The fact that the system has detected only 8 images properly with their real cells and no false cells out of the whole dataset was a key issue to consider if the system is to be robust enough for clinical use. Further work is required to correctly identify only the true cells in the images. The current algorithms have detected 80% of the true cells, and considerable work was required to reduce the false cells down to minimum.
148
The work done measured some of the thresholds according to the images available. These thresholds might slightly change with the change of the clinic. Creating a system before the proposed detection system to automatically adjust these thresholds will make this system more effective to work on any image from any clinic.
If accuracy of the detected Day 2 cells can be sufficiently improved for clinical use it will be possible to implement this on an embedded system that can be placed inside the incubator. The system will take pictures of the embryos while they are inside the incubator at the appropriate time (at Day 2), and then detect, classify and finally grade the embryos.
The work in this thesis focused on the detection of cells of Day 2 embryos. However, work can be done to automatically measure the fragmentation percentage, which can be used along with the number of cells count detected in order to perform an automatic grading system for embryos at other stages of development.
Finally, the merger between BPS and IPS can be tried out using the concept of fuzzy logic. These two values can be considered as the variables of the fuzzy system and an appropriate membership function can be developed to produce the desired output, which in this case would be the measure of goodness.
Bibliography
Awcock, G. & Thomas, R. 1995, Applied Image Processing, Macmillan.
Beuchat, A. & Thevenaz, P, 2008, 'Quantitative morphometrical characterization of human pronuclear zygotes.', Human Reproduction, vol 23, no. 9, pp. 1983-1992.
Bikhet, S., Darwish, A. & Tolba, H. 2000, 'Segmentation and classification of white blood cells', ICASSP '00.
Bqczkowski, T., Kurzawa, R. & W., G. 2004, 'Methods of embryo scoring in in vitro fertilization', Reproductive Biology, vol 4, no. 1, pp. 5-22.
Canny 1986, 'A Computational Approach to Edge Detection', IEEE Transactions on Pattern
Analysis and Machine Intelligence, vol 8, pp. 679-714.
Cummins, J., Breen, T. & Harison, K, 1986, 'A formula for scoring human embryo growth rates in in vitro fertilization: its value in predicting pregnancy and in comparison with visual estimates of embryo quality.', In vitro Fertilization and Embryo Transfer, vol 3, no. 5, pp. 284-295.
Dokras, A., Sargent, I. & Barlow, D. 1993, 'Human blastocyst grading: an indicator of development potential?', Human Reproduction., vol 8, no. 12, pp. 2119-2127.
Filho, E., Noble, J. & Wells, D. 2010, 'A review on automatic analysis of human embryo microscope images.', The open Biomedical Engineering, vol 4, pp. 170-171.
150
Gardner D.K., SE,MD 2005, 'Single blastocyst transfer: a prospective randomized trial',
Fertility and Sterility, no. 81, pp. 551-555.
Giusti, A, Corani, G & L.M, G 2009, 'Segmentation of human zygotes in hoffman modulation contrast images.', Medical Image Understanding and Analysis (MIUA'09).
Gonzalez, R. & Woods, R. 2002, Digital Image Processing (2nd ed.)., Prentice Hall.
Heske T., HJN 1999, Fuzzy logic for real world design, Annabooks, San Diego.
Hnida, C., Engenheiro, E. & Ziebe, S. 2004, 'Computer-controlled, multilevel, morphometric analysis of blastomere size as biomarker of fragmentation and multinuclearity in human embryos.', Human Reproduction, vol 19, no. 2, pp. 288-293.
Hough, P. V. C., Method and Means for Recognizing Complex Patterns, US Patents
3969654, 1996
Kass M., WA,TD 1988, 'Snakes: active Contour Model', International journal of Computer
Vision, pp. 321-331.
Lau, T. & Bischof, W. 1991, 'Automated detection of breast tumors using the asymmetry approach.', Computers and Biomedical Research., vol 24, no. 3, pp. 271-295.
Miroslaw, L, Chorazyczewski, A, Buchholz, F & Kittler, R 2005, 'Correlation-based Method for Automatic Mitotic Cell Detection in Phase Contrast Microscopy', in Computer
Recognition Systems.
Morales, DA., Bengoetxea, E. & Larranaga, P. 2008, 'Automatic segmentation of zona pellucida in human embryo images applying an active contour model.', Medical Image
Understanding and Analysis (MIUA'08).
Puissant, F., Rysselberge, M. & Barlow, P, 1987, 'Embryo scoring as a prognostic tool in ivf treatment.', Human Reproduction., vol 2, no. 8, pp. 705-708.
Ross, N., Pritchard, C. & Rubin, D. 2006, , Medical and Biological Engineering and
Computing., vol 44, no. 5, pp. 427-436.
Sallam, H 2001, Infertility and Assisted Conception.
Scott, L., Alvero, R. & M., L. 2000, 'The morphology of human pronuclear embryos is positively related to blastocyst development and implantation.', Human Reproduction., vol 15, no. 11, pp. 2394-2403.
Staessen, C., Camus, M. & Bollen, N. 1992, 'The relationship between embryo quality and the occurrence of multiple pregnancies.', Fertility and Sterility., vol 57, no. 3, pp. 262-630.
Steer, C., Mills, C. & Tan, S. 1992, 'The cumulative embryo score: a predictive embryo scoring technique to select the optimal number of embryos to transfer in an in vitro
fertilization and embryo transfer programme.', Human Reproduction., vol 7, no. 1, pp. 117- 119.
Vromen, J. & McCane, B. 2009, 'Red blood cell segmentation from sem images.', Image and
Vision Computing IVCNZ '09., New Zealand.
Zeibe, S., Peterson, K. & Lindenberg, S. 1997, 'Embryo morphology or cleavage stage: how to select the best embryos for transfer after in vitro fertilization.', Human Reproduction., vol 12, no. 7, pp. 1545-1549.
152
APPENDIX A
Images Dataset
Image 1 Image 9 Image 17 Image 25 Image 2 Image 10 Image 18 Image 26 Image 3 Image 11 Image 19 Image 27 Image 4 Image 12 Image 20 Image 28 Image 5 Image 13 Image 21 Image 29 Image 6 Image 14 Image 22 Image 7 Image 15 Image 23 Image 8 Image 16 Image 24154
APPENDIX B
This Appendix contains the results of the Hough Transform technique that was discussed in Chapter 5. Some of these results were depicted in the Chapter while the rest are shown in the tables in the following pages.
The results of the detected cells of each of the three types of edge detectors (Sobel, new algorithm and convolution mask) are given in each table. However, these results include the x-centre, y-centre, radius and finally the matching value which corresponds to the value of accumulator. The detected cell is also shown as a blue circle plotted on the image.
173
APPENDIX C
This Appendix contains the results of the proposed technique in Chapter 6. These results are the results of applying the correlation technique and getting the values of the maximum correlation
179
APPENDIX D
This Appendix contains the results of the proposed technique in Chapter 6. These results are the results of applying the correlation technique and getting the values all the maximum correlation cooeffient. The two techniques used were the SAD and the NCC.