||Mammographic Lesion Detection System Using Convolutional
Neural Network and Empirical Mode Decomposition
||Department of BioMedical Engineering
乳房 X 光攝影
empirical mode decomposition
乳癌是20歲到59歲女性中，最常見致死率也很高的癌症，死亡率僅低於肺癌或支氣管相關癌症。乳癌的癌細胞是會轉移的，因此在其轉移前檢測到乳癌可能可以挽救許多生命。乳房X光攝影是早期篩檢最主要的工具，但影像判讀不易而導致篩檢的高召回率及過度診斷的結果。為了降低錯誤召回的發生機率，自1993年開始，建置電腦輔助診斷系統的研究越來越多。在開發電腦輔助診斷系統的過程中，特徵提取及選擇最為關鍵，目的是在不影響系統性能的情況下，減少所需運算的參數。本研究使用經驗模態分析 (Empirical mode decomposition, EMD) 進行特徵提取。經驗模態分析是一種可以將非線性非穩定的時變訊號分解成多個不同頻率的本質模態函數 (Intrinsic mode function, IMF) 的分析方法。分析二維數據的經驗模態分析算法很多，本研究將使用傳統的多維度經驗模態分析 (Multi-dimensional EMD, MEMD)及基於Radon transform的二維經驗模態分析 (Two-dimensional EMD, 2DEMD) 兩種算法。由於過去多用視覺評估來自不同經驗模態分析算法的二維本質模態函數，本研究根據本質模態函數本身之特性提出評估指標。經評估後證實，二維經驗模態分析 (2DEMD) 之性能表現要比傳統的多維度經驗模態分析好。隨後將經驗模態分析應用於乳房X光攝影之影像上，提取本質模態函數後將其用於訓練模型。分別將原始影像，提取後之所有本質模態函數，及移除第n個本質模態函數後之還原影像用於訓練模型，並比較模型性能結果。結果發現，當移除含有訓練目標之相關資訊的本質模態函數時，模型性能有明顯下降。相反的，若移除對於訓練目標無意義之本質模態函數時，模型性能則會上升。未來可以針對不同的病變特徵提取不同本質模態函數進行訓練模型，以提高電腦輔助診斷系統之性能。
Breast cancer, which is the most common and lethal cancer in women around the age of 20-59, has the mortality rate only lower than lung/bronchus cancer. Breast cancer is metastatic cancer, hence cancer detection before the metastasis might save many lives. Mammography is the primary screening tool for early detection of breast cancer, but the difficulty in image interpretation causes a high recall rate and overdiagnosis. To decrease the rate of false-positive, the amount of research on the computer-aided diagnosis (CAD) system has been growing since 1993. Empirical mode decomposition (EMD) will be used in this research to reveal the characteristics of the different lesions in the breast. EMD decomposes the original signal into a finite number of components called intrinsic mode function (IMF). There are many algorithms of EMD to decompose the 2D data. In this study, the traditional multi-dimensional EMD and Radon transform-based 2DEMD were employed. However, the evaluation of 2D IMFs is limited since there are fewer exact metrics. This study proposed some modified metrics, which depend on the characteristics of the IMFs, to improve the method. The result indicated that the performance of 2DEMD is better than multi-dimensional EMD. Next, 2DEMD was applied to mammograms to extract the IMFs. These IMFs would be used as the training data for the model. Finally, several CAD systems performances were compared in pairs: CNN with original images, CNN with all IMFs, and CNN with the reconstructed image without n-th IMF. The model performance drastically reduced if the training data was lacking meaningful IMFs fewer meaningless IMFs would improve the model performance. Concerning prospects for the future, this system provides a promising method to train the model with more meaningful IMFs to different lesion types, hence further improve the performance of the CAD system.
Chapter 1. Introduction 1
Chapter 2. Literature Review 4
2.1. Mammographic Screening 4
2.1.1. Breast Density Classification 5
2.1.2. Breast Lesion in Mammography 8
2.1.3. Final Assessment Category 9
2.1.4. High Recall Rate 10
2.2. Computer-Aided Diagnosis/Detection (CAD) system 11
2.2.1. Feature Extraction/Selection 12
2.2.2. Deep Learning 16
2.2.3. Current Deep Learning Research on Mammography 18
Chapter 3. Material and Methods 19
3.1. Research Framework 19
3.2. Data Acquisition 20
3.2.1. Digital Database for Screening Mammography (DDSM) 20
3.2.2. Curated Breast Imaging Subset of DDSM (CBIS-DDSM) 20
3.2.3. Mammographic Image Analysis Society (MIAS) database 21
3.2.4. ROI Extraction Method 21
3.3. Experiment Flowchart 22
3.4. Feature Extraction with Empirical Mode Decomposition (EMD) 23
3.4.1. Intrinsic Mode Function (IMF) 23
3.4.2. Empirical Mode Decomposition (EMD) 23
3.4.3. Multi-dimensional Ensemble Empirical Mode Decomposition (MEEMD) 27
3.4.4. 2D Empirical Mode Decomposition (2DEMD) 29
3.5. The Metrics to Evaluate the Performance of 2D IMFs 31
3.5.1. Metrics 1: The Overlapping Degree of the Amplitude and Phase of IMF in the Spectra Domain. 31
3.5.2. Metrics 2: The Quality of the Reconstructed Image from IMFs and Residue 33
Chapter 4. Result and Discussion 35
4.1. The Comparison of the EMD Algorithm on 2D Data 35
4.1.1. The Visual Inspection 35
4.1.2. The Performance Metrics 39
4.2. Training Result with the whole image 43
4.2.1. The Training Result on MIAS database 43
4.2.2. The Training Result of DDSM Database 45
4.3. Training Result with ROI image 49
Chapter 5. Conclusion 54
 R. L. Siegel, K. D. Miller, and A. Jemal, “Cancer statistics, 2020,” CA. Cancer J. Clin., vol. 70, no. 1, pp. 7–30, 2020.
 B. Sai and J. Xiang, “Disseminated tumour cells in bone marrow are the source of cancer relapse after therapy,” J. Cell. Mol. Med., vol. 22, no. 12, pp. 5776–5786, 2018.
 “Breast Cancer Screening Methods,” Our Bodies Ourselves. https://www.ourbodiesourselves.org/book-excerpts/health-article/breast-cancer-screening-methods/ (accessed Feb. 07, 2019).
 K. C. Oeffinger et al., “Breast Cancer Screening for Women at Average Risk: 2015 Guideline Update From the American Cancer Society,” JAMA, vol. 314, no. 15, pp. 1599–1614, Oct. 2015.
 A. Oliver et al., “A review of automatic mass detection and segmentation in mammographic images,” Med. Image Anal., vol. 14, no. 2, pp. 87–110, Apr. 2010.
 J. Yanase and E. Triantaphyllou, “A systematic survey of computer-aided diagnosis in medicine: Past and present developments,” Expert Syst. Appl., vol. 138, p. 112821, Dec. 2019.
 N. E. Huang et al., “The Empirical Mode Decomposition and the Hilbert Spectrum for Nonlinear and Non-Stationary Time Series Analysis,” Proc. Math. Phys. Eng. Sci., vol. 454, no. 1971, pp. 903–995, 1998.
 C. Y. Chen, “Mammographic Feature Characterization Using Two‐Dimensional Empirical Mode Decomposition,” National Cheng Kung University, Tainan, Taiwan, 2014.
 H. Zonderland and R. Smithuis, “BI-RADS for mammography and ultrasound 2013,” Updat. Version Available Radiol. Nlbreastbi-Rads--Mammogr.--Ultrasound-2013 Publ. Oct., vol. 8, 2014.
 D. A. Spak, J. S. Plaxco, L. Santiago, M. J. Dryden, and B. E. Dogan, “BI-RADS® fifth edition: A summary of changes,” Diagn. Interv. Imaging, vol. 98, no. 3, pp. 179–190, Mar. 2017.
 V. A. McCormack, “Breast Density and Parenchymal Patterns as Markers of Breast Cancer Risk: A Meta-analysis,” Cancer Epidemiol. Biomarkers Prev., vol. 15, no. 6, pp. 1159–1169, Jun. 2006.
 A. T. Wang, C. M. Vachon, K. R. Brandt, and K. Ghosh, “Breast Density and Breast Cancer Risk: A Practical Review,” Mayo Clin. Proc., vol. 89, no. 4, pp. 548–557, Apr. 2014.
 J. H. Youk, E.-K. Kim, K. H. Ko, and M. J. Kim, “Asymmetric Mammographic Findings Based on the Fourth Edition of BI-RADS: Types, Evaluation, and Management,” RadioGraphics, vol. 29, no. 1, pp. e33–e33, Jan. 2009.
 P. J. Ho et al., “Factors associated with false-positive mammography at first screen in an Asian population,” PLOS ONE, vol. 14, no. 3, p. e0213615, Mar. 2019.
 C. S. Giess, A. Wang, I. K. Ip, R. Lacson, S. Pourjabbar, and R. Khorasani, “Patient, Radiologist, and Examination Characteristics Affecting Screening Mammography Recall Rates in a Large Academic Practice,” J. Am. Coll. Radiol., vol. 16, no. 4, Part A, pp. 411–418, Apr. 2019.
 D. Ping Tian, “A review on image feature extraction and representation techniques,” Int. J. Multimed. Ubiquitous Eng., vol. 8, no. 4, pp. 385–396, 2013.
 E. U. Nathaniel, N. J. George, and S. E. Etuk, “Determination of Instantaneous Frequencies of Low Plasma Waves in the Magnetosheath Using Empirical Mode Decomposition (EMD) and Hilbert Transform (HT),” Atmospheric Clim. Sci., vol. 2013, Aug. 2013.
 I. Zyout and R. Togneri, “A computer-aided detection of the architectural distortion in digital mammograms using the fractal dimension measurements of BEMD,” Comput. Med. Imaging Graph. Off. J. Comput. Med. Imaging Soc., vol. 70, pp. 173–184, 2018.
 V. Bajaj, M. Pawar, V. K. Meena, M. Kumar, A. Sengur, and Y. Guo, “Computer-aided diagnosis of breast cancer using bi-dimensional empirical mode decomposition,” Neural Comput. Appl., vol. 31, no. 8, pp. 3307–3315, Aug. 2019.
 V. Nagarajan, E. C. Britto, and S. M. Veeraputhiran, “Feature extraction based on empirical mode decomposition for automatic mass classification of mammogram images,” Med. Nov. Technol. Devices, vol. 1, p. 100004, Mar. 2019.
 Z. Wu, N. Huang, and X. Chen, “The Multi-Dimensional Ensemble Empirical Mode Decomposition Method.,” Adv. Adapt. Data Anal., vol. 1, pp. 339–372, Jul. 2009.
 Z. Wu and N. E. Huang, “Ensemble empirical mode decomposition: a noise-assisted data analysis method,” Adv. Adapt. Data Anal., vol. 01, no. 01, pp. 1–41, Jan. 2009.
 T. Hey, K. Butler, S. Jackson, and J. Thiyagalingam, “Machine learning and big scientific data,” Philos. Transact. A Math. Phys. Eng. Sci., vol. 378, no. 2166, Mar. 2020.
 M. D. Zeiler and R. Fergus, “Visualizing and Understanding Convolutional Networks,” ArXiv13112901 Cs, Nov. 2013, Accessed: Jul. 03, 2020. [Online]. Available: http://arxiv.org/abs/1311.2901.
 K. Simonyan and A. Zisserman, “Very Deep Convolutional Networks for Large-Scale Image Recognition,” ArXiv14091556 Cs, Apr. 2015, Accessed: Jul. 03, 2020. [Online]. Available: http://arxiv.org/abs/1409.1556.
 C. Szegedy et al., “Going Deeper with Convolutions,” ArXiv14094842 Cs, Sep. 2014, Accessed: Jul. 03, 2020. [Online]. Available: http://arxiv.org/abs/1409.4842.
 K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image Recognition,” ArXiv151203385 Cs, Dec. 2015, Accessed: Jul. 03, 2020. [Online]. Available: http://arxiv.org/abs/1512.03385.
 “Evolution of CNN Architectures: LeNet, AlexNet, ZFNet, GoogleNet, VGG and ResNet,” OpenGenus IQ: Learn Computer Science, Jan. 21, 2019. https://iq.opengenus.org/evolution-of-cnn-architectures/ (accessed May 04, 2019).
 S. S. Aboutalib, A. A. Mohamed, W. A. Berg, M. L. Zuley, J. H. Sumkin, and S. Wu, “Deep Learning to Distinguish Recalled but Benign Mammography Images in Breast Cancer Screening,” Clin. Cancer Res., Oct. 2018.
 D. A. Ragab, M. Sharkas, S. Marshall, and J. Ren, “Breast cancer detection using deep convolutional neural networks and support vector machines,” PeerJ, vol. 7, p. e6201, Jan. 2019.
 H. Sun et al., “AUNet: attention-guided dense-upsampling networks for breast mass segmentation in whole mammograms,” Phys. Med. Biol., vol. 65, no. 5, p. 055005, Feb. 2020.
 L. Shen, L. R. Margolies, J. H. Rothstein, E. Fluder, R. McBride, and W. Sieh, “Deep Learning to Improve Breast Cancer Detection on Screening Mammography,” Sci. Rep., vol. 9, no. 1, Art. no. 1, Aug. 2019.
 S. M. McKinney et al., “International evaluation of an AI system for breast cancer screening,” Nature, vol. 577, no. 7788, pp. 89–94, Jan. 2020.
 I. Zyout and R. Togneri, “A computer-aided detection of the architectural distortion in digital mammograms using the fractal dimension measurements of BEMD,” Comput. Med. Imaging Graph., vol. 70, pp. 173–184, Dec. 2018.
 S. S. Aboutalib, A. A. Mohamed, W. A. Berg, M. L. Zuley, J. H. Sumkin, and S. Wu, “Deep Learning to Distinguish Recalled but Benign Mammography Images in Breast Cancer Screening,” Clin. Cancer Res., vol. 24, no. 23, pp. 5902–5909, Dec. 2018.
 H. Chougrad, H. Zouaki, and O. Alheyane, “Deep Convolutional Neural Networks for breast cancer screening,” Comput. Methods Programs Biomed., vol. 157, pp. 19–30, Apr. 2018.
 E. A. Scuccimarra, “ConvNets for Detecting Abnormalities in DDSM Mammograms,” Medium, Jul. 20, 2018. https://medium.com/@ericscuccimarra/convnets-for-classifying-ddsm-mammograms-1739e0fe8028 (accessed Apr. 10, 2019).
 K. Bowyer et al., “The digital database for screening mammography,” in Third international workshop on digital mammography, 1996, vol. 58, p. 27.
 R. S. Lee, F. Gimenez, A. Hoogi, K. K. Miyake, M. Gorovoy, and D. L. Rubin, “A curated mammography data set for use in computer-aided detection and diagnosis research,” Sci. Data, vol. 4, no. 1, Art. no. 1, Dec. 2017.
 J. Suckling, The mammographic image analysis society digital mammogram database Digital Mammography ed AG Gale, SM Astley, DR Dance and AY Cairns. Amsterdam: Elsevier, 1994.
 D. Laszuk, O. Cadenas, and S. J. Nasuto, “Objective Empirical Mode Decomposition metric,” in 2015 38th International Conference on Telecommunications and Signal Processing (TSP), Jul. 2015, pp. 504–507.
 L. Salwa and R. Mohammed, “Novel phase-based descriptor using bispectrum for texture classification,” Pattern Recognit. Lett., vol. 100, pp. 1–5, Dec. 2017.
 A. V. Oppenheim and J. S. Lim, “The importance of phase in signals,” Proc. IEEE, vol. 69, no. 5, pp. 529–541, 1981.
 H. Kanders, T. Mellqvist, M. Garrido, K. Palmkvist, and O. Gustafsson, “A 1 Million-Point FFT on a Single FPGA,” IEEE Trans. Circuits Syst. Regul. Pap., vol. 66, no. 10, pp. 3863–3873, Oct. 2019.
 P. Skurowski, “Towards the new concept of linear image quality assessment. Review of phase correlation formulas,” Stud. Inform., vol. 35, pp. 111–131, Dec. 2014.