本研究提出了一種有效的人工智慧模型 (Artificial Intelligence model, AI model) 用於偵測動靜脈廔管 (Arteriovenous Fistula, AVF) 狹窄。 由於我們收集到的血管聲音是透過便宜的非侵入式收音設備, 價格不到 30 美元, 可以讓一般民眾在家也可以自我檢測血管是否狹窄。 本研究提出的 AI 模型 (Concatenated model) 是整合兩種不同的輸入特徵 (例如: 血管聲音的智慧型轉換) 以及相對應的分類模型 (Resnet 50 以及 ANN) 。 本研究的兩種不同輸入特徵分別是照片型資料以及數值型資料。 照片型資料是將音檔透過短時距傅立葉 (Short-Time-Fourier-Transform, STFT) 轉換後得出的頻譜圖。 數值型資料是去計算音檔的樣本熵 (Sample Entropy)。 同時透過實驗設計 (Design of Experiment) 找出本研究提出的 AI 模型最佳的參數組合。 本研究提出的AI 模型在統計上的績效、創新性以及完整的架構都優於現有的動靜脈廔管阻塞研究。 此外本研究提出許多明確的圖表讓人可以簡單清楚的了解本研究所使用到的方法。 根據本研究提出的 AI 模型以及便宜的收音設備, 我們相信讓一般民眾在家中自我檢測 AVF 狹窄是指日可待的。

This paper proposes an effective artificial intelligence (AI) model used to automatically detect arteriovenous fistula (AVF) stenosis in home-care use. The vascular sounds are collected from inexpensive non-invasive audio recordings, priced at less than USD 30 each. The proposed AI model is a concatenated model based on two new input features (as functions of vascular sounds) and two associated
classification models (ResNet50 and ANN). The two new input features are “image feature” via short-time-Fourier-transform (STFT) and “texture feature” via sample entropy. Also design of experiment (DOE) is applied to obtain optimal hyper-parameters for the proposed model. The proposed AI model outperforms all or most existing AVF stenosis detection models in terms of the statistical effectiveness, creativity, and thorough analysis. Moreover, we propose insightful graphs to make it easy for others to understand the essence underlining the associated methods. Based on the proposed AI model and the inexpensive audio recording device, the prospects of “developing a reliable method for detecting AVF stenosis in home-care use” are promising.

[1] Y. M. Akay, M. Akay, W. Welkowitz, J. L. Semmlow, and J. B. Kostis, “Noninvasive acoustical detection of coronary artery disease: A comparative study of signal processing methods,” IEEE Transactions on Biomedical Engineering, vol. 40, no. 6, pp. 571–578, 1993.
[2] M. Allon and M. L. Robbin, “Increasing arteriovenous fistulas in hemodialysis patients: Problems and solutions,” Kidney international, vol. 62, no. 4, pp. 1109–1124, 2002.
[3] S. Chin, B. Panda, M. Damaser, and S. Majerus, “Stenosis characterization and identification for dialysis vascular access,” 2018 IEEE Signal Processing in Medicine and Biology Symposium (SPMB), IEEE, 2018, pp. 1–5.
[4] Grochowina, Marcin and Leniowska, Lucyna, “The new method of the selection of features for the k-nn classifier in the arteriovenous fistula state estimation,” 2016 Federated Conference on Computer Science and Information Systems (FedCSIS), IEEE, 2016, pp. 281–285.
[5] M. Grochowina, “Design and implementation of a device supporting automatic diagnosis of arteriovenous fistula,” 2018 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA), IEEE, 2018, pp. 186–190.
[6] M. Grochowina, L. Leniowska, and A. Gala-Błądzińska, “The prototype device for non-invasive diagnosis of arteriovenous fistula condition using machine learning methods,” Scientific Reports, vol. 10, no. 1, pp. 1–11, 2020.
[7] M. Grochowina, K. Wojnar, and L. Leniowska, “An application of wavelet transform for non-invasive evaluation of arteriovenous fistula state,” 2019 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA), IEEE, 2019, pp. 214–217.
[8] D. Higashi, K. Nishijima, K. Furuya, K. Tanaka, and S. Shin, “Classification of shunt murmurs for diagnosis of arteriovenous fistula stenosis,” 2018 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), IEEE, 2018, pp. 665–669.
[9] N. E. Huang, Z. Shen, S. R. Long, M. C. Wu, H. H. Shih, Q. Zheng, N.-C. Yen, C. C. Tung, and H. H. Liu, “The empirical mode decomposition and the hilbert spectrum for nonlinear and non-stationary time series analysis,” Proceedings of the Royal Society of London. Series A: mathematical, physical and engineering sciences, vol. 454, no. 1971, pp. 903–995, 1998.
[10] Q. Ji, J. Huang, W. He, and Y. Sun, “Optimized deep convolutional neural networks for identification of macular diseases from optical coherence tomography images,” Algorithms, vol. 12, no. 3, p. 51, 2019.
[11] H. Kato, M. Kiryu, Y. Suzuki, O. Sakata, and M. Fukasawa, “Improvement of artificial auscultation on hemodialysis stenosis by the estimate of stenosis site and the hierarchical categorization of learning data,” IEICE TRANSACTIONS on Information and Systems, vol. 100, no. 1, pp. 175–180, 2017.
[12] K. Konner, B. Nonnast-Daniel, and E. Ritz, “The arteriovenous fistula,” Journal of the American Society of Nephrology, vol. 14, no. 6, pp. 1669–1680, 2003.
[13] S. J. Majerus, T. Knauss, S. Mandal, G. Vince, and M. S. Damaser, “Bruit enhancing phonoangiogram filter using sub-band autoregressive linear predictive coding,” 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), IEEE, 2018, pp. 1416–1419.
[14] K. Ota, Y. Nishiura, S. Ishihara, H. Adachi, T. Yamamoto, and T. Hamano, “Evaluation of hemodialysis arteriovenous bruit by deep learning,” Sensors, vol. 20, no. 17, p. 4852, 2020.
[15] B. Panda, S. Chin, S. Mandal, and S. Majerus, “Skin-coupled pvdf microphones for noninvasive vascular blood sound monitoring,” 2018 IEEE Signal Processing in Medicine and Biology Symposium (SPMB), IEEE, 2018, pp. 1–4.
[16] B. Panda, S. Mandal, and S. J. Majerus, “Flexible, skin coupled microphone array for point of care vascular access monitoring,” IEEE transactions on biomedical circuits and systems, vol. 13, no. 6, pp. 1494–1505, 2019.
[17] W. T. Song, C. Lai, and Y.-Z. Su, “A statistical robust glaucoma detection framework combining retinex, cnn, and doe using fundus images,” IEEE Access, vol. 9, pp. 103772–103783, 2021.
[18] USRDS, “United states renal data system,” USRDS, 2021.
[19] Wang, H. Y. and Wu, C. H. and Chen, C. Y. & Lin, B. S., “Novel noninvasive approach for detecting arteriovenous fistula stenosis,” IEEE Transactions on Biomedical Engineering, vol. 61, no. 6, pp. 1851–1857, 2014.
[20] Y.-Y. Wang, C.-D. Kan, W.-L. Chen, and K.-S. Cheng, “A rapid assessment method on fistula stenosis staging for hemodialysis patients,” World Congress on Medical Physics and Biomedical Engineering 2018, Springer, 2019, pp. 475–479.
[21] C. Wei-Ling, L. Chia-Hung, and K. Chung-Dann, “Assessment of arteriovenous shunt pathway function and hypervolemia for hemodialysis patients by using integrated rapid screening system,” Advances in Technology Innovation, vol. 2, no. 2, p. 46, 2017.
[22] A. Zeiler, R. Faltermeier, I. R. Keck, A. M. Tomé, C. G. Puntonet, and E. W. Lang, “Empirical mode decomposition-an introduction,” The 2010 International Joint Conference on Neural Networks (IJCNN), IEEE, 2010, pp. 1–8.