隨著飲食習慣的改變，衍生出來許多膽胰管阻塞的疾病，造成膽汁無法排出，因此微創的膽汁引流治療方式也越來越多，並透過X光影像或EUS(Endoscopic Ultrasound)來做精準治療。其中X光影像下的治療為較常用的方式，但基於X光影像對比度低、結構重疊複雜，手術過程中，醫師不容易找到導絲，故醫師需要在有限的時間內，使用醫療器械在膽胰管內完成膽汁引流的治療，變得更具有挑戰性。
本研究於ERCP(Endoscopic Retrograde Cholangiopancreatography)手術的模擬器下，進行模型訓練與測試，並提出以MobileNetV2的encoder結合U-Net的decoder體現出很好的效能，來實現導絲的影像分割，再以PCA(Principal Component Analysis)取得特徵向量及特徵值，來找到導絲的方向及頂點位置，並加上明顯的標記，此研究讓醫師於手術過程中能快速看到所控制的導絲行走到膽胰管患部位置，進而減少醫師眼睛疲勞所造成的手術失誤及避免膽胰管穿孔的發生;並且在嵌入式系統下能執行輕量型的神經網路系統，此ERCP練習的模擬器變得更便捷，同時期待此模擬器可以普及化到每個醫院，讓更多的醫師能隨時隨地都可以做練習，縮短ERCP學習曲線，提升手術技能，降地手術的風險。

With the evolution of dietary habits, there has been a notable increase in biliary and pancreatic duct obstructions, leading to impaired bile drainage. Consequently, minimally invasive biliary drainage procedures have become more prevalent and are often guided by imaging techniques such as X-ray fluoroscopy or endoscopic ultrasound (EUS). Among these, X-ray fluoroscopy is the more commonly used modality; however, due to the low contrast and complex overlapping anatomical structures in X-ray images, it remains challenging for physicians to accurately locate the guidewire during the procedure. As a result, it becomes increasingly difficult to complete bile drainage within the limited operative window, making the procedure more technically demanding.
In this study, we developed and validated a model using an ERCP (Endoscopic Retrograde Cholangiopancreatography) simulator for both training and testing purposes. We propose a deep learning framework that integrates a MobileNetV2-based encoder with a U-Net decoder, demonstrating promising performance in guidewire segmentation. Additionally, Principal Component Analysis (PCA) is employed to extract feature vectors and eigenvalues to determine the orientation and tip location of the guidewire. To enhance intraoperative visibility, a prominent visual marker is added at the guidewire tip. This system enables surgeons to quickly identify the position of the guidewire within the diseased region of the biliary-pancreatic duct, reducing visual fatigue and minimizing surgical errors and the risk of ductal perforation.
Furthermore, the lightweight neural network is optimized for execution on embedded systems, making the ERCP simulator more portable and accessible. This advancement holds the potential to promote widespread adoption in hospitals, allowing physicians to engage in practice anytime and anywhere. Ultimately, this approach aims to shorten the ERCP learning curve, enhance surgical proficiency, and lower the risk associated with clinical procedures.

[1] A, Vandini., et al., “Robust guidewire tracking under large deformations combining segment-like features,” Medical Image Analysis, Vol. 38, pp. 150-164, 2017.
[2] R, Yamashita., et al., “Convolutional neural networks: An overview and application in radiology,” Insights Imaging, Vol. 9, pp. 611-629, 2018.
[3] O, Ronneberger., P, Fischer., T, Brox., “U-Net: Convolutional networks for biomedical image segmentation,” Medical Image Computing and Computer-Assisted Intervention, pp. 234–241, 2015.
[4] N, Siddique., et al., “U-Net and its variants for medical image segmentation: heory and applications,” IEEE Access, Vol. 9, pp. 82031, 2021.
[5] M. Sandler., et al., “MobileNetV2: Inverted residuals and linear bottlenecks,” 2018 IEEE Conference on Computer Vision and Pattern Recognition,”pp. 4510–4520, 2018.
[6] Priyanka., D, Kumar., Feature extraction and selection of kidney ultrasound images using GLCM and PCA,” Procedia Computer Science, Vol. 167, pp. 1722-1731, 2020.
[7] J, Dumonceau., et al., “ERCP-related adverse events: European society of gastrointestinal endoscopy guideline,” Endoscopy, Vol. 52, pp. 127-149, 2020.
[8] B, Zhang., et al., Real-time guidewire tracking and segmentation in intraoperative x-ray,”SPIE Medical Imaging, Vol. 12034, 2022.
[9] J, Lee., J, Park., Y, Lee., Toward efficient cancer detection on mobile devices,” IEEE Access, Vol. 13, pp. 34613-34626, 2025.
[10] H, Ma., et al.,“Dynamic coronary roadmapping via catheter tip tracking in X-ray fluoroscopy with deep learning based bayesian filtering,” Medical Image Anaalysis, Vol. 61, 2020.
[11] A, Ramadani., et al., “A survey of catheter tracking concepts and methodologies,”Medical Image Anlysis, Vol. 82, 2022.
[12] A, P, P, Ayme., et al., “Advancements in minimally invasive surgical techniques: A comprehensive review,” Salud, Ciencia y Tecnologia, Vol. 04, 2024.
[13] B, C, Bag., H, K, Maity., C, Koley., “U-Net mobileNetV2: Medical image segmentation using deep neural network,” Journal of Mechanics of Continua and Mathematical Sciences, Vol.18, pp. 21-29, 2023.
[14] S, Narayanan, S., S, K, Bag., S, Singh., “Enhanced diabetic retinopathyanalysis: U-Net-based lesion segmentation coupled with MobileNetV2 for feature extraction and efficientnetb0 for classification,” 2024 Third International Conference on Intelligent Techniques in Control, Optimization and Signal Processing, pp. 1-5, 2024.
[15] A, Jenynof., T, Ahmad., “Image to image steganography using U-Net architecture with mobileNet convolutional neural network,” 14th International Conference on Computing Communication and Networking Technologies, pp. 1-7, 2023.
[16] Z, Riaz., et al., “Lung tumor image segmentation from computer tomography images using MobileNetV2 and transfer learning,” Bioengineering, Vol. 10, pp. 981, 2023.
[17] N, A, Qureshi., et al., “Application of Principal Component Analysis (PCA) to medical data,” Indian Journal of Science and Technology, Vol. 10, No. 20, 2017.
[18] J,W, Leung., W, Lee., R, Wilson., “Comparison of accessory performance using a novel ERCP mechanical simulator,” Endoscopy, Vol. 40, pp. 983-988, 2008.
[19] P, A , Testoni1., et al., “Papillary cannulation and sphincterotomy techniques at ERCP: European Society of Gastrointestinal Endoscopy clinical guideline,” Endoscopy, Vol. 48, pp. 657–683, 2016.
[20] W, C, Liao., et al., “International consensus recommendations for difficult biliary access,” Gastrointestinal Endoscopy, Vol. 85, No. 2, 2017.
[21] M, Al-Mansour., et al., “Surgeon-performed endoscopic retrograde cholangiopancreatography. Outcomes of 2392 procedures at two tertiary care centers,”Surg Endosc, Vol. 32, No. 6, pp. 2871–2876, 2018.
[22] Y, Kim., et al., “Simulator-based training method in gastrointestinal endoscopy training and currently available simulators,” Clin Endosc, Vol. 56, pp. 1-13, 2023.
[23] J, Bittner IV., et al., “Face and construct validity of a computer-based virtual reality simulator for ERCP,” Gastrointestinal Endoscopy, Vol. 71, No. 2, 2010.
[24] K, Matthes., J, Cohen., “The neo-papilla: a new modification of porcine ex vivo simulators for ERCP training (with videos) ,” Gastrointestinal Endoscopy, Vol. 64, No. 4, 2006.
[25]R, Hubrecht., E, Carter., “The 3Rs and humane experimental technique: Implementing change,” Animals, Vol. 09, pp. 754, 2019.
[26] V, Chandrasekhara., et al., “Adverse events associated with ERCP,” Gastrointestinal Endoscopy, Vol. 85, No. 1, 2017.
[27] K, O’Donoghue., H, A, Jaeger., P, Cantillon-Murphy., “A radiolucent electromagnetic tracking system for use with intraoperative X-ray imaging,” Sensors, Vol. 21, pp. 3357, 2021.
[28] S, Y, Han., et al., “Efficacy of a newly developed guidewire for selective biliary cannulation: A multicenter randomized controlled trial,” Journal of clinical medicine, Vol. 12, pp. 3440, 2023.
[29] D, Satoh., H, Matsukawa., S, Shiozaki.,“The optimal type and management of biliary drainage in patients with obstructive jaundice who undergo pancreaticoduodenectomy,”Jouronal of in vivo, Vol. 36, pp. 391-397, 2022.
[30] M, Okuno., et al., “Endoscopic nasobiliary drainage tube placement through a periampullary perforation for management of intestinal leak and necrotizing pancreatitis,” VIDEOGIE, Vol. 8, No. 2, 2023.
[31] T, H, Hung., et al.,“Endoscopic papillary balloon dilation decreases the risk of bleeding in cirrhotic patients compared with endoscopic biliary sphincterotomy, Medicine, Vol. 98, Issue. 30, 2019.
[32] S, Masuda., K, Koizumi., K, Shionoya., “Comprehensive review on small common bile duct stones,” World Journal of Gastroenterology, Vol. 29, Issue. 13, 2023.
[33] X, Liu., et al., “Deep learning-based target tracking with X-ray images for radiotherapy: a narrative review,” Quantitative Imaging in Medicine and Surgery, Vol. 14, No. 3, 2024.
[34] B, Zhang ., et al., “Real-time guidewire tracking and segmentation in intraoperative X-ray,” arXiv, 2404. 08805v1, 2024.
[35] A, Ramadani ., et al., “A survey of catheter tracking concepts and methodologies,”Medical Image Analysis, Vol. 82, 2022.
[36] A, Lapušinskij., et al., “The application of hough transform and canny edge detector methods for the visual detection of cumuliform clouds,” Sensors, Vol. 21, p. 5821, 2021.
[37] Q, Zhang., et al., “An optimized optical-flow-based method for quantitative tracking of ultrasound-guided right diaphragm deformation,” BMC Medical Imaging, Vol. 23, p. 108, 2023.
[38] J, Kummert., et al., “Efficient reject options for particle filter object tracking in medical applications,” Sensors, Vol. 21, p. 2114, 2021.
[39] A, Arjas., et al., “Neural network kalman filtering for 3-D object tracking from linear array ultrasound data,” IEEE Transactions On Ultrasonics, Ferroelectrics, And Frequency Control, Vol. 69, No.5, 2022.
[40] Y, Ruan., et al., “Principal component analysis of photoplethysmography signals for improved gesture recognition,” Sec. Neuroprosthetics, Vol. 16, 2022.
[41] R, S, Sandhya Devi., V, R, Vijay Kumar., P. Sivakumar., “A review of image classification and object detection on machine learning and deep learning techniques,” 5th International Conference on Electronics, Communication and Aerospace Technology, pp. 1-8, 2022.
[42] T, Kerala., “Different approaches for segmentation,” Proceedings of the Fifth International Conference on Communication and Electronics Systems, pp. 938-943, 2020.
[43] A, B, Amioud., et al., “Object Detection Using Deep Learning, CNNs and Vision Transformers: A Review,” IEEE Access, Vol, 11, pp. 35478-35516, 2023.
[44] J, M, Molina., et al., “Advances in instance segmentation: Technologies, metrics and applications in computer vision,” Neurocomputing, Vol, 625, 2025.
[45] Y, Nie., Y, Chen., J, Guo., “An improved CNN model in image classification application on water turbidity,” Scientific Reports, Vol. 15, Article number:11264, 2025.
[46] Y, Wang., et al., “The influence of the activation function in a convolution neural network model of facial expression recognition,” Applied sciences, Vol. 10, pp. 1897, 2020.
[47] A, H, Nizamani., Z, Chen., A, A, Nizamani., “Feature‑enhanced fusion of U‑Net based improved brain tumor images segmentation,” Journal of Cloud Computing, Vol. 12, Issue. 170, 2023.
[48] A, G, Howard., et al., “MobileNets: Efficient convolutional neural networks for mobile vision applications,” Google Inc., 2017.
[49] A, Kanadath., J, A, A, Jothi., S, Urolagin., “Histopathology image segmentation using MobileNetV2 based U-Net model,” International Conference on Intelligent Technologies, pp. 1-8, 2021.
[50] B, C, Bag., H, K, Maity., C, Koley., “U-Net MobilenetV2: Medical image segmentation using deep neural network,” Journal of Mechanics of Continua and Mathematical Sciences, Vol. 18, pp. 21-99, 2023.
[51] A, Howard., et al., “Searching for MobileNetV3,” arXvix:1905.02244v5, 2019.
[52] D, Qin., et al., “MobileNetV4: Universal Models for the Mobile,” arXvix:2404. 10518v2, 1905.02244v5, 2024.
[53] J, Wang., et al., “PCA-U-Net based breast cancer nest segmentation from microarray hyperspectral images,” Fundamental Research, Vol. 1, Issue. 5, pp. 631-640, 2021.
[54] Y, Peng., M, Sonka., D,Z, Chen., “U-Net V2: Rethinking The Skip connections of U-Net for medical image segmentation,” arXiv:2311.17791v2, 2024.
[55] I, Krispin-Avraham., R, Orfaig., B, Z, Bobrovsky., “Real-time 3D object detection using InnovizOne LiDAR and low-power Hailo-8 AI accelerator, arXiv:2412.05594v1, 2024.
[56] P, Dini., L, Diana., A, Elhanashi., “Overview of AI-models and tools in embedded IIoT applications,” Electronics, Vol. 13, Issue. 12, 2024.
[57] G, Casiez, N, Roussel, D, Vogel., “One Euro filter: A simple speed-based low-pass filter for noisy input in interactive systems,” CHI '12: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 2527-2530, 2012.
[58] B, A, Awaluddin., C, T, Chao., J, S, Chiou., “A hybrid image augmentation technique for use and environment-Independent hand gesture recognition based on deep learning,”Mathematics, Vol. 12, Issue. 1393, 2024.
[59] L, Zhang., Z, Liu., L, Zhang., “Technical note: Generalizable and promptable artificial intelligence model to augment clinical delineation in radiation oncology,” Med Phys, Vol. 51, Issue. 3, pp. 2187-2199, 2024.
[60] N, Ratcliffe., R, Newport., “The effect of visual, spatial and temporal manipulations on embodiment and action,” Front. Hum. Neurosci, Vol. 11, 2017.
[61] S, Sivamani., S, I, Chon., D, Y, Choi., “Importance of adaptive photometric augmentation for different convolutional neural network,” Computers, Materials & Continua, Vol. 72, No. 3, 2022.
[62] D, P, Kingma., J, L, Ba., “Adam: A method for stochastic optimization, arXiv:1412.6980v9, 2017.
[63] R, Azad., et al., “ Loss function in the era of semantic segmentation: A survey and outlook ,” arXiv:2312.05391v1.
[64] A, U, Ruby., et al., “ Binary cross entropy with deep learning technique for image classification,” International Journal of Advanced Trends in Computer Science and Engineering, Vol. 9, No.4, 2020.
[65] R, Zhao., et al., “ Rethinking dice loss for medical image segmentation,” 2020 IEEE International Conference on Data Mining, pp. 851-860, 2020.
[66] Y, Mu., et al., “ The combined focal loss and dice loss function improves the segmentation of beta-sheets in medium-resolution cryo-electron-microscopy density maps,” Bioinformatics Advances, Vol. 4, Issue 1, 2024.
[67] D, Kumar K., et al., “Early melanoma detection and classification using CNN and confusion matrix analysis,” International Conference on System, Computation, Automation and Networking, pp. 1-5, 2024.
[68] A, Wibowo., et al., “Lightweight encoder-decoder model for automatic skin lesion segmentation,” Informatics in Medicine Unlocked, Vol. 25, pp. 100640, 2021.
[69] L, Wei., Z, Ma., C, Yang., “Advances in the neural network quantization: A comprehensive review,” Applied Sciences, Vol. 14, Issue. 17, 2024.
[70] G, N, Cahyo., C, Anam., K, Adi., “Effect of noise on the robustness of MobileNetV2+U-Net semantic segmentation model for MRI images,” IJSRST, Vol. 10, No. 6, 2023.
[71] F, Valeri., M, Bartolucci., E, Cantoni., “UNet and MobileNet CNN-based model observers for CT protocol optimization: Comparative performance evaluation by means of phantom CT images,” Journal of Medical Imaging, Vol. 10, Issue. S1, 2023.
[72] K, Fu., et al., “Text detection for natural scene based on MobileNet V2 and U-Net,” ICMA, pp. 1560-1564, 2019.
[73] J, H, Hwang., B, Cho., D, H, Choi “Feature map analysis of neural networks for the application of vacant parking slot detection,” Applied Sciences, Vol. 13, Issue. 10342, 2023.
[74] L, Qu., X, Zhu., B, Li., “Progressive feature fusion and refinement network for substation rotating object detection,” SPIES, pp. 2356-2360, 2022.
[75] J, Knapheide., B, Stabernack., M, Kuhnke., “A high throughput MobileNetV2 FPGA implementation based on a flexible architecture for depthwise separable convolution,”FPL, pp. 277-283, 2020.
[76] Y, Yousfi., et al., “ImageNet pre-trained CNNs for JPEG steganalysis,” WIFS‘2020, pp. 1-6, 2020.
[77] X, X, Yin., L, Sun., Y, Fu., “Retracted U-Net-based medical image segmentation,”Journal of Healthcare Engineering, Vol. 2022, Issue. 1, 2022.
[78] Y, Furusho., K,i Ikeda., “Theoretical analysis of skip connections and batch normalization from generalization and optimization perspectives,” APSIPA Transactions on Signal and Information Processing, Vol. 9, Issue. 1, 2020.
[79] S, Aslam., T, F, Rabie., “Principal component analysis in image classification : A review,” 2023 ASET, pp. 1-7, 2023.
[80] H, R, Frost., “Eigenvectors from eigenvalues sparse principal component analysis,”Journal of Computational and Graphical Statistics, Vol. 31, No.2, pp. 486-501, 2022.
[81] Y, Deng., F, Gao., H, Chen., “Angle estimation for knee joint movement based on PCA -relm algorithm,” Symmetry, Vol. 12, Issue. 1, 2020.
[82] C, Busu., M, Busu., “An application of the Kalman filter recursive algorithm to estimate the Gaussian errors by minimizing the symmetric loss function, Symmetry., Vol. 13, Issue. 2, 2021.
[83] G, Prethija., J, Katiravan., “Delving into transfer learning within U-Net for refined retinal vessel segmentation: An extensive hyperparameter analysis,” Photodiagnosis and Photodynamic Therapy, Vol. 53, 2025.
[84] C, I, Kwon., D, H, Koh., T, J, Song., “Technical reports of endoscopic retrograde cholangiopancreatography guidewires on the basis of physical properties,”Clinical Endoscopy, Vol. 53, Issue.1, pp. 65-72, 2025.
[85] C, C, Chong., P, Chiu., T, Tan., “Correlation of CBD/CHD angulation with recurrent cholangitis in patients treated with ERCP,” Endoscopy International Open, Vol. 04, Issue. 01, pp. E62-E67, 2016.
[86] D, Alfieri., C, Delogu., S, Mazza., “The role and appropriate selection of guidewires in biliopancreatic endoscopy,” Medicina, Vol. 61, Issue. 5.