3 March 2014 Classification algorithm of ovarian tissue based on co-registered ultrasound and photoacoustic tomography
Author Affiliations +
Abstract
Human ovarian tissue features extracted from photoacoustic spectra data, beam envelopes and co-registered ultrasound and photoacoustic images are used to characterize cancerous vs. normal processes using a support vector machine (SVM) classifier. The centers of suspicious tumor areas are estimated from the Gaussian fitting of the mean Radon transforms of the photoacoustic image along 0 and 90 degrees. Normalized power spectra are calculated using the Fourier transform of the photoacoustic beamformed data across these suspicious areas, where the spectral slope and 0-MHz intercepts are extracted. Image statistics, envelope histogram fitting and maximum output of 6 composite filters of cancerous or normal patterns along with other previously used features are calculated to compose a total of 17 features. These features are extracted from 169 datasets of 19 ex vivo ovaries. Half of the cancerous and normal datasets are randomly chosen to train a SVM classifier with polynomial kernel and the remainder is used for testing. With 50 times data resampling, the SVM classifier, for the training group, gives 100% sensitivity and 100% specificity. For the testing group, it gives 89.68± 6.37% sensitivity and 93.16± 3.70% specificity. These results are superior to those obtained earlier by our group using features extracted from photoacoustic raw data or image statistics only.
© (2014) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Hai Li, Patrick D. Kumavor, Umar Alqasemi, Quing Zhu, "Classification algorithm of ovarian tissue based on co-registered ultrasound and photoacoustic tomography", Proc. SPIE 8943, Photons Plus Ultrasound: Imaging and Sensing 2014, 894349 (3 March 2014); doi: 10.1117/12.2040467; https://doi.org/10.1117/12.2040467
PROCEEDINGS
7 PAGES


SHARE
Back to Top