BKBanu Kılıç

Banu Kılıç

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Banu Kılıç has a total of 25 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Photograph-Based Skeletal Malocclusion Screening with Geometric Machine Learning: External Validation Across Turkish and Non-European Cohorts

    by Aslı Eker Davut, Mehmet Cihan Sakman, Ahmet Yağcı, Berrak Çakmak, Emmanuel Jr Dumu, Yunus Emre Uçaker, Cenk Ahmet Akcan, Özgür Kılıç, Tuğba Önal-Süzek, Banu Kılıç

    A photograph-based screening classifier for skeletal malocclusion maps 100 engineered geometric feature vectors, derived from a side-profile smartphone photograph, to orthodontic class labels, enabling non-invasive triage without radiographic imaging. This study reports external validation across two independent cohorts that contributed no cases to training. Two binary tasks were evaluated; Class III against non-Class III, and Class I against Class II; each with its own development set and model, applied without modification at the default threshold. One external cohort was Turkish, drawn from an institution not represented in the development set; the other was a non-European cohort from a Belgian private orthodontic clinic. Sensitivity, specificity and the area under the receiver operating characteristic curve (AUC) were pre-specified as primary outcome measures, since none depends on outcome prevalence. In the Class III task, sensitivity was similar across all three evaluation sets (0.801, 0.836, 0.848; all adjusted p>0.7), while specificity fell from 0.766 in the Turkish external cohort to 0.697 in the non-European cohort. AUC was 0.902 and 0.835 respectively (difference 0.066; p=0.029). In the Class I and Class II task the corresponding values were 0.875 and 0.723 (difference 0.152; p<0.001). A subgroup analysis showed lower sensitivity in children than adults across both cohorts (0.71–0.80 versus 0.89–0.90). AUC was lower in the non-European cohort for all three models in both tasks. The observed performance differences are consistent with a population-specific generalizability effect rather than model-specific failure and support mandatory external validation on target populations before clinical deployment.

  • Open Access

    Evaluation of orthognathic surgery planning with artificial intelligence: a prospective, comparative study

    by Enes Temizkan, B. Zeynep Yörük, Banu Kılıç

    PurposeThis study aimed to evaluate the accuracy of cephalometric analyses performed by deep learning-based AI programs (NemoCeph 2D, OrthoDx, AudaxCeph, and WebCeph) by comparing their results with the gold standard measurements obtained from 3D CT scans in orthognathic surgery patients. Materials and methodsOrthognathic surgery candidates underwent pretreatment 3D CT scans. These scans were manually analyzed using 3D cephalometric software to establish goldstandard landmark positions. Two-dimensional cephalometric images were then derived from the 3D scans, and deep learning–based AI programs automatically identified the landmarks on these images. The AI-generated measurements were compared with the 3D gold standard, and the differences were analyzed statistically.ResultsWhile the ANB angle showed no significant differences between the methods (p=0.061), other measurements—including SNA, SNB, Wits appraisal, Y Axis Angle, and various facial height ratios—showed significant discrepancies (p<0.05).ConclusionAI-based cephalometric analyses showed notable errors compared with the 3D CT gold standard. These findings suggest that deep learning algorithms require further refinement before they can be reliably used for orthognathic surgery planning.


  • Open Access

    Comparative effectiveness of remineralization agents on attachment-associated enamel demineralization in clear aligner patients: A 6-Month DIAGNOdent-Based controlled clinical trial

    by Hasan Yasin Ünal, Banu Kılıç, Evrim Eliguzeloglu Dalkilic

    Background/objectives: Clear aligner attachments increase the risk of white spot lesions (WSL), with a 35.5% incidence in adolescents. The number of anterior attachments is an independent risk factor (OR = 2.192). Despite 17 million patients treated worldwide, no study has quantitatively assessed demineralization around attachment margins. To compare the effectiveness of CPP-ACP, nano-hydroxyapatite (nHAp), and fluoride varnish versus a control for attachment-associated demineralization using DIAGNOdent monitoring. Methods: This prospective controlled clinical trial evaluated 52 patients; 45 were enrolled, and 40 completed after five pre-baseline withdrawals. Participants were allocated to four groups ( n = 10 each): Group A (control, fluoride toothpaste), Group B (CPP-ACP daily), Group C (nHAp professional + home gel), and Group D (fluoride varnish quarterly). DIAGNOdent measurements were taken around attachments at baseline, 1, 3, and 6 months. Linear mixed-effects models analyzed group differences ( p < 0.05). Results: All 40 participants completed the 6-month study (100% retention). Baseline values were comparable ( p = 0.48). Mean DIAGNOdent changes at 6 months were: Control + 4.35 [95%CI: 3.77–4.93], CPP-ACP − 4.02 [− 4.47 to − 3.56], nHAp − 5.36 [− 5.99 to − 4.72], Fluoride − 4.45 [− 4.78 to − 4.11]. The main group effect was significant ( p < 0.001; η² p = 0.974). All treatments outperformed the control ( p < 0.001). nHAp showed the greatest reduction, superior to CPP-ACP ( p < 0.001) and fluoride ( p = 0.019). Fluoride exhibited a biphasic pattern. Conclusions: All remineralization protocols significantly reduced demineralization versus control, with nHAp demonstrating the highest efficacy. Early preventive intervention and regular monitoring are recommended in attachment-bearing patients. Larger randomized trials are warranted. Trial registration: ClinicalTrials.gov NCT07229105. Registered 13 November 2025. Retrospectively registered.


  • Open Access

    Comparative evaluation of the effect of different cleaning agents on colour and surface roughness of Invisalign clear aligners: a cross-over randomized controlled trial

    by Banu Kılıç, Şeymanur Canpolat, Merve Öztürk

    Background: Recent advancements in orthodontics have led to the widespread use of clear aligners, appreciated for their aesthetic and comfort advantages. However, standardized cleaning instructions remain undefined, and maintaining the surface integrity of aligners is critical for treatment efficacy and patient satisfaction. Although several studies have evaluated the effects of cleaning agents, comprehensive clinical evidence regarding their impact on the colour stability and surface roughness of Invisalign aligners is still lacking. This study aims to address this gap by comparing the effects of five commonly used cleaning agents on these key properties. Objectives: To assess the influence of various cleaning agents on both the colour and surface roughness of clear aligners. Materials and methods: This randomized crossover clinical trial included 23 participants with good oral hygiene, each requiring more than 12 aligners. Participants were randomly assigned to use two different cleaning agents in two consecutive treatment phases. The agents tested were: (1) Invisalign Cleaning Crystals (Align Technology, USA), (2) Efferdent Anti-Bacterial Denture Cleanser (Prestige Brands, USA), (3) Sensodyne Rapid Relief toothpaste (GlaxoSmithKline, UK), (4) Signal White Now Whitening Toothpaste (Unilever, USA), and (5) colourant-free clear soap (La Petit Marseillaise, France). Colour measurements (ΔE) were obtained using a spectrophotometer based on the CIE Lab* system, and surface roughness was evaluated with an optical profilometer. Bonferroni-corrected post hoc pairwise comparisons were conducted using Jamovi version 2.6.26. Results: Colourant-free clear soap caused the least colour change, while the Cleaning Crystals and Whitening Toothpaste showed the highest. Control (unused) aligners had greater surface roughness than cleaned aligners. Among cleaning agents, the Cleaning Crystals group exhibited significantly higher roughness than most others ( p < 0.05). Conclusions: Cleaning agents influence the aesthetic and physical properties of aligners to varying degrees. Colourant-free clear soap demonstrated minimal impact on both colour and surface integrity and may be recommended as a safe and inexpensive option for daily aligner care. Trial registration: ClinicalTrials.gov Identifier: NCT05213650. Registered 17 January 2022. Retrospectively registered. This randomized controlled trial adheres to the CONSORT 2010 guidelines for reporting.


  • Open Access

    A family-centered orthodontic screening approach using a machine learning-based mobile application

    by Banu Kılıç, Ahmed Hassan İbRahim, Selahattin Aksoy, Mehmet Cihan Sakman, Gül Sude Demircan, Tuğba Önal-Süzek


  • Open Access

    COMPARATIVE ANALYSIS OF THREE MACHINE LEARNING MODELS FOR EARLY PREDICTION OF SKELETAL CLASS-III MALOCCLUSION FROM PROFILE PHOTOS

    by Selahattin Aksoy, Banu Kiliç, Tuğba SÜZEK

    The pre-adolescent growth period is the best time for the skeletal Class-III malocclusion treatment. Diagnosis and treatment during this period continue to be a complex orthodontic problem. Class-III malocclusion is complicated to treat with braces frequently requiring surgical intervention after a pubertal growth spurt. In addition, delayed recognition of the problem will yield significant functional, aesthetic, and psychological concerns. This study presents the first fully automated machine learning method to accurately diagnose Class-III malocclusion applied across mobile images, to the best of our knowledge. For this purpose, we comparatively evaluated three machine learning approaches: a deep learning algorithm, a machine learning algorithm, and a rule-based algorithm. We collected a novel profile image data set for this analysis along with their formal diagnosis from 435 orthodontics patients. The most successful method among the three was the machine learning method, with an accuracy of %76.


  • IN VITRO COMPARISON OF INDIRECT BONDING METHODS USING DIFFERENT ISOLATION AND ADHESIVE SYSTEMS WITH EACH OTHER AND WITH THE DIRECT BONDING TECHNIQUE

    by Banu Kiliç, Ercan Cenk Doruk

    Aim: To compare the shear bond strength and adhesive remnant index of a new indirect bonding isolation method for different adhesive systems with the direct bonding technique. Methods: 100 premolar and 10 centrally extracted human teeth were bonded with one direct and four indirect bonding methods, and then subjected to 1000 thermal cycles between 5°C and 55°C. ARI was analyzed by the chi-square test. The Shear Bond strength was analyzed by the analysis of variance (ANOVA) and the Tukey multiple comparison (HSD) test. The bracket-enamel interface area of two teeth from each group was examined under SEM. Results: All groups provided clinically acceptable bond strengths. In the traditional isolation + light-cured bonding group with the highest bonding values identified, the scores for all failures were recorded as 3 and this group differs from all other groups in respect of ARI scores (p<0.05). In addition, it was identified from SEM images that there were air bubbles in the adhesive layer in the groups where chemical adhesives were used. Conclusion. When tape is used as the isolation material, the laboratory process stage is shortened and simplified. Therefore, if the custom base indirect bonding technique is to be used, this new method can be used with light-cured bonding agent for isolation.

  • Open Access

    Comparative Analysis of three machine learning models for Early Prediction of Skeletal Class-III Malocclusion from Profile Photos

    by Banu Kiliç, Tuğba Önal-Süzek, Selahattin Aksoy

    The pre-adolescent growth period is the best time for the skeletal Class-III malocclusion treatment. Diagnosis and treatment during this period continue to be a complex orthodontic problem. Class-III malocclusion is complicated to treat with braces frequently requiring surgical intervention after a pubertal growth spurt. In addition, delayed recognition of the problem will yield significant functional, aesthetic, and psychological concerns. This study presents the first fully automated machine learning method to accurately diagnose Class-III malocclusion applied across mobile images, to the best of our knowledge. For this purpose, we comparatively evaluated three machine learning approaches: a deep learning algorithm, a machine learning algorithm, and a rule-based algorithm. We collected a novel profile image data set for this analysis along with their formal diagnosis from 435 orthodontics patients. The most successful method among the three was the machine learning method, with an accuracy of %76.