..................................................................................................................................... 2 Acknowledgment ....................................................................................................................... 3 List of Figures ............................................................................................................................ 6 List of Tables ............................................................................................................................. 7 Abbreviations ............................................................................................................................. 8 Introduction ........................................................................................................................ 9 1.1. Background ................................................................................................................. 9 1.2. Motivation ................................................................................................................... 9 1.3. Objectives .................................................................................................................... 9 1.4. Scope ......................................................................................................................... 10 1.5. Related Works ........................................................................................................... 10 1.6. Outline of the report .................................................................................................. 10 Outline of Deliverables .................................................................................................... 11 2.1. Data collection tool ................................................................................................... 11 2.2. Automated diagnostic system for bacterial vaginosis ............................................... 11 Methodology .................................................................................................................... 11 3.1. Overview ................................................................................................................... 11 3.2. Prerequisites .............................................................................................................. 13 3.2.1. Hardware ............................................................................................................ 13 3.2.2. Software ............................................................................................................. 13 3.3. Data Collection .......................................................................................................... 13 3.3.1. Smear Images ..................................................................................................... 14 3.3.2. High-level Interpretation of Images ................................................................... 15 3.3.3. Detailed Labelling of Images and the tool Clicklable ....................................... 15 3.4. Image Processing....................................................................................................... 17 3035209241 Chi Ian Tang Page 5 of 32 3.4.1. Pre-processing .................................................................................................... 17 3.4.2. Segmentation...................................................................................................... 19 3.5. Classification ............................................................................................................. 19 3.5.1. Artificial Neural networks ................................................................................. 20 3.5.2. Convolutional neural networks .......................................................................... 21 3.6. Interpretation ............................................................................................................. 22 3.7. Performance Evaluation ............................................................................................ 22 Results .............................................................................................................................. 23 4.1. Work completed ........................................................................................................ 23 4.1.1. Data collection ................................................................................................... 23 4.1.2. Data Labelling Tool ........................................................................................... 23 4.1.3. Diagnostic system .............................................................................................. 23 4.2. Performance of the first diagnostic tool prototype .................................................... 23 Limitations and Future Direction ..................................................................................... 25 5.1. Problems identified ................................................................................................... 25 5.1.1. Limited amount of data ...................................................................................... 25 5.1.2. Similarity among bacteria .................................................................................. 26 5.1.3. High variation in smear images ......................................................................... 26 5.1.4. Discrepancies in morphology ............................................................................ 27 5.2. Future Direction ........................................................................................................ 27 5.3. Project Schedule ........................................................................................................ 27 5.3.1. Schedule of Phase 1 ........................................................................................... 28 5.3.2. Schedule of Phase 2 ........................................................................................... 28 5.3.3. Schedule of Phase 3 ........................................................................................... 28 Conclusion ....................................................................................................................... 29 References ........................................................................................................................ 30 Classification for Pathological Images Using Machine Learning Intermediate Report Page 6 of 32 List of Figures Figure 1. A Gram-stained vaginal smear ................................................................................. 14 Figure 2. A Gram-stained bacterial colony smear ................................................................... 14 Figure 3. The main user-interface of the data collection tool, Clicklable. ............................... 15 Figure 4. The popup menu in Clicklable ................................................................................. 16 Figure 5. Annotation settings in Clicklable ............................................................................. 16 Figure 6. An example of convolution of an image with a convulsion filter ............................ 17 Figure 7. The segmentation of a blood smear image ............................................................... 19 Figure 8. The mathematical model of a neuron ....................................................................... 20 Figure 9. The diagram of a human neuron ............................................................................... 20 Figure 10. A 3-layer neural network ........................................................................................ 20 Figure 11. Example of a network with many convolutional layers ......................................... 21 Figure 12A. Segmented area containing a bacterium type “Lactobacillus” ............................ 26 Figure 12B. Segmented area containing a bacterium type “Gardnerella” ............................... 26 Figure 13A. A patient smear image in image batch 1 .............................................................. 26 Figure 13B. A patient smear image in image batch 2 .............................................................. 26 3035209241 Chi Ian Tang Page 7 of 32 List of Tables Table 1. The Nugent Scoring system ....................................................................................... 12 Table 2. The interpretation of the Nugent Score...................................................................... 12 Table 3. Validation results on type of bacteria in segmented images of the first classifier prototype .................................................................................................................................. 24 Table 4. Testing results on degree of infection of the first classifier prototype ...................... 24 Table 5. Testing results on Nugent Score of the first classifier prototype ............................... 25 Classification for Pathological Images Using Machine Learning Intermediate Report Page 8 of 32 Abbreviations BV Bacterial vaginosis DBSCAN Density-based spatial clustering of applications with noise HIV Human immunodeficiency virus MSER Maximally stable extremal regions SQL Structured Query Language 3035209241 Chi Ian Tang Page 9 of 32 Introduction 1.1. Background Bacterial infection is a common medical condition in humans, and several pathogens were found to be responsible for the development of malignant tumours [1]. Bacterial vaginosis (BV), one of the most common bacterial infections in the vagina, was estimated to affect tens of millions of people in the United States of America alone [2]. The prevalence of this infection varies by countries and can be as high as 50% in women at reproductive age [1]. Studies have also shown that this infection in