本研究提出一種結合Transformer深度學習模型與石英晶體微天平電子鼻量測尿液訊號的方法，用以非侵入式篩檢大腸直腸癌及腺瘤性息肉，完成的概念證明研究。本研究共納入53名經大腸鏡確診大腸直腸癌受測者、30名腺瘤性息肉受測者與38名健康對照受測者，從每位受測者收集一份尿液樣本，並以電子鼻對每份尿液樣本進行五次揮發性有機化合物量測。每份樣本的五次電子鼻訊號經過演算法判斷後，以多數決的方式決定最後的預測結果，再進行正確率的計算。本研究之演算法流程為電子鼻訊號首先經基線校正與正規化處理，作為演算法輸入以進行特徵萃取，並以Transformer作為核心分類模型進行分類分析。為了驗證Transformer演算法的優越性，本研究比較了機器學習以及深度學習的演算法效能。在機器學習方面，電子鼻訊號首先透過線性判別分析及時序線性判別分析、因果線性判別分析進行特徵工程，再以四種機器學習模型：決策樹、k近鄰演算法、隨機森林、支援向量機進行辨識。在深度學習方面，本研究採用一維卷積神經網路及ResNet作為對照模型比較性能。本研究對於每一個尿液樣本量測五次，然後將量測結果以演算法辨識後，採用多數決策略決定樣本屬於何類；結果顯示整體準確率為93.4%、靈敏度91.4%、特異度96.3%，進一步超越ResNet採用多數決策略後的表現(準確率86.8%)。總和而言，本研究為首篇結合尿液樣本的電子鼻訊號與Transformer演算法進行三類辨識研究(大腸直腸癌、腺瘤性息肉、健康對照受測者)，且受試者樣本數為目前文獻中最多的一篇。研究成果如準確率、靈敏度、特異度在三類分類任務上的表現數值更優於傳統二類(大腸直腸癌、健康對照受測者)研究，上述研究成果凸顯了本研究具有較大的應用範圍、受測者數量多、高準確率的研究優點。

This study presents a method that combines a Transformer-based deep learning model with electronic-nose measurements of urine samples for noninvasive screening of colorectal cancer (CRC) and adenomatous polyps (ADE). Fifty-three patients with colonoscopy-confirmed CRC, thirty patients with ADE, and thirty-eight healthy controls (HC) each provided one urine sample; every sample was measured five times by the electronic nose. For each sample, the five measurement outputs were processed by the algorithm and the final prediction was determined via majority vote before computing overall accuracy. In our analytic pipeline, raw sensor signals undergo baseline removal and normalization, which serve as inputs for feature extraction; the Transformer model then performs the core classification task. To demonstrate the superiority of the Transformer, we compared its performance against both machine-learning and alternative deep-learning approaches. In the machine-learning arm, features were engineered using Linear Discriminant Analysis (LDA), Temporal LDA, and Causal LDA, and then classified with decision trees (DT), k-nearest neighbors (k-NN), random forests (RF), and support vector machines (SVM). In the deep-learning arm, one-dimensional convolutional neural networks (1D CNN) and ResNet were used as benchmark models. Applying majority vote over five repeated measurements per sample yielded an overall accuracy of 93.4%, sensitivity of 91.4%, and specificity of 96.3%, outperforming the ResNet-based majority-vote approach (accuracy 86.8%). To our knowledge, this is the first three-class study—colorectal cancer, adenomatous polyps, and healthy control—using electronic-nose urine measurements with a Transformer classifier, and it features the largest subject cohort reported to date. Moreover, its three-class performance metrics exceed those of conventional two-class studies, underscoring the broader applicability, larger sample size, and high accuracy of our approach.

[1] J. Petimar, S. A. Smith-Warner, B. Rosner, A. T. Chan, E. L. Giovannucci, and F. K. Tabung, “Adherence to the world cancer research Fund/American institute for cancer research 2018 recommendations for cancer prevention and risk of colorectal cancer,” Cancer epidemiology, biomarkers & prevention, vol. 28, no. 9, p. 1469-1479, 2019.
[2] M. Begum et al., “Global colorectal cancer research, 2007‐2021: outputs and funding,” International Journal of Cancer, vol. 152, no. 3, p. 470-479, 2023.
[3] T. A. C. S. m. a. e. c. team, “Key statistics for colorectal cancer,” American Cancer Society, https://www.cancer.org/cancer/types/colon-rectal-cancer/about/key-statistics.html, 2025 (accessed May. 2025).
[4] P. Rawla, T. Sunkara, and A. Barsouk, “Epidemiology of colorectal cancer: incidence, mortality, survival, and risk factors,” Gastroenterology Review/Przegląd Gastroenterologiczny, vol. 14, no. 2, p. 89-103, 2019.
[5] D. K. Rex et al., “Quality indicators for colonoscopy,” Official journal of the American College of Gastroenterology| ACG, vol. 110, no. 1, p. 72-90, 2015.
[6] J.-H. Yeh et al., “Performance of the fecal immunochemical test in detecting advanced colorectal neoplasms and colorectal cancers in people aged 40–49 years: a systematic review and meta-analysis,” Cancers, vol. 15, no. 11, p. 3006, 2023.
[7] J. Meklin, K. Syrjänen, and M. Eskelinen, “Colorectal cancer screening with traditional and new-generation fecal immunochemical tests: a critical review of fecal occult blood tests,” Anticancer research, vol. 40, no. 2, p. 575-581, 2020.
[8] J. M. Carethers, “Fecal DNA testing for colorectal cancer screening,” Annual Review of Medicine, vol. 71, no. 1, p. 59-69, 2020.
[9] J. R. Bailey, A. Aggarwal, and T. F. Imperiale, “Colorectal cancer screening: stool DNA and other noninvasive modalities,” Gut and Liver, vol. 10, no. 2, p. 204, 2016.
[10] N. D'Souza, A. Brzezicki, and M. Abulafi, “Faecal immunochemical testing in general practice,” (in eng), Br J Gen Pract, vol. 69, no. 679, p. 60-61, Feb 2019, doi: 10.3399/bjgp19X700853.
[11] F. Monedeiro, M. Monedeiro-Milanowski, T. Ligor, and B. Buszewski, “A review of GC-based analysis of non-invasive biomarkers of colorectal cancer and related pathways,” Journal of Clinical Medicine, vol. 9, no. 10, p. 3191, 2020. [Online]. Available: https://www.mdpi.com/2077-0383/9/10/3191.
[12] Y. Qiu et al., “Urinary metabonomic study on colorectal cancer,” Journal of Proteome Research, vol. 9, no. 3, p. 1627-1634, 2010/03/05 2010, doi: 10.1021/pr901081y.
[13] D. B. Liesenfeld et al., “Changes in urinary metabolic profiles of colorectal cancer patients enrolled in a prospective cohort study (ColoCare),” Metabolomics, vol. 11, no. 4, p. 998-1012, 2015/08/01 2015, doi: 10.1007/s11306-014-0758-3.
[14] R. P. Arasaradnam et al., “Detection of colorectal cancer (CRC) by urinary volatile organic compound analysis,” PloS one, vol. 9, no. 9, p. e108750, 2014.
[15] K. Maštovská and S. J. Lehotay, “Practical approaches to fast gas chromatography–mass spectrometry,” Journal of Chromatography A, vol. 1000, no. 1-2, p. 153-180, 2003.
[16] A. D. Wilson, “Application of electronic-nose technologies and VOC-biomarkers for the noninvasive early diagnosis of gastrointestinal diseases,” Sensors, vol. 18, no. 8, p. 2613, 2018. [Online]. Available: https://www.mdpi.com/1424-8220/18/8/2613.
[17] H. Tyagi, E. Daulton, A. S. Bannaga, R. P. Arasaradnam, and J. A. Covington, “Non-invasive detection and staging of colorectal cancer using a portable electronic nose,” Sensors, vol. 21, no. 16, p. 5440, 2021.
[18] E. Westenbrink et al., “Development and application of a new electronic nose instrument for the detection of colorectal cancer,” Biosensors and Bioelectronics, vol. 67, p. 733-738, 2015.
[19] P. Gupta et al., “Prediction of colon cancer stages and survival period with machine learning approach,” Cancers, vol. 11, no. 12, p. 2007, 2019. [Online]. Available: https://www.mdpi.com/2072-6694/11/12/2007.
[20] M. Min, S. Su, W. He, Y. Bi, Z. Ma, and Y. Liu, “Computer-aided diagnosis of colorectal polyps using linked color imaging colonoscopy to predict histology,” Scientific Reports, vol. 9, no. 1, p. 2881, 2019/02/27 2019, doi: 10.1038/s41598-019-39416-7.
[21] D. Sarwinda, R. H. Paradisa, A. Bustamam, and P. Anggia, “Deep learning in image classification using residual network (ResNet) variants for detection of colorectal cancer,” Procedia Computer Science, vol. 179, p. 423-431, 2021/01/01/ 2021, doi: https://doi.org/10.1016/j.procs.2021.01.025.
[22] A. Vaswani et al., “Attention is all you need,” Advances in neural information processing systems, vol. 30, p. 4-6, 2017.
[23] F. Pedregosa et al., “Scikit-learn: machine learning in python,” the Journal of machine Learning research, vol. 12, p. 2825-2830, 2011.
[24] B. Deepa and K. Ramesh, “Epileptic seizure detection using deep learning through min max scaler normalization,” Int. J. Health Sci, vol. 6, p. 10981-10996, 2022.
[25] S. Balakrishnama and A. Ganapathiraju, “Linear discriminant analysis-a brief tutorial,” Institute for Signal and information Processing, vol. 18, no. 1998, p. 1-8, 1998.
[26] Š.-B. Jūratė and D. Kȩstutis, “Linear discriminant analysis of multivariate spatial-temporal regressions,” Scandinavian Journal of Statistics, vol. 32, no. 2, p. 281-294, 2005. [Online]. Available: http://www.jstor.org/stable/4616878.
[27] K. Yoshida, M. Inui, T. Yairi, K. Machida, M. Shioya, and Y. Masukawa, “Identification of causal variables for building energy fault detection by semi-supervised LDA and decision boundary analysis,” in 2008 IEEE International Conference on Data Mining Workshops, 2008: IEEE, pp. 164-173.
[28] Z. Li, F. Liu, W. Yang, S. Peng, and J. Zhou, “A survey of convolutional neural networks: analysis, applications, and prospects,” IEEE transactions on neural networks and learning systems, vol. 33, no. 12, p. 6999-7019, 2021.
[29] S. Targ, D. Almeida, and K. Lyman, “Resnet in resnet: generalizing residual architectures,” arXiv preprint arXiv:1603.08029, 2016.
[30] L. Ying, “Decision tree methods: applications for classification and prediction,” Shanghai archives of psychiatry, vol. 27, no. 2, p. 130, 2015.
[31] G. Guo, H. Wang, D. Bell, Y. Bi, and K. Greer, “KNN model-based approach in classification,” in On The Move to Meaningful Internet Systems 2003: CoopIS, DOA, and ODBASE: OTM Confederated International Conferences, CoopIS, DOA, and ODBASE 2003, Catania, Sicily, Italy, November 3-7, 2003. Proceedings, 2003: Springer, pp. 986-996.
[32] L. Breiman, “Random forests,” Machine learning, vol. 45, p. 5-32, 2001.
[33] V. Jakkula, “Tutorial on support vector machine (svm),” School of EECS, Washington State University, vol. 37, no. 2.5, p. 3, 2006.
[34] M. P. Pauli, C. Pohl, and M. Golz, “Balanced leave-one-subject-out cross-validation for microsleep classification,” Current Directions in Biomedical Engineering, vol. 7, no. 2, p. 147-150, 2021.
[35] T.-T. Wong and P.-Y. Yeh, “Reliable accuracy estimates from k-fold cross validation,” IEEE Transactions on Knowledge and Data Engineering, vol. 32, no. 8, p. 1586-1594, 2019.
[36] F. Fauzi, A. Rianjanu, I. Santoso, and K. Triyana, “Gas and humidity sensing with quartz crystal microbalance (QCM) coated with graphene-based materials–A mini review,” Sensors and Actuators A: Physical, vol. 330, p. 112837, 2021.
[37] K. E. van Keulen, M. E. Jansen, R. W. Schrauwen, J. J. Kolkman, and P. D. Siersema, “Volatile organic compounds in breath can serve as a non‐invasive diagnostic biomarker for the detection of advanced adenomas and colorectal cancer,” Alimentary pharmacology & therapeutics, vol. 51, no. 3, p. 334-346, 2020.