生物標記 (biomarker) 指對正常生物過程、致病過程或治療干預造成的反應進行測量得到的特徵指標，如特定的DNA片段，蛋白質等，在疾病的診斷及研究上有著重要的作用。微陣列晶片 (microarray chip) 提供了一個高效率測量生物標記的方法，而網格化 (gridding) 則是微陣列晶片分析流程中不可或缺的一個步驟。
網格化是在微陣列晶片掃描出來的圖片上，找出每一個生物標記的座標位置。以往的網格化多採用傳統影像處理來進行圖片的最佳化。但是微陣列晶片的排版、圖片狀況各有差異，這些最佳化方法在某些有高密度雜訊的資料集上的成效並不佳。為此，本研究提出了一個基於深度學習的網格化方法，利用物件偵測 (object detection) 任務中常用的深度殘差網路 (Resnet)，不需要經過圖片的最佳化可預測生物標記的座標。本研究亦利用多個版型不同的微陣列晶片資料集進行預訓練 (pre-train)，並且使用資料增補 (data augmentation) 來增加訓練資料的多樣性。
實驗結果顯示預訓練與資料增補對於網格化的結果都有顯著的幫助。我們的方法在存在高密度雜訊的資料集中，準確率由原本的不足50% 提升至99.3%，並且在其他資料集上也維持相當的水準。我們亦將此網格化的方法套用至一個covid-19晶片的生物特徵篩選流程，使此流程達到完全自動化，並且能夠篩選出與以往相似的生物特徵。

Biomarker, an indicator that is able to measure metabolism, the process of development of disease, treatment, and at the same time be able to help making clinical decision. Microarray offers an effective and convenient way to find biomarkers, where gridding is a necessary step of microarray analysis procedure.
Gridding is the task of locating each biosensor on a microarray chip image. Previous researches often utilized rule-based image processing algorithms such as equalization and de-noising to enhance the image. However, these rule-based algorithms don’t perform well on datasets with high density noises. Therefore, this study proposes a deep learning-based gridding method, using Resnet to directly predict the coordinates of biosensors without any image enhancement. The proposed model is pre-trained on microarray datasets of different layouts with data augmentations to increase the variety of training samples.
Experimental results in this study show that both pre-training and data augmentation improve gridding accuracy, especially on datasets with severe noises. This study also tested the proposed gridding to automate the biomarker detection on a microarray dataset of covid-19. The experimental results show that the automatic pipeline delivers similar results as the original pipeline with manual gridding.

[1] S. M. Joseph and P. S. Sathidevi, "A Fully Automated Gridding Technique for Real Composite cDNA Microarray Images," IEEE access, 24 2 2020.
[2] GenePix, "GenePix Pro 6.0 Manual," [Online]. Available: https://ipmb.sinica.edu.tw/microarray/index.files/GenePix_Pro_6.0_Manual_RevL.pdf. [Accessed 28 6 2021].
[3] P. Bajcsy, “Gridline: Automatic Grid Alignment in DNA Microarray Scans,” IEEE TRANSACTIONS ON IMAGE PROCESSING, 1 2004.
[4] M. Ceccarelli and G. Antoniol, “A Deformable Grid-Matching Approach for Microarray Images,” IEEE TRANSACTIONS ON IMAGE PROCESSING, 10 10 2006.
[5] D. Bariamis and D. K. Iakovidis, “M3G: Maximum Margin Microarray Gridding,” Bioinformatics, 2010.
[6] K. He, J. Sun, and X. Tang, “Single Image Haze Removal Using Dark Channel Prior,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 9 9 2010.
[7] W. S. McCulloch and W. Pitts , "A logical calculus of the ideas immanent in nervous actitity," The bulletin of mathematical biophysics, pp. 115-133, 4 5 1943.
[8] K. He, X. Zhang, S. Ren, and J. Sun, "Deep Residual Learning for Image Recognition," Computer Vision and Pattern Recognition, 10 12 2015.
[9] A. Kolesnikov, et al., “Big Transfer (BiT): General Visual Representation Learning,” Computer Vision and Pattern Recognition, 5 5 2020.
[10] J. Devlin, M. W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” Computation and Language, 24 5 2019.
[11] "Gene Expression Omnibus (GEO)," National Center for Biotechnology Information, [Online]. Available: http://www.ncbi.nlm.nih.gov/geo/. [Accessed 1 12 2020].
[12] P. X. Du, et al., “Development and Application of Human Coronavirus Protein Microarray for Specificity Analysis,” Analytial Chemistry, 19 5 2021.
[13] P. C. Chen, G. D. Syu, K. H. Chung, Y. W. Chen, S. Y. Tsai, C. S. Chen, “Antibody Profiling of Bipolar Disorder Using Escherichia coli Proteome Microarrays,” Molecular and Cellular Proteomics, 1 3 2015.
[14] H. C. Kuo, et al., “Antibody Profiling of Kawasaki Disease Using Escherichia coli Proteome Microarrays,” Molecular & Cellular Proteomics, 3 2018.
[15] S. N. Hsieh, “A new microarray analyzer based on deep learning,” 7 2020.
[16] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards Real-Time Object,” Computer Vision and Pattern Recognition, 6 1 2016.
[17] D. P. Kingma and J. Ba, “Adam: A Method for Stochastic Optimization,” 3rd International Conference for Learning Representations, 23 6 2015.
[18] L. N. Smith, “Cyclical Learning Rates for Training Neural Networks,” Computer Vision and Pattern Recognition, 4 4 2017.
[19] Z. Gan, et al., “Multilevel Segmentation Optimized by Physical Information for Gridding of Microarray Images,” IEEE Access, 19 2 2019.