Chang-Su Na
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
Research Subjects
Chang-Su Na has a total of 213 co-authors. Here are the top co-authors by number of shared publications:
- Hong-Seok Son
Department of Biotechnology, College of Life Sciences and Biotechnology, Korea University, Seoul 02841, Republic of Korea
9 shared publications - Byeong-Il Lee
Industry 4.0 Convergence Bionics Engineering, Pukyong National University, Busan 48513, Republic of Korea
7 shared publications - Myunggi Yi
Industry 4.0 Convergence Bionics Engineering, Pukyong National University, Busan 48513, Republic of Korea
7 shared publications - Hi-Joon Park
College of Korean Medicine, Kyung Hee University, Seoul 02453, Republic of Korea
6 shared publications - Seon-Jong Kim
Department of Korean Medicine Rehabilitation, Dongshin University Mokpo Korean Medicine Hospital
6 shared publications - Chan-Hun Choi
College of Korean Medicine, Dongshin University, 185 Geonjae-ro, Naju-si, Jeollanam-do, 58245, Republic of Korea
5 shared publications
Latest Publications
- Open Access
- 4 August 2026
- Published by Korean Society for Microbiology and Biotechnology in Journal of Microbiology and Biotechnology
- Vol. 36  (8) , 1-12
Banhasasim-tang Attenuates Corticosterone-Induced Behavioral Despair-like Responses and Is Associated with Gut Microbiota and Serum Metabolomic Changes in Mice
Seung-Ho Seo, Sang-Mi Kang, Siwon Kang, Yanghee You, Yujin Lee, Mee-Hyun Lee, and Chang-Su Na. J. Microbiol. Biotechnol. 2026;36:1-12. https://doi.org/10.4014/jmb.2606.06018
- 4 August 2026
- Published by Korean Society for Microbiology and Biotechnology in Journal of Microbiology and Biotechnology
- Vol. 36  (8) , 1-12
Open AccessBanhasasim-tang Attenuates Corticosterone-Induced Behavioral Despair-like Responses and Is Associated with Gut Microbiota and Serum Metabolomic Changes in Mice
Seung-Ho Seo, Sang-Mi Kang, Siwon Kang, Yanghee You, Yujin Lee, Mee-Hyun Lee, and Chang-Su Na. J. Microbiol. Biotechnol. 2026;36:1-12. https://doi.org/10.4014/jmb.2606.06018
- Open Access
- 31 July 2026
- Published by The Society of Korean Medicine Rehabilitation in Journal of Korean Medicine Rehabilitation
- Vol. 36  (3) , 117-126
Integrative Korean Medicine Treatment Including Craniosacral Therapy for Tinnitus with Brain Fog: A CARE-Compliant Case Report
Geun Jeong Kim, M.S., Yoon Ah Kim, M.S., Ye Rynn Jeong, M.S., Rae On Jang, K.M.D., Jae-Uk Sul, Ph.D., Jin Bong Choi, K.M.D., Chang-Su Na, Ph.D.. J Korean Med Rehabi 2026;36:117-26. https://doi.org/10.18325/jkmr.2026.36.3.117
- 31 July 2026
- Published by The Society of Korean Medicine Rehabilitation in Journal of Korean Medicine Rehabilitation
- Vol. 36  (3) , 117-126
Open AccessIntegrative Korean Medicine Treatment Including Craniosacral Therapy for Tinnitus with Brain Fog: A CARE-Compliant Case Report
Geun Jeong Kim, M.S., Yoon Ah Kim, M.S., Ye Rynn Jeong, M.S., Rae On Jang, K.M.D., Jae-Uk Sul, Ph.D., Jin Bong Choi, K.M.D., Chang-Su Na, Ph.D.. J Korean Med Rehabi 2026;36:117-26. https://doi.org/10.18325/jkmr.2026.36.3.117
- Open Access
- 19 April 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16  (8) , 3968
Precision Without Complexity: A Comparative Study of YOLO26 Pose Variants for Distal Arm Landmark Detection
Accurate anatomical landmark localization in clinical images requires millimeter-level spatial precision, yet whether increasing model scale improves such precision in structured medical imaging tasks remains unclear. Five YOLO26 pose-estimation variants (N, S, M, L, and X) were evaluated on 3679 RGB distal-arm images from 262 participants under a standardized overhead imaging protocol, with five anatomical landmarks annotated across the proximal forearm, mid-forearm, and hand. Localization error was quantified in millimeters using ArUco-marker-based pixel-to-millimeter calibration; all models were initialized from COCO-pretrained weights, fine-tuned under identical conditions, and assessed using COCO-style detection metrics and physically grounded localization error. Detection performance saturated across all scales (mAP@0.5 = 99.5%), while localization performance differed substantially; YOLO26N achieved the lowest mean error (2.76 ± 0.96 mm) and the highest proportion of predictions within 4 mm (88.0%), whereas YOLO26X produced the highest mean error (4.08 ± 2.59 mm) despite a 26.9× higher computational cost. Landmark-wise analysis revealed a consistent proximal-to-distal error gradient, with the largest degradation at anatomically ambiguous proximal landmarks in larger models. These findings suggest that increasing model capacity does not improve clinically meaningful localization precision in structured distal-arm imaging, and lightweight models may offer the most favorable accuracy-efficiency trade-off in resource-constrained clinical settings.
- 19 April 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16  (8) , 3968
Open AccessPrecision Without Complexity: A Comparative Study of YOLO26 Pose Variants for Distal Arm Landmark Detection
Accurate anatomical landmark localization in clinical images requires millimeter-level spatial precision, yet whether increasing model scale improves such precision in structured medical imaging tasks remains unclear. Five YOLO26 pose-estimation variants (N, S, M, L, and X) were evaluated on 3679 RGB distal-arm images from 262 participants under a standardized overhead imaging protocol, with five anatomical landmarks annotated across the proximal forearm, mid-forearm, and hand. Localization error was quantified in millimeters using ArUco-marker-based pixel-to-millimeter calibration; all models were initialized from COCO-pretrained weights, fine-tuned under identical conditions, and assessed using COCO-style detection metrics and physically grounded localization error. Detection performance saturated across all scales (mAP@0.5 = 99.5%), while localization performance differed substantially; YOLO26N achieved the lowest mean error (2.76 ± 0.96 mm) and the highest proportion of predictions within 4 mm (88.0%), whereas YOLO26X produced the highest mean error (4.08 ± 2.59 mm) despite a 26.9× higher computational cost. Landmark-wise analysis revealed a consistent proximal-to-distal error gradient, with the largest degradation at anatomically ambiguous proximal landmarks in larger models. These findings suggest that increasing model capacity does not improve clinically meaningful localization precision in structured distal-arm imaging, and lightweight models may offer the most favorable accuracy-efficiency trade-off in resource-constrained clinical settings.
- Open Access
- 26 January 2026
- Published by Korean Society for Microbiology and Biotechnology in Journal of Microbiology and Biotechnology
- Vol. 36  (1) , 1-12
Multi-Layered Antidepressant Mechanisms of Gami-Soyosan in a Corticosterone-Induced Mouse Model
Seung-Ho Seo, Seungeun Yeo, Sang-Mi Kang, Yang-Hee You, Siwon Kang, Yujin Lee, Youngshik Choe, and Chang-Su Na. J. Microbiol. Biotechnol. 2026;36:1-12. https://doi.org/10.4014/jmb.2509.09033
- 26 January 2026
- Published by Korean Society for Microbiology and Biotechnology in Journal of Microbiology and Biotechnology
- Vol. 36  (1) , 1-12
Open AccessMulti-Layered Antidepressant Mechanisms of Gami-Soyosan in a Corticosterone-Induced Mouse Model
Seung-Ho Seo, Seungeun Yeo, Sang-Mi Kang, Yang-Hee You, Siwon Kang, Yujin Lee, Youngshik Choe, and Chang-Su Na. J. Microbiol. Biotechnol. 2026;36:1-12. https://doi.org/10.4014/jmb.2509.09033
- Open Access
- 3 January 2026
- Published by Springer Nature in Gut Pathogens
- Vol. 18  (1)
Multi-aspect therapeutic effects of a polyphenolic herbal formulation Cirsium japonicum, Scutellaria baicalensis, Paeonia japonica, and Glycyrrhiza uralensis on ulcerative colitis: inflammation modulation, gut microbiota remodeling, and metabolite profiling
Ulcerative Colitis (UC) is a chronic illness that commonly demands the use of medication, sometimes for long term. In a DSS mouse model, we examined 5-amin
- 3 January 2026
- Published by Springer Nature in Gut Pathogens
- Vol. 18  (1)
Open AccessMulti-aspect therapeutic effects of a polyphenolic herbal formulation Cirsium japonicum, Scutellaria baicalensis, Paeonia japonica, and Glycyrrhiza uralensis on ulcerative colitis: inflammation modulation, gut microbiota remodeling, and metabolite profiling
Ulcerative Colitis (UC) is a chronic illness that commonly demands the use of medication, sometimes for long term. In a DSS mouse model, we examined 5-amin
- Open Access
- 27 November 2025
- Published by MDPI AG in Healthcare
- Vol. 13  (23) , 3093
MetaAcuPoint: MetaHuman-Generated Synthetic Data for Hand Acupoint Localization
Background: Precise localization of acupuncture points (acupoints) is crucial for the clinical success of Traditional Korean Medicine (TKM). Traditional methods that rely on visual inspection and palpation are subjective and prone to inter- and intra-observer differences, making standardization challenging. The progress of data-driven localization techniques is also limited by the scarcity of annotated datasets and inconsistent labeling quality. Objective: This study presents MetaAcuPoint, a synthetic dataset created to overcome these limitations by providing high-fidelity, anatomically consistent hand images for acupoint localization. Methods: MetaAcuPoint was generated using MetaHuman avatars within Unreal Engine, resulting in 900 RGB hand images. Anatomically aligned, bone-attached sockets were implemented for five diagnostically relevant hand acupoints, ensuring millimeter-level precision and spatial consistency across various hand poses. Dataset validity was assessed by training a high-resolution network (HRNet-W48) within the MMPose framework and testing its performance on real-world forearm images. Results: The synthetic-trained model achieved a mean distance error (MDE) of 5.67 ± 3.13 pixels, closely aligning with the real-data baseline at 4.81 ± 2.85 pixels. Adding synthetic samples to real data further enhanced performance (MDE: 4.95 pixels). In contrast, manually annotated synthetic images yielded poorer results (MDE: 12.76 pixels), emphasizing the advantages of automated anatomical annotation. Generalization tests across four external datasets confirmed that the synthetic data-trained model outperformed the real-data-trained model, maintaining higher accuracy (MDE: 5.84–6.45 mm vs. 10.63–15.80 mm). Conclusions: MetaAcuPoint demonstrates the first example of synthetic-to-real generalization for hand acupoint localization. By combining photorealistic rendering with anatomically grounded annotation, the dataset offers a reliable resource to promote standardized, data-driven approaches in acupuncture research and practice.
- 27 November 2025
- Published by MDPI AG in Healthcare
- Vol. 13  (23) , 3093
Open AccessMetaAcuPoint: MetaHuman-Generated Synthetic Data for Hand Acupoint Localization
Background: Precise localization of acupuncture points (acupoints) is crucial for the clinical success of Traditional Korean Medicine (TKM). Traditional methods that rely on visual inspection and palpation are subjective and prone to inter- and intra-observer differences, making standardization challenging. The progress of data-driven localization techniques is also limited by the scarcity of annotated datasets and inconsistent labeling quality. Objective: This study presents MetaAcuPoint, a synthetic dataset created to overcome these limitations by providing high-fidelity, anatomically consistent hand images for acupoint localization. Methods: MetaAcuPoint was generated using MetaHuman avatars within Unreal Engine, resulting in 900 RGB hand images. Anatomically aligned, bone-attached sockets were implemented for five diagnostically relevant hand acupoints, ensuring millimeter-level precision and spatial consistency across various hand poses. Dataset validity was assessed by training a high-resolution network (HRNet-W48) within the MMPose framework and testing its performance on real-world forearm images. Results: The synthetic-trained model achieved a mean distance error (MDE) of 5.67 ± 3.13 pixels, closely aligning with the real-data baseline at 4.81 ± 2.85 pixels. Adding synthetic samples to real data further enhanced performance (MDE: 4.95 pixels). In contrast, manually annotated synthetic images yielded poorer results (MDE: 12.76 pixels), emphasizing the advantages of automated anatomical annotation. Generalization tests across four external datasets confirmed that the synthetic data-trained model outperformed the real-data-trained model, maintaining higher accuracy (MDE: 5.84–6.45 mm vs. 10.63–15.80 mm). Conclusions: MetaAcuPoint demonstrates the first example of synthetic-to-real generalization for hand acupoint localization. By combining photorealistic rendering with anatomically grounded annotation, the dataset offers a reliable resource to promote standardized, data-driven approaches in acupuncture research and practice.
- Open Access
Advancing Medical Training with Mixed Reality and Haptic Feedback Simulator for Acupuncture Needling
Traditional acupuncture training often lacks consistent, objective feedback, while current extended reality (XR) solutions rarely include quantitative assessment. This study developed and evaluated a feedback-enabled mixed reality (MR) acupuncture simulator to improve skill acquisition through depth-responsive guidance. The system, used on Microsoft HoloLens 2, combines a MetaHuman-based virtual patient with expert-designed acupoint geometries. It provides depth-dependent vibrotactile cues via a wearable haptic device and calculates a composite score from normalized metrics, including insertion depth, angular deviation, tip-to-center distance, and task duration. Ten participants (eight novices and two experts) performed needle tasks at LI4, LI11, and TE3 across two sessions. Mean depth error decreased from 6.41 mm to 3.58 mm, and task time from 9.29 s to 6.83 s. At LI11, beginners improved in achieved depth (16.24 ± 1.88 mm to 19.74 ± 1.23 mm), reduced angular deviation (27.83° to 15.34°), and shortened completion time (38.77 s to 13.28 s). Experts outperformed novices (69.25 ± 21.64 vs. 56.26 ± 23.37), confirming construct validity. Usability evaluation showed a mean overall score of 4.46 ± 0.51 and excellent reliability (McDonald’s ω = 0.93). These results demonstrate that expert-informed scoring and depth-responsive haptic feedback substantially enhance accuracy, efficiency, and learning confidence, validating the system’s technical robustness and educational readiness for clinical acupuncture training.
Open AccessAdvancing Medical Training with Mixed Reality and Haptic Feedback Simulator for Acupuncture Needling
Traditional acupuncture training often lacks consistent, objective feedback, while current extended reality (XR) solutions rarely include quantitative assessment. This study developed and evaluated a feedback-enabled mixed reality (MR) acupuncture simulator to improve skill acquisition through depth-responsive guidance. The system, used on Microsoft HoloLens 2, combines a MetaHuman-based virtual patient with expert-designed acupoint geometries. It provides depth-dependent vibrotactile cues via a wearable haptic device and calculates a composite score from normalized metrics, including insertion depth, angular deviation, tip-to-center distance, and task duration. Ten participants (eight novices and two experts) performed needle tasks at LI4, LI11, and TE3 across two sessions. Mean depth error decreased from 6.41 mm to 3.58 mm, and task time from 9.29 s to 6.83 s. At LI11, beginners improved in achieved depth (16.24 ± 1.88 mm to 19.74 ± 1.23 mm), reduced angular deviation (27.83° to 15.34°), and shortened completion time (38.77 s to 13.28 s). Experts outperformed novices (69.25 ± 21.64 vs. 56.26 ± 23.37), confirming construct validity. Usability evaluation showed a mean overall score of 4.46 ± 0.51 and excellent reliability (McDonald’s ω = 0.93). These results demonstrate that expert-informed scoring and depth-responsive haptic feedback substantially enhance accuracy, efficiency, and learning confidence, validating the system’s technical robustness and educational readiness for clinical acupuncture training.
- Open Access
- 1 November 2025
- Published by MDPI AG in Journal of Imaging
- Vol. 11  (11) , 388
Evaluating Feature-Based Homography Pipelines for Dual-Camera Registration in Acupoint Annotation
Reliable acupoint localization is essential for developing artificial intelligence (AI) and extended reality (XR) tools in traditional Korean medicine; however, conventional annotation of 2D images often suffers from inter- and intra-annotator variability. This study presents a low-cost dual-camera imaging system that fuses infrared (IR) and RGB views on a Raspberry Pi 5 platform, incorporating an IR ink pen in conjunction with a 780 nm emitter array to standardize point visibility. Among the tested marking materials, the IR ink showed the highest contrast and visibility under IR illumination, making it the most suitable for acupoint detection. Five feature detectors (SIFT, ORB, KAZE, AKAZE, and BRISK) were evaluated with two matchers (FLANN and BF) to construct representative homography pipelines. Comparative evaluations across multiple camera-to-surface distances revealed that KAZE + FLANN achieved the lowest mean 2D error (1.17 ± 0.70 px) and the lowest mean aspect-aware error (0.08 ± 0.05%) while remaining computationally feasible on the Raspberry Pi 5. In hand-image experiments across multiple postures, the dual-camera registration maintained a mean 2D error below ~3 px and a mean aspect-aware error below ~0.25%, confirming stable and reproducible performance. The proposed framework provides a practical foundation for generating high-quality acupoint datasets, supporting future AI-based localization, XR integration, and automated acupuncture-education systems.
- 1 November 2025
- Published by MDPI AG in Journal of Imaging
- Vol. 11  (11) , 388
Open AccessEvaluating Feature-Based Homography Pipelines for Dual-Camera Registration in Acupoint Annotation
Reliable acupoint localization is essential for developing artificial intelligence (AI) and extended reality (XR) tools in traditional Korean medicine; however, conventional annotation of 2D images often suffers from inter- and intra-annotator variability. This study presents a low-cost dual-camera imaging system that fuses infrared (IR) and RGB views on a Raspberry Pi 5 platform, incorporating an IR ink pen in conjunction with a 780 nm emitter array to standardize point visibility. Among the tested marking materials, the IR ink showed the highest contrast and visibility under IR illumination, making it the most suitable for acupoint detection. Five feature detectors (SIFT, ORB, KAZE, AKAZE, and BRISK) were evaluated with two matchers (FLANN and BF) to construct representative homography pipelines. Comparative evaluations across multiple camera-to-surface distances revealed that KAZE + FLANN achieved the lowest mean 2D error (1.17 ± 0.70 px) and the lowest mean aspect-aware error (0.08 ± 0.05%) while remaining computationally feasible on the Raspberry Pi 5. In hand-image experiments across multiple postures, the dual-camera registration maintained a mean 2D error below ~3 px and a mean aspect-aware error below ~0.25%, confirming stable and reproducible performance. The proposed framework provides a practical foundation for generating high-quality acupoint datasets, supporting future AI-based localization, XR integration, and automated acupuncture-education systems.
- Open Access
- 19 October 2025
- Published by Springer Nature
- p. 86-97
Evaluating the Impact of Backbone Networks and Input Resolution on Forearm Acupoint Localization
Accurate localization of anatomical keypoints is fundamental to a variety of clinical and research applications, including acupoint therapy, physical rehabilitation, and biomechanical motion analysis. In this study, we investigate the impact of different deep learning backbone architectures and input image resolutions on the localization accuracy of five clinically important forearm acupoints (LI11, LI10, TE5, LI4, and TE3) using a top-down pose estimation framework. Six backbone models, High-Resolution Network (HRNet), Lite High-Resolution Network (Lite-HRNet), Residual Network (ResNet), Swin Transformer, Visual Geometry Group (VGG) network, and Vision Transformer (ViT), are benchmarked across three input resolutions: 256 × 256, 384 × 384, and 512 × 512 pixels. Localization performance is evaluated using normalized Euclidean distance and its standard deviation, providing a resolution-independent measure of both accuracy and consistency. Among the tested models, HRNet consistently yields the most precise and stable predictions, achieving the lowest average error of 0.009 at the 512 × 512 resolution. Lite-HRNet and ResNet also demonstrate strong performance, particularly at higher resolutions, offering a compelling balance between accuracy and computational efficiency. While ViT exhibits higher error rates at lower resolutions, it significantly improves at 512 × 512, closing the performance gap with CNN-based models. Swin Transformer, however, records the highest average error across all resolutions, indicating its limited suitability for fine-grained localization tasks without additional architectural adaptation. Overall, our findings highlight the importance of backbone selection and resolution scaling in medical keypoint detection tasks. HRNet delivers over 50% lower normalized error compared to the weakest-performing model. These insights provide practical guidance for developing robust, high-precision AI systems for clinical and digital health applications, where anatomical accuracy is essential.
- 19 October 2025
- Published by Springer Nature
- p. 86-97
Open AccessEvaluating the Impact of Backbone Networks and Input Resolution on Forearm Acupoint Localization
Accurate localization of anatomical keypoints is fundamental to a variety of clinical and research applications, including acupoint therapy, physical rehabilitation, and biomechanical motion analysis. In this study, we investigate the impact of different deep learning backbone architectures and input image resolutions on the localization accuracy of five clinically important forearm acupoints (LI11, LI10, TE5, LI4, and TE3) using a top-down pose estimation framework. Six backbone models, High-Resolution Network (HRNet), Lite High-Resolution Network (Lite-HRNet), Residual Network (ResNet), Swin Transformer, Visual Geometry Group (VGG) network, and Vision Transformer (ViT), are benchmarked across three input resolutions: 256 × 256, 384 × 384, and 512 × 512 pixels. Localization performance is evaluated using normalized Euclidean distance and its standard deviation, providing a resolution-independent measure of both accuracy and consistency. Among the tested models, HRNet consistently yields the most precise and stable predictions, achieving the lowest average error of 0.009 at the 512 × 512 resolution. Lite-HRNet and ResNet also demonstrate strong performance, particularly at higher resolutions, offering a compelling balance between accuracy and computational efficiency. While ViT exhibits higher error rates at lower resolutions, it significantly improves at 512 × 512, closing the performance gap with CNN-based models. Swin Transformer, however, records the highest average error across all resolutions, indicating its limited suitability for fine-grained localization tasks without additional architectural adaptation. Overall, our findings highlight the importance of backbone selection and resolution scaling in medical keypoint detection tasks. HRNet delivers over 50% lower normalized error compared to the weakest-performing model. These insights provide practical guidance for developing robust, high-precision AI systems for clinical and digital health applications, where anatomical accuracy is essential.
- Open Access
- 19 October 2025
- Published by Springer Nature
- p. 165-177
High-Fidelity Synthetic MetaAcuPoint Depth (MAP-d) Dataset for Acupoint Localization Using MetaHuman Avatars
Manual annotation of anatomical landmarks, such as acupuncture points or acupoints, is labor-intensive, prone to variability, and limited in clinical datasets. To overcome these constraints, this study introduces the MetaAcuPoint depth dataset (MAP-d dataset), a novel high-fidelity synthetic RGB-D dataset generated using Unreal Engine 5.4 and Epic Games’ MetaHuman avatars. The dataset provides pixel-aligned RGB-D image pairs with anatomically consistent annotations for five clinically relevant acupoints: LI4, TE3, TE5, LI10, and LI11. Reusable skeletal sockets ensure annotation reproducibility across diverse hand morphologies. Domain randomization in pose, skin tone, and skeletal structure enhances the dataset’s demographic and anatomical variability. A lightweight convolutional neural network benchmarked the dataset using RGB-only and RGB-D inputs. While absolute localization accuracy remains modest due to the model’s simplicity, depth augmentation consistently improved performance, yielding a 22.9% reduction in mean distance error and a 57.6% increase in PCK@10. These results confirm the effectiveness of the dataset design. MAP-d provides a scalable resource for developing acupoint localization models, with future potential in clinical training, augmented reality, and real-world deployment.
- 19 October 2025
- Published by Springer Nature
- p. 165-177
Open AccessHigh-Fidelity Synthetic MetaAcuPoint Depth (MAP-d) Dataset for Acupoint Localization Using MetaHuman Avatars
Manual annotation of anatomical landmarks, such as acupuncture points or acupoints, is labor-intensive, prone to variability, and limited in clinical datasets. To overcome these constraints, this study introduces the MetaAcuPoint depth dataset (MAP-d dataset), a novel high-fidelity synthetic RGB-D dataset generated using Unreal Engine 5.4 and Epic Games’ MetaHuman avatars. The dataset provides pixel-aligned RGB-D image pairs with anatomically consistent annotations for five clinically relevant acupoints: LI4, TE3, TE5, LI10, and LI11. Reusable skeletal sockets ensure annotation reproducibility across diverse hand morphologies. Domain randomization in pose, skin tone, and skeletal structure enhances the dataset’s demographic and anatomical variability. A lightweight convolutional neural network benchmarked the dataset using RGB-only and RGB-D inputs. While absolute localization accuracy remains modest due to the model’s simplicity, depth augmentation consistently improved performance, yielding a 22.9% reduction in mean distance error and a 57.6% increase in PCK@10. These results confirm the effectiveness of the dataset design. MAP-d provides a scalable resource for developing acupoint localization models, with future potential in clinical training, augmented reality, and real-world deployment.