細菌感染是全球醫療體系面臨的重大挑戰，而能否快速且準確地鑑別感染源，對臨床診斷與治療決策具有關鍵影響。針對此需求，本研究提出一套基於多任務學習（Multi-Task Learning, MTL）架構的電子鼻分類模型，用以分類細菌代謝產生之揮發性有機化合物（Volatile Organic Compounds, VOCs）相關的氣體訊號，藉此實現多面向的細菌辨識任務。該模型可同時執行三項分類任務：革蘭氏染色分類、菌種鑑別與濃度分級，有效提升推論效率與任務整合性。
相較於過往文獻中常見的傳統機器學習方法，需仰賴人工特徵擷取並搭配分類演算法進行氣味辨識，本文採用深度學習模型可自動學習特徵，簡化處理流程，進一步提升模型使用效率與分類準確度。本模型以一維卷積神經網路為特徵萃取骨幹，並融合通道加權與特徵加權等注意力機制，以加強對關鍵氣味特徵的辨識能力。資料前處理方面，採用基線移除與 最大最小正規化處理，有效減少背景干擾並提升模型穩定性。實驗部分選用五種常見致病菌株（Escherichia coli、Klebsiella pneumoniae、Pseudomonas aeruginosa、Streptococcus agalactiae 與 Staphylococcus aureus），並在六種濃度（102、103、104、105、106、107 CFU/mL）條件下蒐集 VOC 訊號資料。透過 10-fold 交叉驗證進行模型訓練與評估，並與其他文獻中的氣體分類模型進行比較，於三項任務中分別達成 98.83%、98.92% 與 85.83% 的平均準確率，展現本模型在多元任務中的優異表現與潛力。
為進一步驗證系統的實用性與可部署性，本研究將訓練完成的模型經由後訓練量化轉換為 8-bit 精度，並部署至低功耗嵌入式平台 ESP32-S3。結合緊湊型 MEM感測器模組，實現模型在裝置端的即時推論功能，無須仰賴高階運算平台即可完成樣本預測。根據實測結果，量化模型仍能維持高分類準確率，並於 ESP32-S3 上達到平均單次推論時間約 1798 毫秒，證實其於資源受限場域中的可行性。
此外，本研究亦針對本系統與先前開發之 QCM 電子鼻系統進行比較，評估兩者在體積、功耗、硬體成本與推論平台效能上的差異，結果顯示 MEMS 系統在可攜性與應用彈性方面具明顯優勢。綜合而言，本研究成功建構一套具備多任務辨識能力、可部署於嵌入式裝置的電子鼻系統，未來有望應用於臨床感染初篩、智慧醫療設備與行動健康監控等多元場域。

Bacterial infections pose a major challenge to global healthcare systems, where the ability to rapidly and accurately identify the causative pathogen is critical for clinical diagnosis and treatment decisions. To address this need, this study proposes a multi-task learning (MTL)-based electronic nose (E-nose) classification model designed to analyze volatile organic compounds (VOCs) produced by bacterial metabolism. The model simultaneously performs three classification tasks—Gram stain categorization, bacterial species identification, and concentration level estimation—thereby enhancing inference efficiency and integrating multiple diagnostic objectives within a single framework.
Unlike conventional machine learning approaches that rely on handcrafted feature extraction and separate classifiers, our method employs a deep learning model that automatically learns relevant features, simplifying the analysis process and improving both usability and accuracy. The core architecture is based on a one-dimensional convolutional neural network, augmented with channel and feature attention mechanisms to enhance the detection of critical odor characteristics. For data preprocessing, baseline correction and min-max normalization are applied to reduce background noise and improve model robustness.
Experiments were conducted using VOC samples collected from five clinically relevant bacterial strains (Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, Streptococcus agalactiae, and Staphylococcus aureus), across six concentration levels (10² to 10⁷ CFU/mL). The model was evaluated using 10-fold cross-validation, achieving average accuracies of 98.83% (Gram stain), 98.92% (species), and 85.83% (concentration), demonstrating excellent multi-task performance.
To further validate the system's practicality, the trained model was quantized to 8-bit precision via post-training quantization and deployed on a low-power embedded platform (ESP32-S3). Combined with a compact MEMS gas sensor module, real-time on-device inference was achieved without the need for high-performance computing hardware. Evaluation results show that the quantized model maintains high classification accuracy, with an average inference time of approximately 1798ms per sample on the ESP32-S3, confirming its feasibility for deployment in resource-constrained environments.
Additionally, the proposed MEMS-based system was compared to a previously developed QCM-based E-nose system in terms of size, power consumption, hardware cost, and inference performance. Results indicate significant advantages in portability and deployment flexibility for the MEMS platform. In conclusion, this study successfully establishes a deployable, multi-task-capable E-nose system, with promising potential for use in clinical pre-screening, smart healthcare devices, and mobile health monitoring applications.

[1] J. Soni, S. Sinha, and R. Pandey, “Understanding bacterial pathogenicity: A closer look at the journey of harmful microbes,” Frontiers in Microbiology, vol. 15, p. 1370818, 2024.
[2] B. Dyck, M. Unterberg, M. Adamzik, and B. Koos, “The impact of pathogens on sepsis prevalence and outcome,” Pathogens, vol. 13, no. 1, p. 89, 2024.
[3] K. S. Ikuta, L. R. Swetschinski, G. R. Aguilar, F. Sharara, T. Mestrovic, A. P. Gray, N. D. Weaver, E. E. Wool, C. Han, and A. G. Hayoon, “Global mortality associated with 33 bacterial pathogens in 2019: A systematic analysis for the global burden of disease study 2019,” The Lancet, vol. 400, no. 10369, pp. 2221-2248, 2022.
[4] C. Zhang, X. Fu, Y. Liu, H. Zhao, and G. Wang, “Burden of infectious diseases and bacterial antimicrobial resistance in china: A systematic analysis for the global burden of disease study 2019,” The Lancet Regional Health–Western Pacific, vol. 43, 2024.
[5] S. Choi, J. Son, D. K. Oh, J. W. Huh, C.-M. Lim, and S.-B. Hong, “Rapid response system improves sepsis bundle compliances and survival in hospital wards for 10 years,” Journal of Clinical Medicine, vol. 10, no. 18, p. 4244, 2021.
[6] R. Setyawati, A. Astuti, T. P. Utami, S. Adiwijaya, and D. M. Hasyim, “The importance of early detection in disease management,” Journal of World Future Medicine, Health and Nursing, vol. 2, no. 1, pp. 51-63, 2024.
[7] C. Deusenbery, Y. Wang, and A. Shukla, “Recent innovations in bacterial infection detection and treatment,” ACS Infectious Diseases, vol. 7, no. 4, pp. 695-720, 2021.
[8] C. Singhal, J. G. Bruno, A. Kaushal, and T. K. Sharma, “Recent advances and a roadmap to aptamer-based sensors for bloodstream infections,” ACS Applied Bio Materials, vol. 4, no. 5, pp. 3962-3984, 2021.
[9] K. S. Gracias and J. L. McKillip, “A review of conventional detection and enumeration methods for pathogenic bacteria in food,” Canadian journal of microbiology, vol. 50, no. 11, pp. 883-890, 2004.
[10] A. E. Budding, M. Hoogewerf, C. M. Vandenbroucke-Grauls, and P. H. Savelkoul, “Automated broad-range molecular detection of bacteria in clinical samples,” Journal of clinical microbiology, vol. 54, no. 4, pp. 934-943, 2016.
[11] I. C. Santos, Z. L. Hildenbrand, and K. A. Schug, “Applications of maldi-tof ms in environmental microbiology,” Analyst, vol. 141, no. 10, pp. 2827-2837, 2016.
[12] Z. Wang, L. Russon, L. Li, D. C. Roser, and S. R. Long, “Investigation of spectral reproducibility in direct analysis of bacteria proteins by matrix‐assisted laser desorption/ionization time‐of‐flight mass spectrometry,” Rapid Communications in Mass Spectrometry, vol. 12, no. 8, pp. 456-464, 1998.
[13] D. J. Shin, N. Andini, K. Hsieh, S. Yang, and T.-H. Wang, “Emerging analytical techniques for rapid pathogen identification and susceptibility testing,” Annual review of analytical chemistry, vol. 12, no. 1, pp. 41-67, 2019.
[14] R. Patel, “Maldi-tof ms for the diagnosis of infectious diseases,” Clinical chemistry, vol. 61, no. 1, pp. 100-111, 2015.
[15] J. Wang, J. Tao, Z. Dong, and J. Zhu, “Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry: An innovative tool for rapid identification of hylurgus ligniperda, an invasive pest,” Forests, vol. 15, no. 4, p. 729, 2024.
[16] B. Buszewski, M. Kęsy, T. Ligor, and A. Amann, “Human exhaled air analytics: Biomarkers of diseases,” Biomedical chromatography, vol. 21, no. 6, pp. 553-566, 2007.
[17] R. Cabo, K. Kozik, M. Milanowski, S. Hernes, A. Slettan, M. Haugen, S. Ye, R. Blomhoff, and M. A. Mansoor, “A simple high-performance liquid chromatography (hplc) method for the measurement of pyridoxal-5-phosphate and 4-pyridoxic acid in human plasma,” Clinica Chimica Acta, vol. 433, pp. 150-156, 2014.
[18] J. Dawson, K. Green, H. Lazarowicz, P. Cornford, and C. Probert, “Analysis of urinary volatile organic compounds for prostate cancer diagnosis: A systematic review,” BJUI compass, vol. 5, no. 9, pp. 936-947, 2024.
[19] M. Śmiełowska, T. Ligor, W. Kupczyk, J. Szeliga, M. Jackowski, and B. Buszewski, “Screening for volatile biomarkers of colorectal cancer by analyzing breath and fecal samples using thermal desorption combined with gc-ms (td-gc-ms),” Journal of Breath Research, vol. 17, no. 4, p. 047102, 2023.
[20] I.-A. Ratiu, T. Ligor, V. Bocos-Bintintan, J. Szeliga, K. Machała, M. Jackowski, and B. Buszewski, “Gc-ms application in determination of volatile profiles emitted by infected and uninfected human tissue,” Journal of Breath Research, vol. 13, no. 2, p. 026003, 2019.
[21] L. Gao, R. Yang, J. Zhang, M. Sheng, Y. Sun, B. Han, and G. Kai, “Gas chromatography-ion mobility spectrometry for the detection of human disease: A review,” Analytical Methods, 2024.
[22] J. Song, R. Li, R. Yu, Q. Zhu, C. Li, W. He, and J. Liu, “Detection of vocs in exhaled breath for lung cancer diagnosis,” Microchemical Journal, vol. 199, p. 110051, 2024.
[23] H. Yang, M. Wu, X. Shen, Y. Lai, D. Wang, C. Ma, X. Ye, C. Sun, J. Cao, and C. Sun, “A comprehensive review of vocs as a key indicator in food authentication,” eFood, vol. 6, no. 3, p. e70057, 2025.
[24] J.-S. Hung, “Integration of electronic olfaction sensor transient response and machine learning for metabolic odor recognition of different inoculum concentrations and mixed cultures of bacteria in an in vitro cultivation environment,” Master's Thesis, Department of Biomedical Engineering, National Cheng Kung University, Tainan, Taiwan, 2023.
[25] M. Wang and Y. Chen, “Electronic nose and its application in the food industry: A review,” European Food Research and Technology, vol. 250, no. 1, pp. 21-67, 2024.
[26] S. Mor, B. Gunay, M. Zanotti, M. Galvani, S. Pagliara, and L. Sangaletti, “Current opportunities and trends in the gas sensor market: A focus on e-noses and their applications in food industry,” Chemosensors, vol. 13, no. 5, p. 181, 2025.
[27] H. Chen, D. Huo, and J. Zhang, “Gas recognition in e-nose system: A review,” IEEE transactions on biomedical circuits and systems, vol. 16, no. 2, pp. 169-184, 2022.
[28] J. Xu, K. Liu, and C. Zhang, “Electronic nose for volatile organic compounds analysis in rice aging,” Trends in Food Science & Technology, vol. 109, pp. 83-93, 2021.
[29] M. Tomić, M. Šetka, L. Vojkůvka, and S. Vallejos, “Vocs sensing by metal oxides, conductive polymers, and carbon-based materials,” Nanomaterials, vol. 11, no. 2, p. 552, 2021.
[30] B. VA, M. Subramoniam, and L. Mathew, “Noninvasive detection of copd and lung cancer through breath analysis using mos sensor array based e-nose,” Expert review of molecular diagnostics, vol. 21, no. 11, pp. 1223-1233, 2021.
[31] X. Yang, M. Li, X. Ji, J. Chang, Z. Deng, and G. Meng, “Recognition algorithms in e-nose: A review,” IEEE Sensors Journal, vol. 23, no. 18, pp. 20460-20472, 2023.
[32] S. Okur, M. Sarheed, R. Huber, Z. Zhang, L. Heinke, A. Kanbar, C. Wöll, P. Nick, and U. Lemmer, “Identification of mint scents using a qcm based e-nose,” Chemosensors, vol. 9, no. 2, p. 31, 2021.
[33] A. Das and R. Manjunatha, “Modeling of graphene oxide coated qcm sensor for e-nose application,” in Recent trends in materials: Select proceedings of ictmim 2022: Springer, 2022, pp. 179-188.
[34] Y. Yang, C. L. Meng, C. Zhang, and T. Ya, “Research on detection technology of bio-electronic nose based on saw devices,” in 2022 International Conference on Computing, Communication, Perception and Quantum Technology (CCPQT), 2022: IEEE, pp. 328-333.
[35] A. Afzal, N. Iqbal, A. Mujahid, and R. Schirhagl, “Advanced vapor recognition materials for selective and fast responsive surface acoustic wave sensors: A review,” Analytica chimica acta, vol. 787, pp. 36-49, 2013.
[36] J. R. Askim, M. Mahmoudi, and K. S. Suslick, “Optical sensor arrays for chemical sensing: The optoelectronic nose,” Chemical Society Reviews, vol. 42, no. 22, pp. 8649-8682, 2013.
[37] Y.-C. Tung, “Development of an ai odor recognition edge computing system using qcm-based electronic nose and applying in pneumonia pathogens odors recognition under in vitro culture environment,” Master's Thesis, Department of Biomedical Engineering, National Cheng Kung University Tainan, Taiwan, Republic of China (R.O.C.) 2022.
[38] E. Bonah, X. Huang, R. Yi, J. H. Aheto, R. Osae, and M. Golly, “Electronic nose classification and differentiation of bacterial foodborne pathogens based on support vector machine optimized with particle swarm optimization algorithm,” Journal of Food Process Engineering, vol. 42, no. 6, p. e13236, 2019.
[39] W. B. Gonçalves, W. S. R. Teixeira, A. N. d. C. E. Sampaio, O. A. Martins, E. P. Cervantes, M. d. S. R. Mioni, J. Gruber, and J. G. Pereira, “Combination of the electronic nose with microbiology as a tool for rapid detection of salmonella,” Journal of Microbiological Methods, vol. 212, p. 106805, 2023.
[40] P. P. Shinde and S. Shah, “A review of machine learning and deep learning applications,” in 2018 Fourth international conference on computing communication control and automation (ICCUBEA), 2018: IEEE, pp. 1-6.
[41] N. K. Chauhan and K. Singh, “A review on conventional machine learning vs deep learning,” in 2018 International conference on computing, power and communication technologies (GUCON), 2018: IEEE, pp. 347-352.
[42] Z. Ye, Y. Liu, and Q. Li, “Recent progress in smart electronic nose technologies enabled with machine learning methods,” Sensors, vol. 21, no. 22, p. 7620, 2021.
[43] X. Zhao, Z. Wen, X. Pan, W. Ye, and A. Bermak, “Mixture gases classification based on multi-label one-dimensional deep convolutional neural network,” IEEE access, vol. 7, pp. 12630-12637, 2019.
[44] Q. Wang, H. Qi, and F. Liu, “Time series prediction of e-nose sensor drift based on deep recurrent neural network,” in 2019 Chinese Control Conference (CCC), 2019: IEEE, pp. 3479-3484.
[45] S. Kang, Q. Zhang, Z. Li, C. Yin, N. Feng, and Y. Shi, “Determination of the quality of tea from different picking periods: An adaptive pooling attention mechanism coupled with an electronic nose,” Postharvest Biology and Technology, vol. 197, p. 112214, 2023.
[46] J. Luo, Z. Zhu, W. Lv, J. Wu, J. Yang, M. Zeng, N. Hu, Y. Su, R. Liu, and Z. Yang, “E-nose system based on fourier series for gases identification and concentration estimation from food spoilage,” IEEE Sensors Journal, vol. 23, no. 4, pp. 3342-3351, 2023.
[47] S. Zhang, Y. Cheng, D. Luo, J. He, A. K. Wong, and K. Hung, “Channel attention convolutional neural network for chinese baijiu detection with e-nose,” IEEE Sensors Journal, vol. 21, no. 14, pp. 16170-16182, 2021.
[48] S. Ruder, “An overview of multi-task learning in deep neural networks,” arXiv preprint arXiv:1706.05098, 2017.
[49] D. Yu, X. Wang, H. Liu, and Y. Gu, “A multitask learning framework for multi-property detection of wine,” IEEE Access, vol. 7, pp. 123151-123157, 2019.
[50] T. Wang, H. Zhang, Y. Wu, W. Jiang, X. Chen, M. Zeng, J. Yang, Y. Su, N. Hu, and Z. Yang, “Target discrimination, concentration prediction, and status judgment of electronic nose system based on large-scale measurement and multi-task deep learning,” Sensors and Actuators B: Chemical, vol. 351, p. 130915, 2022.
[51] C.-H. Wang, K.-Y. Huang, Y. Yao, J.-C. Chen, H.-H. Shuai, and W.-H. Cheng, “Lightweight deep learning: An overview,” IEEE consumer electronics magazine, vol. 13, no. 4, pp. 51-64, 2022.
[52] S. Kiranyaz, O. Avci, O. Abdeljaber, T. Ince, M. Gabbouj, and D. J. Inman, “1d convolutional neural networks and applications: A survey,” Mechanical systems and signal processing, vol. 151, p. 107398, 2021.
[53] G. Brauwers and F. Frasincar, “A general survey on attention mechanisms in deep learning,” IEEE Transactions on Knowledge and Data Engineering, vol. 35, no. 4, pp. 3279-3298, 2021.
[54] S. Chen, Y. Zhang, and Q. Yang, “Multi-task learning in natural language processing: An overview,” ACM Computing Surveys, vol. 56, no. 12, pp. 1-32, 2024.
[55] Y. Zhang and Q. Yang, “An overview of multi-task learning,” National Science Review, vol. 5, no. 1, pp. 30-43, 2018.
[56] M. Bhagat and B. Bakariya, “A comprehensive review of cross-validation techniques in machine learning,” IJSAT-International Journal on Science and Technology, vol. 16, no. 1, 2025.
[57] V. W. Lumumba, D. Kiprotich, N. Makena, M. Kavita, and M. Mpaine, “Comparative analysis of cross-validation techniques: Loocv, k-folds cross-validation, and repeated k-folds cross-validation in machine learning models,” Am. J. Theor. Appl. Stat, vol. 13, pp. 127-137, 2024.
[58] A. Berna, “Metal oxide sensors for electronic noses and their application to food analysis,” Sensors, vol. 10, no. 4, pp. 3882-3910, 2010.
[59] J. Song, X. Lin, L. Y. Ee, S. F. Y. Li, and M. Huang, “A review on electrospinning as versatile supports for diverse nanofibers and their applications in environmental sensing,” Advanced Fiber Materials, vol. 5, no. 2, pp. 429-460, 2023.
[60] A. D. Easley, T. Ma, C. I. Eneh, J. Yun, R. M. Thakur, and J. L. Lutkenhaus, “A practical guide to quartz crystal microbalance with dissipation monitoring of thin polymer films,” Journal of Polymer Science, vol. 60, no. 7, pp. 1090-1107, 2022.
[61] D. Johannsmann, A. Langhoff, and C. Leppin, “Studying soft interfaces with shear waves: Principles and applications of the quartz crystal microbalance (qcm),” Sensors, vol. 21, no. 10, p. 3490, 2021.
[62] J. Hassall, C. Coxon, V. C. Patel, S. D. Goldenberg, and C. Sergaki, “Limitations of current techniques in clinical antimicrobial resistance diagnosis: Examples and future prospects,” npj Antimicrobials and Resistance, vol. 2, no. 1, p. 16, 2024.
[63] S. Ahmad, S. Lohiya, A. Taksande, R. J. Meshram, A. Varma, K. Vagha, S. Ahmad Jr, and A. Varma Sr, “A comprehensive review of innovative paradigms in microbial detection and antimicrobial resistance: Beyond traditional cultural methods,” Cureus, vol. 16, no. 6, 2024.
[64] A. Haider, M. Ringer, Z. Kotroczó, C. Mohácsi-Farkas, and T. Kocsis, “The current level of maldi-tof ms applications in the detection of microorganisms: A short review of benefits and limitations,” Microbiology Research, vol. 14, no. 1, pp. 80-90, 2023.
[65] V. Puca, R. Z. Marulli, R. Grande, I. Vitale, A. Niro, G. Molinaro, S. Prezioso, R. Muraro, and P. Di Giovanni, “Microbial species isolated from infected wounds and antimicrobial resistance analysis: Data emerging from a three-years retrospective study,” Antibiotics, vol. 10, no. 10, p. 1162, 2021.
[66] N. Macesic, A.-C. Uhlemann, and A. Y. Peleg, “Multidrug-resistant gram-negative bacterial infections,” The Lancet, vol. 405, no. 10474, pp. 257-272, 2025.
[67] A. Smart, B. de Lacy Costello, P. White, M. Avison, C. Batty, C. Turner, R. Persad, and N. Ratcliffe, “Sniffing out resistance–rapid identification of urinary tract infection-causing bacteria and their antibiotic susceptibility using volatile metabolite profiles,” Journal of pharmaceutical and biomedical analysis, vol. 167, pp. 59-65, 2019.
[68] F. Li, Y. Zheng, C. Zhao, J. Zhu, Y. Hang, Y. Fang, and L. Hu, “Gc–ims facilitates identification of carbapenem-resistant klebsiella pneumoniae in simulated blood cultures,” AMB Express, vol. 14, no. 1, p. 40, 2024.
[69] W. Filipiak, K. Żuchowska, M. Marszałek, D. Depka, T. Bogiel, N. Warmuzińska, and B. Bojko, “Gc-ms profiling of volatile metabolites produced by klebsiella pneumoniae,” Frontiers in Molecular Biosciences, vol. 9, p. 1019290, 2022.
[70] N. Drabińska, B. de Lacy Costello, K. Hewett, A. Smart, and N. Ratcliffe, “From fast identification to resistance testing: Volatile compound profiling as a novel diagnostic tool for detection of antibiotic susceptibility,” TrAC Trends in Analytical Chemistry, vol. 115, pp. 1-12, 2019.
[71] L. Cheng, Q.-H. Meng, A. J. Lilienthal, and P.-F. Qi, “Development of compact electronic noses: A review,” Measurement Science and Technology, vol. 32, no. 6, p. 062002, 2021.
[72] Y. Li, Z. Wang, T. Zhao, H. Li, J. Jiang, and J. Ye, “Electronic nose for the detection and discrimination of volatile organic compounds: Application, challenges, and perspectives,” TrAC Trends in Analytical Chemistry, p. 117958, 2024.
[73] A. Khorramifar, H. Karami, L. Lvova, A. Kolouri, E. Łazuka, M. Piłat-Rożek, G. Łagód, J. Ramos, J. Lozano, and M. Kaveh, “Environmental engineering applications of electronic nose systems based on mox gas sensors,” sensors, vol. 23, no. 12, p. 5716, 2023.
[74] A. Aurora, “Algorithmic correction of mos gas sensor for ambient temperature and relative humidity fluctuations,” IEEE Sensors Journal, vol. 22, no. 15, pp. 15054-15061, 2022.
[75] X. Ding, J. Li, K. Tang, X. Chen, and H. Li, “A highly linear and sensitive qcm humidity sensor based on high-frequency quartz crystal transducer,” IEEE Transactions on Instrumentation and Measurement, vol. 73, pp. 1-9, 2023.
[76] A. Alassi, M. Benammar, and D. Brett, “Quartz crystal microbalance electronic interfacing systems: A review,” Sensors, vol. 17, no. 12, p. 2799, 2017.
[77] Y. Yao, X. Chen, X. Li, X. Chen, and N. Li, “Investigation of the stability of qcm humidity sensor using graphene oxide as sensing films,” Sensors and Actuators B: Chemical, vol. 191, pp. 779-783, 2014.