Automatic Detection of Embryo Location in Medical Imaging Using Trigonometric Rotation for Noise Reduction

The assessment of embryos in vitro is an important tool both in In Vitro Fertilization (IVF) treatment and for research purposes. Traditionally, such assessment has been done manually and demands extensive training and expertise. Interobserver variability limits the use of evaluation, and the manual labor increases the cost of treatment and research. To this end, feature extraction and automatic annotation using computer aided tools would both improve objectivity and save time. Due to the sensitivity of embryos, any evaluation must be performed using non-invasive imaging techniques, which sets a limit for the ability to artificially enhance image contrast, and computer aided assessment of microscopic images under clinical conditions also poses a number of difficulties due to the high level of noise and the unpredictable imaging environment. To overcome this, we propose a framework using a Hough Transform (HT) for the extraction of circular shapes in microscopic embryo images, and also use the algorithm for successful automatic profiling of the pronuclear breakdown (PNB) at syngami and detection of its timing. The best results are achieved when using a HT for embryo outline detection in combination with image variance of the embryo interior. Using our method, we achieve an overall accuracy of 95.0% detecting the position of the embryo outline, and 83.0% detecting the timing of the PNB.