KEYWORDS: Image segmentation, Ultrasonography, Performance modeling, Data modeling, Education and training, Medical imaging, Diagnostics, Image processing, Machine learning, Deep learning
PurposeAccurate segmentation of the endometrium in ultrasound images is essential for gynecological diagnostics and treatment planning. Manual segmentation methods are time-consuming and subjective, prompting the exploration of automated solutions. We introduce “segment anything with inception module” (SAIM), a specialized adaptation of the segment anything model, tailored specifically for the segmentation of endometrium structures in ultrasound images.ApproachSAIM incorporates enhancements to the image encoder structure and integrates point prompts to guide the segmentation process. We utilized ultrasound images from patients undergoing hysteroscopic surgery in the gynecological department to train and evaluate the model.ResultsOur study demonstrates SAIM’s superior segmentation performance through quantitative and qualitative evaluations, surpassing existing automated methods. SAIM achieves a dice similarity coefficient of 76.31% and an intersection over union score of 63.71%, outperforming traditional task-specific deep learning models and other SAM-based foundation models.ConclusionsThe proposed SAIM achieves high segmentation accuracy, providing high diagnostic precision and efficiency. Furthermore, it is potentially an efficient tool for junior medical professionals in education and diagnosis.
Background:
Although ultrasound imaging has been widely applied in medical diagnosis for decades, the data processing remains primitive. Traditional B-mode ultrasound imaging exhibits the amplitude of the scattered ultrasonic signals as brightness of the images, neglecting rich information delivered by the frequency modulation of the signals. While the doctors diagnose upon the whole image with their experiences, the limited local information poses significant difficulties on computer diagnosis.
Methods:
Ultrasonic harmonic imaging employs multiple frequency bands in the imaging strategy, indicating that the frequency variation in the spectrum contains nonlinear vibrations which are specific for given biological tissues. We hypothesize that detailed analysis and characterization of the spectrum enable the software to recognize the signals from different organs or from diseased regions. Wavelet transform was utilized to exhibit the ultrasonic signal in both time and frequency domain, followed by the principal component analysis which extracted the feature of the frequency. Pseudo colors, red, green, and blue, were associated with the first 3 principal components as a colorized augmentation of the ultrasound imaging.
Results and Conclusion:
In the preliminary test, each pixel of the image distinguished itself by frequency characteristics in the wavelet transform. Principal component analysis recognized the major characteristics and presented them in pseudo color images. The hypodermic layers, the kidney, and the surrounding tissues distinguished themselves clearly from one another by the color association. The ultrasonic spectral analysis and augmented visualization technique pioneered the way to intelligent ultrasonic imaging systems and computer-aided diagnosis.
High resolution ultrasound medical imaging requires high frequency transducers, which usually are known with
decreased penetration depth because of high loss in two-way-loop at high frequencies. To obtain high resolution imaging
at large depth, a dual frequency transducer was designed for contrast imaging. Specifically, a 35 MHz receiving
transducer with aperture of 0.6 mm x 0.6 mm was integrated into a 6.5 MHz transmitting transducer with aperture of 0.6
mm x 3 mm. High pressure ultrasound at low frequency was generated by the transducer to excited microbubbles in
tissue. High frequency component of the nonlinear response from microbubbles were received by the 35 MHz transducer
for high resolution imaging at a relatively large depth. The prototyped transducer showed the ability of transmitting
about 2 MPa pressure at 6.5 MHz, under an input of 5-cycle burst at 250 Vpp, which is high enough to generate nonlinear
oscillation of microbubbles. The pulse-echo test showed that the -6 dB bandwidth of the 35 MHz transducer is
34.4% and the loop sensitivity is -38.3 dB. The small aperture, dual frequency ultrasound transducers developed in this
paper are promising for high resolution ultrasound medical imaging.
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.