Melanoma is one of the most serious skin cancers in the world. As circulating tumor cells have been proved to be an important marker of early metastasis of cancer, the detection of circulating tumor cells of melanoma is of great significance for early diagnosis and the monitoring of tumor progression. In vivo photoacoustic flow cytometry (PAFC) is constructed to achieve real-time and non-invasive detection of circulating melanoma cells in vivo. However, as the photoacoustic signals acquired in the detection process are disturbed by various kinds of noise, it is difficult to accurately distinguish the photoacoustic signals of background and circulating tumor cells by the traditional triple mean square deviation method. Therefore, a photoacoustic signal classification method is proposed based on convolutional neural network, which can greatly improve the accuracy of detection. Features of signals are extracted by the convolutional neural network to distinguish photoacoustic signals of melanoma cells and background. We construct a convolutional neural network based on one-dimensional input signals. For training the classifier, a large number of samples are selected. The accuracy rate in the test set can reach 95%. Besides, a neural network is built based on VGG16 model and transfer learning, and the trained classifier can realize the accuracy of 98% in the test set. Experiments show that the method of photoacoustic signal classification based on convolutional neural network greatly improves the accuracy of signal classification, and realizes the rapid and accurate analysis of a large number of data.
Melanoma, developing from melanocytes, is the deadliest type of malignant skin tumors in the world. Due to high light absorption of melanin, rare circulating melanoma cells, as an endogenous marker for metastasis at the early stage, can be quantitatively detected in small superficial vessels of mouse ears by in vivo photoacoustic flow cytometry (PAFC). Before clinical application, the capability of promising PAFC platform should be verified and optimized by mouse vessels, which are similar in size and depth to human vessels. In the current study, compared with optical resolution PAFC (OR-PAFC), we build acoustic resolution PAFC (AR-PAFC) using focused ultrasonic transducer and 1064 nm laser with lower pulse rate, leading to higher detection depth and lower laser power density in mouse models. Besides, based on laser frequency doubling and high absorption coefficient of hemoglobin at 532nm wavelength, the blood vessels can be positioned by lowcost navigation system rather than the expensive system of two coupled lasers or charged coupled device with depth limitation. We confirm that AR-PAFC can be applied to noninvasive label-free counting of circulating melanoma cells in mouse tail veins, and validated by in vitro assays using phantom models, which simulates the scattering and absorption coefficients of living tissue. These results show that AR-PAFC platform has great potential for preoperative diagnosis and postoperative evaluation of melanoma patients.
Melanoma is a malignant tumor whose circulating tumor cell (CTC) count has been shown as a prognostic marker for metastasis development. Therefore detection of circulating melanoma cells plays an important role in monitoring tumor metastasis and prevention after diagnosis. In Vivo Photoacoustic Flow Cytometry (PAFC) is established here to achieve in vivo melanoma inspection, meanwhile guarantees non-invasive and real-time detection.Accurate tumor cell detection is of great significance to achieve highly specific diagnosis and avoid unnecessary medical tests.However, the amount of data detected by PAFC is large and original photoacoustic signal is mixed with various noises.The traditional triple mean square deviation method has lower accuracy and consumes a lot of time in data processing. Here, a classification approach in photoacoustic is proposed, which could discriminate signals and noises based on features extracted from photoacoustic waves compared to normal cells using Support Vector Machines algorithm. Due to similar shape and structure of cells, the photoacoustic signals usually have similar vibration mode. By analyzing the correlations and the signal features in time domain and frequency domain, we finally choose the continuity, amplitude, and photoacoustic waveform pulse width as the features to characterize the signal.More than 600,000 samples were selected as the training set (normalized in advance), and a classifier with a precision of 85% accuracy to filter out the photoacoustic signals rapidly was trained by the support vector machine method.The classifier introduced here has proved to optimize the signal acquisition and reduce signal processing time, realizing real-time detection and real-time analysis in PAFC.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.