微陣列晶片是一種用來進行快速且大量生物檢測的工具，上頭有數千至數萬個如陣列般排列整齊的生物探針，會和檢測樣本中的待測物進行高度專一性的反應。其分析的第一步是使用微陣列晶片掃描機將微陣列晶片掃描成圖檔，圖片上的一個點代表一個生物探針，接著第二步是定位出圖上每個點的座標位置，得到各點的位置資訊後，第三步是算出每個點的強度讀值，此讀值代表的是生物探針與待測物的結合狀況，最後第四步就可以用各點的強度讀值做後續的數據分析。
近年來深度學習蓬勃發展，許多領域都嘗試以深度學習來實現自動化，在微陣列晶片的分析上，也陸續有研究提出基於深度學習的自動對點演算法以及自動強度讀取演算法，用來使微陣列晶片的分析流程更加自動化。本研究觀察過去的自動強度讀取演算法，發現三個會導致模型的讀值預測不準確的問題。為此，本研究以過去的自動強度讀取演算法的架構為基礎，並加以改進，提出了一個基於多頭深度學習的自動強度讀取演算法。
實驗結果顯示本研究提出的改進方法對於微陣列晶片的讀值預測有顯著的幫助。在數據分析階段，過去的自動強度讀取演算法所輸出的讀值，在新冠肺炎資料集上疾病檢測的AUC 只能達到 0.916，而本研究提出的自動強度讀取演算法所輸出的讀值，在同一個資料集的疾病檢測 AUC 可以達到 0.957。本研究提出之自動讀值演算法搭配自動對點演算法，可以達到完全自動化的分析。

Microarray chip is a tool for rapid and large-scale biological detection. There are thousands of biological probes arranged neatly like an array on the chip. The biological probes would have a highly specific reaction with the analyte in the testing samples. To analyze the microarray, the first step is to use the microarray scanner to scan the chip into an image file. Every spot on the image represents a biological probe. The second step is to locate the coordinate of each spot on the image. Subsequently, the third step is to calculate the intensity reading of each spot, which reflects the binding state of biological probes and the analytes. Finally, the fourth step is to use the intensity reading of each spot to do the follow-up data analysis.
In recent years, deep learning has been applied to many fields to fulfill automatic systems. For the microarray analysis, there are some studies based on deep learning are proposed to achieve automatic gridding and automatic intensity reading. However, there are some problems existing in previous automatic intensity reading algorithms being observed. Referring to the structure of previous automatic intensity reading algorithms, this work proposes an automatic intensity reading algorithm based on multi-head deep learning that try to solve the problems leading to inaccurate readout predictions.
The experimental results show that the method proposed in this study are all significantly helpful for the prediction of microarray readouts. In the data analysis stage, the readouts predicted by the previous automatic intensity reading algorithm can only reach 0.916 in the AUC of disease detection in the COVID-19 microarray dataset, while the improved automatic intensity reading algorithm proposed in this research could reach up to 0.957. The automatic intensity reading algorithm proposed in this study can combined with the automatic gridding algorithm to achieve fully automatic analysis for microarray.

[1] P. O. Brown, D. Botstein, “Exploring the new world of the genome with DNA microarrays,” Nature genetics, vol. 21, no. 1, pp. 33-37, 1999
[2] C. S. Chen, H. Zhu, “Protein microarrays,” Biotechniques, vol. 40, no. 4, pp. 423-429, 2006
[3] M. Schena, D. Shalon, R. W. Davis & P. O. Brown, “Quantitative monitoring of gene expression patterns with a complementary DNA microarray,” Science, vol. 270, no. 5235, pp. 467-470, 1995
[4] A. Wachter, S. Bernhardt, T. Beissbarth & U. Korf, “Analysis of reverse phase protein array data: from experimental design towards targeted biomarker discovery,” Microarrays, vol. 4, no. 4, pp. 520-539, 2015
[5] F. R. Sutandy, J. Qian, C. S. Chen & H. Zhu, “Overview of protein microarrays,” Current protocols in protein science, vol. 72, no. 1, pp. 27-1, 2013
[6] S. M. Zhou, L. Cheng, S. J. Guo, H. Zhu & S. C. Tao, “Functional protein microarray: an ideal platform for investigating protein binding property,” Frontiers in biology, vol. 7, no. 4, pp. 336-349, 2012
[7] M. C. Motwani, M. C. Gadiya, R. C. Motwani & F. C. Harris, “Survey of image denoising techniques,” In Proceedings of GSPX, vol. 27, pp. 27-30, 2004
[8] R. Hummel, “Image enhancement by histogram transformation,” Unknown, 1975
[9] S. M. Pizer, E. P. Amburn, J. D. Austin, R. Cromartie, A. Geselowitz, T. Greer & K. Zuiderveld, “Adaptive histogram equalization and its variations,” Computer vision, graphics, and image processing, vol. 39, no. 3, pp. 355-368, 1987
[10] K. Zuiderveld, “Contrast limited adaptive histogram equalization,” In Graphics gems IV, pp. 474-485, 1994
[11] W. S. McCulloch, W. Pitts, “A logical calculus of the ideas immanent in nervous activity,” The bulletin of mathematical biophysics, vol. 5, no. 4, pp. 115-133, 1943
[12] S. Ren, K. He, R. Girshick, & J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal networks,” Advances in neural information processing systems 28, 2015
[13] K. He, G. Gkioxari, P. Dollár & R. Girshick, “Mask r-cnn,” In Proceedings of the IEEE international conference on computer vision, pp. 2961-2969, 2017
[14] Z. Cai, N. Vasconcelos, “Cascade r-cnn: Delving into high quality object detection,” In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 6154-6162, 2018
[15] J. Wang, Y. Yang, J. Mao, Z. Huang, C. Huang & W. Xu, “Cnn-rnn: A unified framework for multi-label image classification,” In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2285-2294, 2016
[16] Y. LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, & L. D. Jackel, “Backpropagation applied to handwritten zip code recognition,” Neural computation, vol. 1, no. 4, pp. 541-551, 1989
[17] Y. LeCun, Y. Bengio, “Convolutional networks for images, speech, and time series,” The handbook of brain theory and neural networks, vol. 3361, no. 10, 1995
[18] Y. LeCun, L. Bottou, Y. Bengio & P. Haffner, “Gradient-based learning applied to document recognition,” Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, 1998
[19] K. Simonyan, A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv preprint arXiv:1409.1556, 2015
[20] C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov & A. Rabinovich, “Going deeper with convolutions,” In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1-9, 2015
[21] K. He, X. Zhang, S. Ren, & J. Sun, “Spatial pyramid pooling in deep convolutional networks for visual recognition,” IEEE transactions on pattern analysis and machine intelligence, vol. 37, no. 9, pp. 1904-1916, 2015
[22] J. Long, E. Shelhamer & T. Darrell, “Fully convolutional networks for semantic segmentation,” In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 3431-3440, 2015
[23] Y. Bengio, P. Simard, & P. Frasconi, “Learning long-term dependencies with gradient descent is difficult,” IEEE transactions on neural networks, vol. 5, no. 2, pp. 157-166, 1994
[24] X. Glorot, Y. Bengio, “Understanding the difficulty of training deep feedforward neural networks,” In Proceedings of the thirteenth international conference on artificial intelligence and statistics, pp. 249-256, 2010
[25] K. He, X. Zhang, S. Ren, & J. Sun, “Deep residual learning for image recognition,” In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770-778, 2016
[26] “Microarray acquisition and analysis software for GenePix microarray scanners user’s guide & tutorial,” ipmb.sinica.edu.tw. https://ipmb.sinica.edu.tw/microarray/index.files/GenePix_Pro_6.0_Manual_RevL.pdf (accessed: May. 31, 2022)
[27] P. Bajcsy, “Gridline: automatic grid alignment DNA microarray scans,” IEEE transactions on image processing, vol. 13, no. 1, pp. 15-25, 2004
[28] M. Ceccarelli, G. Antoniol, “A deformable grid-matching approach for microarray images,” IEEE Transactions on Image Processing, vol. 15, no. 10, pp. 3178-3188, 2006
[29] D. Bariamis, D. K. Iakovidis & D. Maroulis, “M3G: Maximum margin microarray gridding,” BMC bioinformatics, vol. 11, no. 1, pp. 1-11, 2010
[30] S. M. Joseph, P. S. Sathidevi, “A fully automated gridding technique for real composite cDNA microarray images,” IEEE Access, vol. 8, pp. 39605-39622, 2020
[31] L. F. Hsieh, “Automatic microarray gridding using transfer learning,” 2021
[32] S. N. Hsieh, “A new microarray analyzer based on deep learning,” 2020
[33] P. X. Du, Y. Y. Chou, H. M. Santos, B. B. Keskin, M. H. Hsieh, T. S. Ho & G. D. Syu, “Development and Application of Human Coronavirus Protein Microarray for Specificity Analysis,” Analytical chemistry, vol. 93, no. 21, pp. 7690-7698, 2021