Márcio Sarroglia Pinho
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Márcio Sarroglia Pinho has a total of 140 co-authors. Here are the top co-authors by number of shared publications:
- Leonardo Pavanatto
Department of Computer Science, Center for Human-Computer Interaction, Virginia Tech, Blacksburg, VA, USA
5 shared publications - Rafael Rieder
Universidade de Passo Fundo
5 shared publications - Isabel Harb Manssour
PUCRS
4 shared publications - Regis Kopper
Iowa State University, Ames, USA
4 shared publications - Vicenzo Abichequer Sangalli
Pontifical Catholic University of Rio Grande do Sul (PUCRS), Polytechnic School, Virtual Reality Research Group, Porto Alegre, Brazil
4 shared publications - Fabrício Pretto
Universidade do Vale do Taquari - UNIVATES
3 shared publications
Latest Publications
- Open Access
- 14 November 2026
- Published by Wiley in Orthodontics & Craniofacial Research
Machine Learning Prediction of Orthodontic Extraction Patterns Using Multimodal Features Derived From Intraoral Scans and Lateral Cephalometric Radiographs
Objective: To develop and evaluate a machine learning framework for predicting orthodontic extraction patterns using features extracted from pretreatment intraoral scans (IOS) and lateral cephalometric radiographs (LCRs). Materials and Methods: This retrospective study included 420 subjects (mean age, 17.1 years; 61% female) with treatment plans involving dental extractions. Expert consensus‐derived reference labels were established using a validated Delphi process involving four orthodontic experts. The dataset was randomly split into two mutually exclusive subsets: 30% for hyperparameter optimization using grid search with five‐fold cross‐validation and 70% ( n = 294) for model development and evaluation using leave‐one‐out cross‐validation (LOOCV). Performance was assessed using accuracy, sensitivity, specificity, positive and negative predictive values, likelihood ratios and F1 score. Results: The Gradient Boosting classifier trained exclusively on canine‐molar relationship features achieved the highest overall classification accuracy (79.9%), whereas the Random Forest multimodal model incorporating all IOS‐ and LCR‐derived features achieved 78.2% accuracy and did not provide a significant improvement over canine‐molar relationship features alone. SHAP analysis identified canine and molar relationships as the most influential predictors. For the Random Forest multimodal model, class‐specific sensitivity (recall) was 84% for Class I, 77% for Class II, and 48% for Class III. Conclusions: Sagittal canine‐molar relationship features contained the greatest discriminative information for extraction pattern classification in this dataset, and incorporating the full multimodal feature set did not improve predictive performance. The substantially lower sensitivity for the underrepresented Class III pattern highlights the need for larger, more balanced datasets.
- 14 November 2026
- Published by Wiley in Orthodontics & Craniofacial Research
Open AccessMachine Learning Prediction of Orthodontic Extraction Patterns Using Multimodal Features Derived From Intraoral Scans and Lateral Cephalometric Radiographs
Objective: To develop and evaluate a machine learning framework for predicting orthodontic extraction patterns using features extracted from pretreatment intraoral scans (IOS) and lateral cephalometric radiographs (LCRs). Materials and Methods: This retrospective study included 420 subjects (mean age, 17.1 years; 61% female) with treatment plans involving dental extractions. Expert consensus‐derived reference labels were established using a validated Delphi process involving four orthodontic experts. The dataset was randomly split into two mutually exclusive subsets: 30% for hyperparameter optimization using grid search with five‐fold cross‐validation and 70% ( n = 294) for model development and evaluation using leave‐one‐out cross‐validation (LOOCV). Performance was assessed using accuracy, sensitivity, specificity, positive and negative predictive values, likelihood ratios and F1 score. Results: The Gradient Boosting classifier trained exclusively on canine‐molar relationship features achieved the highest overall classification accuracy (79.9%), whereas the Random Forest multimodal model incorporating all IOS‐ and LCR‐derived features achieved 78.2% accuracy and did not provide a significant improvement over canine‐molar relationship features alone. SHAP analysis identified canine and molar relationships as the most influential predictors. For the Random Forest multimodal model, class‐specific sensitivity (recall) was 84% for Class I, 77% for Class II, and 48% for Class III. Conclusions: Sagittal canine‐molar relationship features contained the greatest discriminative information for extraction pattern classification in this dataset, and incorporating the full multimodal feature set did not improve predictive performance. The substantially lower sensitivity for the underrepresented Class III pattern highlights the need for larger, more balanced datasets.
- 20 August 2025
- Published by Springer Nature
- p. 339-353
Using Frameworks to Develop Augmented Reality Training in Industry: A Systematic Review
- 20 August 2025
- Published by Springer Nature
- p. 339-353
Using Frameworks to Develop Augmented Reality Training in Industry: A Systematic Review
- 8 November 2024
- Published by Springer Nature
- p. 1-10
Automatic Detection of Common Gastroenterological Diseases Using a Small Dataset: A Two-Phase Image Processing Method
- 8 November 2024
- Published by Springer Nature
- p. 1-10
Automatic Detection of Common Gastroenterological Diseases Using a Small Dataset: A Two-Phase Image Processing Method
- Open Access
- 3 July 2024
- Published by MDPI AG in Journal of Imaging
- Vol. 10 (7) , 161
A 2.5D Self-Training Strategy for Carotid Artery Segmentation in T1-Weighted Brain Magnetic Resonance Images
Precise annotations for large medical image datasets can be time-consuming. Additionally, when dealing with volumetric regions of interest, it is typical to apply segmentation techniques on 2D slices, compromising important information for accurately segmenting 3D structures. This study presents a deep learning pipeline that simultaneously tackles both challenges. Firstly, to streamline the annotation process, we employ a semi-automatic segmentation approach using bounding boxes as masks, which is less time-consuming than pixel-level delineation. Subsequently, recursive self-training is utilized to enhance annotation quality. Finally, a 2.5D segmentation technique is adopted, wherein a slice of a volumetric image is segmented using a pseudo-RGB image. The pipeline was applied to segment the carotid artery tree in T1-weighted brain magnetic resonance images. Utilizing 42 volumetric non-contrast T1-weighted brain scans from four datasets, we delineated bounding boxes around the carotid arteries in the axial slices. Pseudo-RGB images were generated from these slices, and recursive segmentation was conducted using a Res-Unet-based neural network architecture. The model’s performance was tested on a separate dataset, with ground truth annotations provided by a radiologist. After recursive training, we achieved an Intersection over Union (IoU) score of (0.68 ± 0.08) on the unseen dataset, demonstrating commendable qualitative results.
- 3 July 2024
- Published by MDPI AG in Journal of Imaging
- Vol. 10 (7) , 161
Open AccessA 2.5D Self-Training Strategy for Carotid Artery Segmentation in T1-Weighted Brain Magnetic Resonance Images
Precise annotations for large medical image datasets can be time-consuming. Additionally, when dealing with volumetric regions of interest, it is typical to apply segmentation techniques on 2D slices, compromising important information for accurately segmenting 3D structures. This study presents a deep learning pipeline that simultaneously tackles both challenges. Firstly, to streamline the annotation process, we employ a semi-automatic segmentation approach using bounding boxes as masks, which is less time-consuming than pixel-level delineation. Subsequently, recursive self-training is utilized to enhance annotation quality. Finally, a 2.5D segmentation technique is adopted, wherein a slice of a volumetric image is segmented using a pseudo-RGB image. The pipeline was applied to segment the carotid artery tree in T1-weighted brain magnetic resonance images. Utilizing 42 volumetric non-contrast T1-weighted brain scans from four datasets, we delineated bounding boxes around the carotid arteries in the axial slices. Pseudo-RGB images were generated from these slices, and recursive segmentation was conducted using a Res-Unet-based neural network architecture. The model’s performance was tested on a separate dataset, with ground truth annotations provided by a radiologist. After recursive training, we achieved an Intersection over Union (IoU) score of (0.68 ± 0.08) on the unseen dataset, demonstrating commendable qualitative results.
- Open Access
- 6 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 259-263
A Framework for Monitoring Cargo Movement using Virtual Reality
- 6 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 259-263
Open AccessA Framework for Monitoring Cargo Movement using Virtual Reality
- Open Access
- 6 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 289-293
A Framework for AR Treasure Hunt Design
This paper proposes a framework for the development of Treasure Hunt games using Augmented Reality for educational purposes. The Framework is based on markers and allows the insertion of clues in the form of a simple instruction or an objective question with multiple choices along with 3D models anchored on the marker. Our main contribution was the development of a method for editing the game directly on the display device, without the need to adjust the content in a desktop editor and with low complexity of use. Overall, the framework was able to create Treasure Hunt applications in an easy and intuitive way.
- 6 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 289-293
Open AccessA Framework for AR Treasure Hunt Design
This paper proposes a framework for the development of Treasure Hunt games using Augmented Reality for educational purposes. The Framework is based on markers and allows the insertion of clues in the form of a simple instruction or an objective question with multiple choices along with 3D models anchored on the marker. Our main contribution was the development of a method for editing the game directly on the display device, without the need to adjust the content in a desktop editor and with low complexity of use. Overall, the framework was able to create Treasure Hunt applications in an easy and intuitive way.
- Open Access
- 25 May 2023
- Published by Wiley in Stroke Research and Treatment
- Vol. 2023, 1-8
Conventional Mirror Therapy versus Immersive Virtual Reality Mirror Therapy: The Perceived Usability after Stroke
Background. Stroke is a widespread and complex health issue, with many survivors requiring long-term rehabilitation due to upper-limb impairment. This study is aimed at comparing the perceived usability of two feedback-based stroke therapies: conventional mirror therapy (MT) and immersive virtual reality mirror therapy (VR). Methods. The study involved 45 participants, divided into three groups: the stroke survivors ( ), stroke-free older adults ( ), and young controls ( ). Participants performed two tasks using both MT and VR in a semirandom sequence. Usability instruments (SUS and NASA-TLX) were applied at the end of the activities, along with two experience-related questions. Results. The results indicated that both MT and VR had similar levels of perceived usability, with MT being more adaptable and causing less overall discomfort. Conversely, VR increased the perception of task difficulty and prevented participants from diverting their attention from the mirror-based feedback. Conclusion. While VR was found to be less comfortable than MT, both systems exhibited similar perceived usability. The comfort levels of the goggles may play a crucial role in determining the usability of VR for upper limb rehabilitation after stroke.
- 25 May 2023
- Published by Wiley in Stroke Research and Treatment
- Vol. 2023, 1-8
Open AccessConventional Mirror Therapy versus Immersive Virtual Reality Mirror Therapy: The Perceived Usability after Stroke
Background. Stroke is a widespread and complex health issue, with many survivors requiring long-term rehabilitation due to upper-limb impairment. This study is aimed at comparing the perceived usability of two feedback-based stroke therapies: conventional mirror therapy (MT) and immersive virtual reality mirror therapy (VR). Methods. The study involved 45 participants, divided into three groups: the stroke survivors ( ), stroke-free older adults ( ), and young controls ( ). Participants performed two tasks using both MT and VR in a semirandom sequence. Usability instruments (SUS and NASA-TLX) were applied at the end of the activities, along with two experience-related questions. Results. The results indicated that both MT and VR had similar levels of perceived usability, with MT being more adaptable and causing less overall discomfort. Conversely, VR increased the perception of task difficulty and prevented participants from diverting their attention from the mirror-based feedback. Conclusion. While VR was found to be less comfortable than MT, both systems exhibited similar perceived usability. The comfort levels of the goggles may play a crucial role in determining the usability of VR for upper limb rehabilitation after stroke.
- Open Access
- 24 October 2022
- Published by Association for Computing Machinery (ACM)
- p. 151-156
Open Source Augmented Reality in Data Center Infrastructure Maintenance
The use of Augmented Reality (AR) is increasing in several application domains. However, applying AR in data center infrastructure management is still limited. Although solutions are providing DCIM (Data Center Infrastructure Management) with augmented reality, they are paid commercial applications that do not provide the source code. Therefore, using such solutions, their reproducibility, and their evaluation is difficult in academic and computational research environments. In this vein, this work aims to create a prototype of an open-source AR mobile application for maintenance management in data centers. Moreover, this work provides a theoretical foundation based on a literature review of solutions available in the academic and industry fields. We evaluated the proposed solution with qualitative and quantitative methods, including user feedback on the developed application and comparison to OpenDCIM - a well-known tool for data center management. The results from the qualitative evaluation demonstrate that the proposed solution improves the user experience. However, it is not possible to infer significant productivity gains from the quantitative evaluation.
- 24 October 2022
- Published by Association for Computing Machinery (ACM)
- p. 151-156
Open AccessOpen Source Augmented Reality in Data Center Infrastructure Maintenance
The use of Augmented Reality (AR) is increasing in several application domains. However, applying AR in data center infrastructure management is still limited. Although solutions are providing DCIM (Data Center Infrastructure Management) with augmented reality, they are paid commercial applications that do not provide the source code. Therefore, using such solutions, their reproducibility, and their evaluation is difficult in academic and computational research environments. In this vein, this work aims to create a prototype of an open-source AR mobile application for maintenance management in data centers. Moreover, this work provides a theoretical foundation based on a literature review of solutions available in the academic and industry fields. We evaluated the proposed solution with qualitative and quantitative methods, including user feedback on the developed application and comparison to OpenDCIM - a well-known tool for data center management. The results from the qualitative evaluation demonstrate that the proposed solution improves the user experience. However, it is not possible to infer significant productivity gains from the quantitative evaluation.