臺灣博碩士論文加值系統

圖像處理技術的進步已經徹底改變了各個領域，而醫療保健也不例外。對於傷口的評估和持續監測在醫療保健中扮演著至關重要的角色，因為它們能夠了解患者病情的走向並指導適當的治療策略。傳統的傷口評估方法依賴視覺檢查，需要專業知識並且通常非常耗時。這項研究利用人工智慧（AI）的功能，並借助基於影像處理的深度學習（DL）自動分析傷口特徵，包括大小、顏色和分類。透過這樣的方式，它提供了易於理解的評估結果。在這個項目中，我們提出了一種先進的影像處理方法，特別為醫療保健領域的傷口評估和監測需求而設計。這種方法充分利用了人工智慧和尖端的影像處理技術，提高了傷口影像的質量。透過應用精確和增強的影像處理方法，我們的應用程式呈現出更清晰、更詳細的傷口表現。這進一步幫助醫療保健專業人員準確評估和追蹤傷口的進展。這種創新代表了醫學領域的重大進步，為治療決策提供了準確可靠的見解。
本研究的中心目標是引入一種創新方法，精確計算傷口深度，通過將2D影像轉換為3D熱圖的開創性方法來實現。這種方法論的進步以利用 YOLO 進行傷口檢測和隨後的掩模分割為中心，有望重新定義我們對傷口特徵的理解。分割後，數據進一步轉換為2D影像和3D熱圖，從而能夠預測和計算傷口深度。值得注意的是，這種方法超越了傳統方法，提供了一個複雜的框架，用於監測傷口的進展和預測癒合趨勢。實施這種方法對傷口護理和管理領域做出了重大貢獻，並標誌著在追求更準確和更全面的傷口評估方面邁出了關鍵的一步。這個項目強調了我們致力於推動傷口分析邊緣的承諾，為在當前研究工作最前沿的方式下準確。
關鍵詞 - 深度學習、深度檢測、傷口深度圖、2D 到 3D 映射。

Advancements in image processing techniques have significantly transformed various fields, with healthcare no exception. The assessment and ongoing monitoring of wounds are critical components of medical care, as they inform the trajectory of a patient's condition and guide the formulation of proper treatment strategies. Conventional methods of wound evaluation, reliant on visual inspection, demand specialized ability and are often time-consuming. This research harnesses the capabilities of artificial intelligence (AI) and uses deep learning (DL) grounded in image processing to automate the analysis of wound characteristics, encompassing size, color, and categorization. By doing so, it gives easily interpretable evaluation outcomes.
Within this project, we present an advanced image-processing approach tailored to cater to the healthcare sector's wound assessment and monitoring requirements. This method harnesses the power of artificial intelligence and innovative image processing techniques to elevate the quality of wound imagery. By applying precise and enhanced image processing methods, our application yields crisper, more detailed representations of wounds. This, in turn, aids healthcare professionals in accurately assessing and tracking wound progression. This innovation stands for a significant stride within the medical domain, giving correct and dependable insights to inform treatment decisions.
The central aim of this study is to introduce an innovative method for the precise computation of wound depth, achieved through a pioneering approach of transforming 2D images into 3D heatmaps. This methodological advancement, centered around using YOLO for wound detection and the next mask segmentation to derive wound area, is poised to redefine our understanding of wound characteristics. Following this segmentation, the data is further transformed from 2D images into 3D heat maps, enabling the prediction and calculation of wound depth. Notably, this approach goes beyond conventional methods, offering a sophisticated framework for checking wound progression and predicting healing trends. Implementing this method contributes significantly to the wound care and management field and marks a crucial step forward in pursuing a more correct and comprehensive assessment of wounds. This project underscores our commitment to pushing the boundaries of wound analysis, supplying a robust foundation for the precise determination of wound depth in a manner that stands at the forefront of current research endeavors.
Index Terms - Deep learning, depth detection, wound depth maps, 2D to 3D mapping.

TABLE OF CONTENTS
LIST OF FIGURES 15
LIST OF TABLES 17
ABBREVIATION 18
CHAPTER 1 ORGANIZATION 19
1.1 Motivation 19
1.2 Dissertation Organization 20
CHAPTER 2 INTRODUCTION 21
2.1 Chronic Wound Characteristics 21
2.2 Deep learning for wound image 23
2.3 Depth Prediction for Wound Assessment 28
2.4 Summary 32
CHAPTER 3 RESEARCH METHODS AND IMAGING ARCHITECTURE DESIGN 33
3.1 Wound Detection, Mask Segmentation, and Wound Area Calculation 34
3.2 Calculate the Depth of an Object in a 2-D Image and Map it into 3-D Heatmap 49
3.3 Summary 58
CHAPTER 4 EXPERIMENTAL RESULT AND DISCUSSION 59
4.1 Wound Region Detection Based on Deep Learning Model Discussion 59
4.2 Wound Area Calculation Based on Mask Segmentation 62
4.3 Wound Depth Estimation Using 3D Heatmaps 64
4.4 Prediction and Progress Evaluation 69
4.5 Summary 70
CHAPTER 5 CONCLUSION AND FUTURE WORK 71
5.1 Conclusion 71
5.2 Future Work 72
5.3 Defect Detection Beyond Chronic Wound Analysis 73
REFERENCES 76
BIOGRAPHY 78

REFERENCES
[1]L. E. Edsberg, J. M. Black, M. Goldberg, L. McNichol, L. Moore, M. Sieggreen, “Revised national pressure ulcer advisory panel pressure injury staging system: Revised pressure injury staging system,” J Wound Ostomy Continence Nurs. 43(6) (2016) 585–597.
[2]C.- L. Chin, C.- Y. Li, Y.- M Lai, T. Chen, T.- Y. Sun, J.- C. Lin, CWD2GAN: Generative Adversarial Network of Chronic Wound Depth Detection for Predicting Chronic Wound Depth, International Journal of Pattern Recognition and Artificial Intelligence, vol. 37, no. 3, (2023).
[3]J. S. Mervis and T. J. Phillips, Pressure ulcers: Prevention and management, J. Am. Acad. Dermatol. 81(4) (2019) 893–902.
[4]C. C. V. Stülpnagel, N. D. Silva, M. Augustin, C. V. Montfrans, C. Fife, A.- M. Fagerdahl, A. Gamus, T. M. Klein, C. Blome and R. Sommer, Assessing the quality of life of people with chronic wounds by using the cross-culturally valid and revised Wound-QoL ques- tionnaire, Wound Repair Regen. 29(3) (2021) 452–459.
[5]T. Biswas, M. F. Ahmad Fauzi, F. S. Abas, and H. K. R. Nair, Enhanced CNN Based Super pixel Classification for Automated Wound Area Segmentation, IEEE 8th R10 Humanitarian Technology Conference (2020) (R10-HTC).
[6]C. Cui, K. Thurnhofer-Hemsi, R. Soroushmehr, A. Mishra, J. Gryak, E. Domínguez, K. Najarian, and E. López-Rubio, Diabetic Wound Segmentation using Convolutional Neural Networks, Annu Int Conf IEEE Eng Med Biol Soc. (2019) 1002–1005.
[7]R. Zhang, D. Tian, D. Xu, W. Qian, and Y. Yao, A Survey of Wound Image Analysis Using Deep Learning: Classification, Detection, and Segmentation, IEEE Access, vol. 10, (2022) 79502-79515.
[8]D. Marijanović and D. Filko, A systematic overview of recent methods for non-contact chronic wound analysis, Appl. Sci. 10(21) (2020) 7613.
[9]D. Marijanović, E. K. Nyarko and D. Filko, Wound detection by simple feedforward neural network, Electronics. 11(3) (2022) 329.
[10]J. S. Mervis and T. J. Phillips, Pressure ulcers: Prevention and management, J. Am. Acad. Dermatol. 81(4) (2019) 893–902.
[11]J. Redmon, S. Divvala, R. Girshick, A. Farhadi, You Only Look Once: Unified, Real-Time Object Detection, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2016) 779-788.
[12]J. A.- Sánchez, G. V.- Molina, M. E. L.- Valverde and J. M. R.- Bonilla, Severe diabetic foot infections without systemic inflammatory response syndrome: Prospective validation of a new category, Wound Repair Regen. Acoust. Speech. Signal Process. 30 (5) (2022) 553-559.
[13]J. E. Grey, S. Enoch, research fellow and K. G. Harding, Wound assessment, BMJ. 332(7536) (2006) 285–288.
[14]T. I. Oliver and M. Mutluoglu, Diabetic Foot Ulcer, StatPearls Publishing. (2023).
[15]S. Treneska, E. Zdravevski, I. M. Pires, P. Lameski and S. Gievska, GAN-Based Image Colorization for Self-Supervised Visual Feature Learning, Sensors (Basel). 22(4) (2022) 1599.
[16]R. Shetty, H. Sreekar, S. Lamba and A. K. Gupta, A novel and accurate technique of photographic wound measurement, Indian J. Plast. Surg.45(2) (2012) 425–429.
[17]L. Antony, A. S. Thelly, and J. M. Mathew, Evidence-based Clinical Practice Guidelines for Caregivers of Palliative Care Patients on the Prevention of Pressure Ulcer, Indian J Palliat Care. 29(1) (2023) 75–81.