Investigating Temporal Features of Carotid Intima-Media Thickness from Ultrasound Imaging with Recurrent Neural...

Area of Science:

• Medical imaging analysis
• Cardiovascular disease research
• Artificial intelligence in healthcare

Background:

• Carotid intima-media thickness (cIMT) measurement via ultrasound is crucial for cardiovascular risk assessment.
• Current automated methods using Convolutional Neural Networks (CNNs) often overlook temporal information in ultrasound sequences.
• Exploiting temporal dynamics in cIMT could enhance diagnostic capabilities.

Purpose of the Study:

• To develop a novel framework for analyzing temporal features of cIMT using ultrasound imaging.
• To investigate the application of Recurrent Neural Networks (RNNs) for Region of Interest (ROI) detection in cIMT analysis.
• To explore the potential clinical relevance of time-series cIMT data.

Main Methods:

• Utilized Recurrent Neural Networks (RNNs) for Region of Interest (ROI) detection on consecutive ultrasound frames.
• Generated cIMT time series from sequential ultrasound scans.
• Extracted statistical features (min, max, mean, frequency) from the cIMT time series.
• Validated the method by estimating heart rate from cIMT time series and comparing with clinical records.

Main Results:

• Recurrent Neural Networks (RNNs) demonstrated a 4.75% improvement in ROI detection performance compared to Convolutional Neural Networks (CNNs).
• Estimated heart rates from cIMT time series showed high correlation with patients' clinical data.
• The study successfully extracted additional information from the temporal dynamics of cIMT.

Conclusions:

• The proposed RNN-based framework effectively leverages temporal information in ultrasound imaging for cIMT analysis.
• Temporal features of cIMT provide valuable insights beyond static measurements, enhancing cardiovascular risk assessment.
• The findings suggest significant potential for cIMT time series analysis in future clinical studies and cardiovascular monitoring.