Maxim Shiltsyn
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Maxim Shiltsyn has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Artem Obukhov
Laboratory of VR Simulators, Tambov State Technical University, Tambov 392000, Russia
4 shared publications - Daniil Teselkin
Laboratory of VR Simulators, Tambov State Technical University, Tambov 392000, Russia
3 shared publications - Denis Dedov
Laboratory of VR Simulators, Tambov State Technical University, Tambov 392000, Russia
3 shared publications - Artem Sveshnikov
Laboratory of VR Simulators, Tambov State Technical University, Tambov 392000, Russia
1 shared publication - Elena Yurievna Kutimova
S.N. Fedorov National Medical Research Center MNTK «Eye Microsurgery», Tambov 392000, Russia
1 shared publication - Irina Mikhailovna Chernukha
V. M. Gorbatov Federal Research Center for Food Systems of the Russian Academy of Sciences
1 shared publication
Latest Publications
- Open Access
- 10 September 2026
- Published by MDPI AG in Journal of Imaging
- Vol. 12 (9) , 429
Multi-Camera Analysis of Simulated Porcine Recovery States Based on Three-Dimensional Pose Using Synthetic Data
Continuous objective assessment of pig motor condition, including recovery after anesthesia, requires the joint analysis of posture, locomotion, transitions between states, and short-term adverse events. This study aimed to develop and algorithmically evaluate a multi-camera pipeline for the quantitative analysis of simulated recovery states in a fully synthetic virtual environment. Procedural generation was used to produce 150,000 images annotated with fifteen anatomical keypoints while varying the pose, size, and appearance of the model, camera viewpoints, illumination, environment, and image post-processing parameters. The YOLO11m-pose model was used for two-dimensional pose estimation, after which observations from three synchronized cameras were combined using weighted triangulation. The reconstructed three-dimensional trajectories were processed using a quality-control system, a finite-state machine, and temporal rules for detecting falls, prolonged immobility, and convulsion-like movements. After repartitioning the dataset by three-dimensional pose index, the retrained YOLO11m-pose model achieved a keypoint of 0.8968, a precision of 0.9985, and a recall of 0.9986 on the independent pose-level test set. During the processing of a 15-min three-camera sequence, 98.993% of 405,000 reconstructions satisfied the geometric acceptance criteria, and the median reprojection error was 2.420 pixels. On a first manually annotated 5-min synthetic sequence, the state machine achieved a strict frame-level accuracy of 91.20%, a macro-F1 score of 90.10%, and Cohen’s of 0.860; accuracy outside s neighborhoods of state transitions was 97.43%. On a second 5-min synthetic evaluation sequence with exact Unity-world geometric ground truth, direct three-dimensional evaluation yielded an MPJPE of 31.86 mm at 99.993% landmark coverage. A limited temporal assessment on this second sequence detected both of the two lying-derived prolonged-immobility reference intervals; fall and convulsion-like-movement events were not independently annotated. All training, validation, and end-to-end evaluation data were generated in a virtual environment; therefore, these results establish algorithmic feasibility within the synthetic domain but do not establish performance on real animals.- 10 September 2026
- Published by MDPI AG in Journal of Imaging
- Vol. 12 (9) , 429
Open AccessMulti-Camera Analysis of Simulated Porcine Recovery States Based on Three-Dimensional Pose Using Synthetic Data
Continuous objective assessment of pig motor condition, including recovery after anesthesia, requires the joint analysis of posture, locomotion, transitions between states, and short-term adverse events. This study aimed to develop and algorithmically evaluate a multi-camera pipeline for the quantitative analysis of simulated recovery states in a fully synthetic virtual environment. Procedural generation was used to produce 150,000 images annotated with fifteen anatomical keypoints while varying the pose, size, and appearance of the model, camera viewpoints, illumination, environment, and image post-processing parameters. The YOLO11m-pose model was used for two-dimensional pose estimation, after which observations from three synchronized cameras were combined using weighted triangulation. The reconstructed three-dimensional trajectories were processed using a quality-control system, a finite-state machine, and temporal rules for detecting falls, prolonged immobility, and convulsion-like movements. After repartitioning the dataset by three-dimensional pose index, the retrained YOLO11m-pose model achieved a keypoint of 0.8968, a precision of 0.9985, and a recall of 0.9986 on the independent pose-level test set. During the processing of a 15-min three-camera sequence, 98.993% of 405,000 reconstructions satisfied the geometric acceptance criteria, and the median reprojection error was 2.420 pixels. On a first manually annotated 5-min synthetic sequence, the state machine achieved a strict frame-level accuracy of 91.20%, a macro-F1 score of 90.10%, and Cohen’s of 0.860; accuracy outside s neighborhoods of state transitions was 97.43%. On a second 5-min synthetic evaluation sequence with exact Unity-world geometric ground truth, direct three-dimensional evaluation yielded an MPJPE of 31.86 mm at 99.993% landmark coverage. A limited temporal assessment on this second sequence detected both of the two lying-derived prolonged-immobility reference intervals; fall and convulsion-like-movement events were not independently annotated. All training, validation, and end-to-end evaluation data were generated in a virtual environment; therefore, these results establish algorithmic feasibility within the synthetic domain but do not establish performance on real animals. - Open Access
A Hybrid Verification Method for an Actuated Lower-Limb Exoskeleton Based on Mathematical and Simulation Analyses
This study proposes a hybrid method for pre-design verification of the geometric and biomechanical compatibility of experimentally recorded gait trajectories with a lower-limb exoskeleton. The method integrates joint-space kinematic analysis, CAD-based self-collision detection in a Unity digital twin, anatomical hip and knee constraints, and experimentally recorded gait trajectories. Gait data were obtained from 30 healthy participants at 12 treadmill speeds ranging from 0.5 to 6.0 km/h using multicamera markerless motion capture. Of 54,481 evaluated configurations, 21,450 remained after self-collision detection and 4080 after anatomical filtering, corresponding to 7.49% of the initial discretized domain. The mean proportion of experimental trajectory points excluded because of mechanical self-collisions increased from 2.70% at 0.5 km/h to 30.49% at 6.0 km/h. Walking speed had a significant and large overall effect (, , Kendall’s ), and the median participant-specific Spearman coefficient was . The proposed framework provides a quantitative pre-screening tool for identifying critical gait configurations before powered prototype testing and can support mechanical redesign and definition of controller constraints.Open AccessA Hybrid Verification Method for an Actuated Lower-Limb Exoskeleton Based on Mathematical and Simulation Analyses
This study proposes a hybrid method for pre-design verification of the geometric and biomechanical compatibility of experimentally recorded gait trajectories with a lower-limb exoskeleton. The method integrates joint-space kinematic analysis, CAD-based self-collision detection in a Unity digital twin, anatomical hip and knee constraints, and experimentally recorded gait trajectories. Gait data were obtained from 30 healthy participants at 12 treadmill speeds ranging from 0.5 to 6.0 km/h using multicamera markerless motion capture. Of 54,481 evaluated configurations, 21,450 remained after self-collision detection and 4080 after anatomical filtering, corresponding to 7.49% of the initial discretized domain. The mean proportion of experimental trajectory points excluded because of mechanical self-collisions increased from 2.70% at 0.5 km/h to 30.49% at 6.0 km/h. Walking speed had a significant and large overall effect (, , Kendall’s ), and the median participant-specific Spearman coefficient was . The proposed framework provides a quantitative pre-screening tool for identifying critical gait configurations before powered prototype testing and can support mechanical redesign and definition of controller constraints. - Open Access
- 17 July 2026
- Published by MDPI AG in Technologies
- Vol. 14 (7) , 438
Comparative Analysis of Local Large Language Models for Ranking Higher Education Programmes Based on Applicant Digital Profiles
Choosing among closely related higher-education programmes requires interpretation of heterogeneous applicant data while preserving data confidentiality. This study compares ten locally executable large language model (LLM) configurations as semantic rankers, contextualises their performance against a hybrid term frequency–inverse document frequency (TF–IDF) cosine baseline, and evaluates robustness when relevant competencies are expressed indirectly. The main benchmark comprised 50 anonymised applicant profiles and 10 degree programmes; an additional processed-profile robustness set comprising 10 profiles was used to reduce direct lexical overlap with the programme catalogue. Rankings were evaluated using Accuracy@1, Accuracy@3, normalised discounted cumulative gain at 5 (NDCG@5) and mean reciprocal rank at 5 (MRR@5), while structured-output validity and inference time were assessed for the LLMs. Among the local LLMs, gpt-oss-20b-MXFP4 achieved the highest Accuracy@1 (0.76), whereas gemma-4-E4B-it-Q8_0 achieved the highest Accuracy@3 (0.92), and Ministral-3-14B-Reasoning-2512-Q4_K_M provided a favourable quality–latency balance. On the main benchmark, TF–IDF achieved Accuracy@1 = 0.88 and NDCG@5 = 0.9557, exceeding all LLM configurations and demonstrating a strong lexical signal. On the processed-profile robustness set, the Accuracy@1 of the TF–IDF baseline decreased to 0.60, whereas gemma achieved 0.90. The results provide initial evidence that several local LLM configurations can reproduce observed programme-selection patterns in a limited pilot benchmark. However, the findings should be interpreted as task-specific model-comparison results rather than as full validation of an autonomous career guidance system.
- 17 July 2026
- Published by MDPI AG in Technologies
- Vol. 14 (7) , 438
Open AccessComparative Analysis of Local Large Language Models for Ranking Higher Education Programmes Based on Applicant Digital Profiles
Choosing among closely related higher-education programmes requires interpretation of heterogeneous applicant data while preserving data confidentiality. This study compares ten locally executable large language model (LLM) configurations as semantic rankers, contextualises their performance against a hybrid term frequency–inverse document frequency (TF–IDF) cosine baseline, and evaluates robustness when relevant competencies are expressed indirectly. The main benchmark comprised 50 anonymised applicant profiles and 10 degree programmes; an additional processed-profile robustness set comprising 10 profiles was used to reduce direct lexical overlap with the programme catalogue. Rankings were evaluated using Accuracy@1, Accuracy@3, normalised discounted cumulative gain at 5 (NDCG@5) and mean reciprocal rank at 5 (MRR@5), while structured-output validity and inference time were assessed for the LLMs. Among the local LLMs, gpt-oss-20b-MXFP4 achieved the highest Accuracy@1 (0.76), whereas gemma-4-E4B-it-Q8_0 achieved the highest Accuracy@3 (0.92), and Ministral-3-14B-Reasoning-2512-Q4_K_M provided a favourable quality–latency balance. On the main benchmark, TF–IDF achieved Accuracy@1 = 0.88 and NDCG@5 = 0.9557, exceeding all LLM configurations and demonstrating a strong lexical signal. On the processed-profile robustness set, the Accuracy@1 of the TF–IDF baseline decreased to 0.60, whereas gemma achieved 0.90. The results provide initial evidence that several local LLM configurations can reproduce observed programme-selection patterns in a limited pilot benchmark. However, the findings should be interpreted as task-specific model-comparison results rather than as full validation of an autonomous career guidance system.
- Open Access
- 11 November 2024
- Published by MDPI AG in Technologies
- Vol. 12 (11) , 228
Medical VR Simulator for Pediatric Strabismus Treatment
In the process of treating pediatric strabismus, great difficulties arise with maintaining attention and a high level of motivation in patients. Existing computer programs and medical equipment should be supplemented with more modern tools and approaches based on virtual reality (VR) technologies, ensuring the full immersion of children in the treatment process. The aim of this study is to develop and evaluate the effectiveness of a virtual reality medical simulator for the treatment of pediatric strabismus. The specifics of the realization of ophthalmic exercises for the virtual simulator and the methods for evaluation of the quality of their performance are considered. In the course of experimental research, a control group of 58 people using the VR simulator and a reference group of 59 people receiving the standard process of strabismus treatment were compared. The average value of visual acuity in the control group increased from 66.1% to 80.4% (p = 0.002); the average value of objective strabismus angle decreased from 5° to 0° (p < 0.001). The subjective strabismus angle was also found to improve from 3° to 0° (p < 0.001). In terms of exercise quality metrics, a selected subgroup of participants who have been training for a long time have shown positive dynamics in terms of improved accuracy and a reduction in their average number of errors. In summary, virtual reality technologies demonstrated a statistically significant improvement in the metrics used to evaluate orthoptic treatment in the control group and the superiority of this approach over standard treatment.
- 11 November 2024
- Published by MDPI AG in Technologies
- Vol. 12 (11) , 228
Open AccessMedical VR Simulator for Pediatric Strabismus Treatment
In the process of treating pediatric strabismus, great difficulties arise with maintaining attention and a high level of motivation in patients. Existing computer programs and medical equipment should be supplemented with more modern tools and approaches based on virtual reality (VR) technologies, ensuring the full immersion of children in the treatment process. The aim of this study is to develop and evaluate the effectiveness of a virtual reality medical simulator for the treatment of pediatric strabismus. The specifics of the realization of ophthalmic exercises for the virtual simulator and the methods for evaluation of the quality of their performance are considered. In the course of experimental research, a control group of 58 people using the VR simulator and a reference group of 59 people receiving the standard process of strabismus treatment were compared. The average value of visual acuity in the control group increased from 66.1% to 80.4% (p = 0.002); the average value of objective strabismus angle decreased from 5° to 0° (p < 0.001). The subjective strabismus angle was also found to improve from 3° to 0° (p < 0.001). In terms of exercise quality metrics, a selected subgroup of participants who have been training for a long time have shown positive dynamics in terms of improved accuracy and a reduction in their average number of errors. In summary, virtual reality technologies demonstrated a statistically significant improvement in the metrics used to evaluate orthoptic treatment in the control group and the superiority of this approach over standard treatment.