Photoacoustic Imaging for Ovarian Cancer Detection: System Development and Classification Algorithm

Ovarian cancer is relatively rare but it has the highest mortality with a five-year survival rate of only 30% comparing with other gynecologic cancers. Most of ovarian cancers are diagnosed at late stages because of no efficacious screening techniques. So there is an urgent need to develop new imaging techniques for early stage ovarian cancer detection. Photoacoustic imaging (PAI) inherently combines the merits of optical imaging and ultrasound imaging. In PAI, photoacoustic waves are generated by illuminating tissue samples with a short laser pulse. Photoacoustic waves are then measured by ultrasound transducers to reconstruct optical absorption at ultrasound resolution, which is directly related to tumor angiogenesis. This research mainly focuses on the development of real-time co-registered ultrasound (US)/ photoacoustic tomography (PAT) imaging system and a classification algorithm for diagnosis of malignant vs. benign ovarian tissues. In this study, two versions of US/PAT systems were designed and implemented. To achieve real-time imaging capability for clinical application, efforts had been devoted to hardware structure and software algorithm optimization. We have achieved real-time imaging capability of 15 frames per second for patient studies. The system’s imaging capability is demonstrated in phantom and animal studies. A classification algorithm for diagnosis of malignant vs. benign ovarian tissues is the second topic of this search. Features from US/PAT imaging data, which may be helpful for ovarian cancer diagnosis, are extracted. Feature selection method is applied to select optimal subset for logistic regression classifier and supporter vector machine (SVM) classifier and promising results have obtained. The frame work set by this classification algorithm can be extended by having more features and advanced classifiers in the future