Doppler ultrasound (DU) measurements are used to detect and evaluate venous gas emboli (VGE) formed after decompression. Automated analysis methodologies of these VGE have been developed on varying real-world datasets of limited size and without ground truth values preventing objective evaluation. We develop and report a method to generate synthetic post-dive data using DU signals collected in both precordium and subclavian vein with varying degrees of bubbling matching field-standard grading metrics. This method is adaptable, modifiable, and reproducible, allowing for researchers to tune the produced dataset for their desired purpose. We provide the baseline Doppler recordings and code required to generate synthetic data for researchers to reproduce our work and improve upon it. We also provide a set of pre-made synthetic post-dive DU data spanning six scenarios representing the Spencer and Kisman-Masurel (KM) grading scales as well as precordial and subclavian DU recordings. By providing a method for synthetic post-dive DU data generation, we aim to improve and accelerate the development of signal processing techniques for VGE analysis in Doppler ultrasound. 

Baseline data:

1) Human Precordial Doppler Data: The study collected precordial Doppler ultrasound recordings from 16 healthy volunteers who gave informed consent and were free of heart murmurs. The recordings were obtained using two different Doppler ultrasound systems, Continuous and Pulsed wave (CW and PW) with the same precordial transducer. The audio signals were recorded using an Analog-to-Digital converter and Audacity software. Ten subjects were measured with both CW and PW systems, and five were measured using only the CW system. Each recording lasted between 3:30 to 7 minutes. The study was approved by the Duke Health Institutional Review Board.

2) Human Subclavian Doppler Data: Subclavian Doppler ultrasound data was collected from 75 healthy volunteers. The data was de-identified and the study was approved by the DAN institutional review board (#024-19-22) with informed consent provided by the subjects. The O'Dive continuous wave Doppler device was used to perform the subclavian Doppler ultrasound, with measurements taken on both the left and right subclavian veins at rest. The data was recorded on an iPad at a sampling frequency of 48 kHz, and each recording lasted between 18-20 seconds.

3) Bubble-only Doppler recordings: A tissue-mimicking flow phantom was created using porcine gelatin, 1-propanol, and distilled water. A peristaltic pump was connected to the phantom vessel inlet and outlet, with a water reservoir to allow bubbles to dissipate without recirculation. Doppler ultrasound measurements were acquired using a Siemens/Acuson Sequoia C512 with a transducer set to pulse-wave Doppler mode, and the Doppler audio output was recorded using Audacity with a sampling frequency of 44.1 kHz. The pump was set to various flow rates, and the transducer was placed at different angles relative to the water flow. An air-filled syringe was used to inject individual bubbles into the flowing water, with recordings of 25-second duration taken for each parameter combination. An additional experiment was performed with the pump flow randomly varying between 540-900 mL/min at each specified angle, producing a total of 45 recordings. Individual bubbles were isolated using a thresholding algorithm and saved as a .mat file.

Fully synthetic data:

Using the baseline human Doppler audio, a procedure was created that performs a mixing of audio with bubbles inserted into the blood flow signals. The amount of bubbling was determined using a modified Kisman-Masurel scale and includes data augmentation to increase the variability for bubbles and cardiac sounds. Different processing methods were employed for precordial such that bubbles can be found in any part of the cardiac cycle or can only be heard in between cycles (to represent bubbles that are most likely to be heard).

Data was generated using 2 grading scales: Kisman-Masurel (22 classes) or Spencer (5 classes).

To generate synthetic data Matlab must be used. Otherwise, all data is saved as .wav files and can be opened with any audio processing software.