of post-exam feedback. Materials and Methods The sessions were carried out among dental hygienist students from Umea University, Sweden and from University of Oslo, Norway. The post-exam feedback in Umea was given simultaneously with the exam results, while the session in Oslo was held after the students had received their results. The study was based on a qualitative approach, with two student focus group interviews conducted in each study context (Umea, n=13/Oslo, n=5). All the interviews were led by an unbiased external researcher. The interviews were transcribed verbatim, and a content analysis was performed to identify patterns in participant experiences and perceptions. The teachers subjective experience and thoughts of the activity will be reported. Results (Preliminary) Students from both contexts, with and without previous experience, appreciated the post-exam feedback, citing the learning benefits by clearing up uncertainties and revealing knowledge gaps and misconceptions. Students from Umea reported a somewhat stressful experience having simultaneously received their grades and during the feedback session, a tension that was not reported in Oslo. Most students suggested that the post-exam feedback sessions should always be carried out and last longer. Conclusions Post-exam feedback appears to support student learning. The reported learning outcomes depend on how the session is organized and this requires further investigation.

• Markussen, Hege; Torgersen, Gerald R. & Møystad, Anne (2021). Alternative clinical practice using recorded video demonstrations for oral radiology - a pandemic measure. Full text in Research Archive
• Lange, Thomas de; Ahlqvist, Jan; Jäghagen, Eva Levring; Torgersen, Gerald R. & Møystad, Anne (2021). Students’ perceptions of post-exam feedback in oral radiology—A comparative study from two dental hygienist educational settings. Full text in Research Archive
• Vik, Alix Young; Torgersen, Gerald R.; Giertsen, Elin & Skudutyte-Rysstad, Rasa (2021). Use of a student response system for teaching radiographic caries detection and treatment planning. Full text in Research Archive
• Torgersen, Gerald R. & Hol, Caroline (2018). Tannpleieren – en ekspert i strålevern. Tannstikka : tidsskrift for Norsk tannpleierforening. ISSN 1501-8636. 2018(2), p. 8–14. Full text in Research Archive
• Torgersen, Gerald R.; Lange, Thomas de & Møystad, Anne (2018). Can the use of student created videos enhance learning in oral radiology? Student experiences based on a pilot study. Full text in Research Archive
• Lange, Thomas de; Torgersen, Gerald R. & Møystad, Anne (2017). Digitising Summative assessment in dental education, benefits and challenges. Full text in Research Archive
• Torgersen, Gerald R.; Møystad, Anne & Lange, Thomas de (2017). Increasing clinical relevance in oral radiology through computerIncreasing clinical relevance in oral radio. Full text in Research Archive
• Arvidsson, Linda Zamoline; Hol, Caroline; Torgersen, Gerald R. & Wikan, Kristine (2017). Workshop om strålevern i tannhelseprofesjonsutdanningene. Den norske tannlegeforenings Tidende. ISSN 0029-2303. 127(11), p. 1016–1017. Full text in Research Archive
• Torgersen, Gerald R.; Arvidsson, Linda Zamoline & Kristine, Wikan (2017). Workshop om strålevern i tannhelseprofesjonsutdanningene. Tannstikka : tidsskrift for Norsk tannpleierforening. ISSN 1501-8636. p. 26–27. Full text in Research Archive
• Torgersen, Gerald R.; Møystad, Anne; Hol, Caroline; Hellén-Halme, Kristina; Strindberg, Jerker Edén & Näsström, Karin [Show all 7 contributors for this article] (2016). CBCT constancy controls: How user-friendly is the new phantom? Oral surgery, oral medicine, oral pathology and oral radiology. ISSN 2212-4403. 121(3), p. 343–344. doi: 10.1016/j.oooo.2015.11.014. Full text in Research Archive
• Torgersen, Gerald R.; Møystad, Anne; Hol, Caroline; Hellén-Halme, Kristina; Strindberg, Jerker Edén & Näsström, Karin [Show all 7 contributors for this article] (2015). CBCT constancy controls: How user friendly is the new phantom? A pilot study. Full text in Research Archive
• Torgersen, Gerald R. (2013). Litt om stråling og strålevern. Full text in Research Archive Show summary

  Kurs for tannhelsesekretærer og tannpleiere: Litt strålefysikk og - biologi Stråledoser Regelverk Praktisk strålevern

• Hol, Caroline; Møystad, Anne; Hellén-Halme, Kristina; Torgersen, Gerald R. & Nilsson, Mats (2013). Workflow and experiences in Norwegian clinicsusing dental CBCT units: A questionnaire study. Full text in Research Archive
• Torgersen, Gerald R. (2013). Litt om stråling og strålevern for pedodontister. Full text in Research Archive Show summary

  Forelesning innen temaene: Strålebiologi Stråledoser Optimalisering innen dental radiologi Med vekt på kommunikasjon med: Foreldre Barn Gravide mennesker

• Torgersen, Gerald R. (2013). Litt om stråling og strålevern. Full text in Research Archive Show summary

  Kort forelesning innen temaene: Litt strålefysikk og - biologi Stråledoser Regelverk Praktisk strålevern

• Nilsson, Mats; Torgersen, Gerald R.; Hellén-Halme, Kristina; Møystad, Anne & Hol, Caroline (2013). A Simple And Inexpensive Phantom For Quality Control Of CBCT Units. Full text in Research Archive
• Torgersen, Gerald R. (2012). Strålevern – ny forskrift – med bakgrunn. Full text in Research Archive
• Torgersen, Gerald R. (2012). Strålevern – ny forskrift – med bakgrunn. Full text in Research Archive
• Torgersen, Gerald R. (2012). Kvalitetskontroller CBCT. Full text in Research Archive
• Torgersen, Gerald R. (2011). Dental Cone Beam CT - Måling av doser og andre parametre QA - Dosesammenlikninger - Pasientdoser. Full text in Research Archive
• Torgersen, Gerald R. (2011). Bildekvalitetskontroll på maxillofaciale CBCT-enheter. Full text in Research Archive
• Torgersen, Gerald Ruiner & Hol, Caroline (2009). Cone Beam CT (CBCT) – en ny modalitet på vei ut til tannlegene. Full text in Research Archive Show summary

  Vi gir en kort introduksjon til teknologien bak CBCT og introduserer noen apparater på markedet. Bruksområder innen maxillofacial radiologi presenteres. En kort status på dagens utbredelse og et anslag av planlagte innkjøp vil bli gitt. Vi ser litt på regelverket for bruk og ettersyn av ustyret, internasjonalt og nasjonalt. Teknologien er ny innen dentale klinikker, derfor mangler det rutiner og rettningslinjer for dosimetri, mottaks- og kvalitetskontroll. Rollen medisinske fysikere skal spille i dette arbeidet diskuteres.

• Hol, Caroline; Olerud, Hilde Merete; Landmark, I.D.; Friberg, Eva G; Torgersen, Gerald Ruiner & Svanæs, Dagfinn Beyer [Show all 7 contributors for this article] (2008). Low-dose MDCT and CBCT: Dose comparison. Full text in Research Archive
• Torgersen, Gerald Ruiner (2006). Anisotrope universmodeller med materie og mørk energi. Universitetet i Oslo. Full text in Research Archive

View all works in NVA