Additional file 1 of Mapping of prostate cancer microvascular patterns using super-resolution ultrasound imaging

Additional file 1: Fig. S1. Zoomed in density map of the clinical data set. The black star is the initial microbubble (MB) location in frame t and white-centre star is the MB location at frame t+1. The line between them corresponds to paired MBs when no rejection criteria are applied. The black circle corresponds to a correct link within a high MB density region (indicates a vessel), whereas the links circled in white correspond to incorrect pairing with MBs “jumping” from one vessel to the next passing through a low MB density region, i.e., inconsistent density within the formed “track”. Fig. S2 (a) The simulated network structure, (b) an example individual frame of the synthetic contrast-enhanced ultrasound video data (c) The microbubble speed maps calculated from the synthetic data set using nearest neighbour (NN), forward model (FM) and vascular model (VM) versions of the tracking algorithm. (d) Average intensity profile through the selected region in part (c) for each processing method. Each vessel is better resolved in the VM version compared to the NN and FM versions. (e) scatter plot comparing average full-width half maximum (FWHM) for each peak in part (d). FM (circles) and VM (x) methods are compared to NN (x-axis). The FWHM calculated by NN and FM models are very similar (mean NN 179.6 ± 25 μm mean FM 179.0 ± 26 μm) (p = 0.5), following y = x, while the VM consistently model produces narrower vessels (mean 140.7 ± 16 μm) (p = 0.001). Fig. S3. Displaying 25–90% of the peak intensity value from the track number maps of the central Corpus Luteum (CL) region (see Figure 2) for Nearest Neighbour (NN) and Vascular model (VM) methods. The central CL is known to comprise very small vessels. The VM version of the code creates more links within this region and shows vessels that are likely to be between 200 and 50 μm in diameter. The NN version produces a few tracks and covers an area of 34 mm2. The area covered inside the CL with VM is 82.2 mm2, showing an increase of 137%. Table S1 The specifications of the computational synthetic Vascular Network (detailed Figure S2 panel a). A tube-diameter range wider than 1.5 mm is provided, which corresponds to tubes with realistic blood speeds ranging between 3.8 and 23.87 mm/sec. Table S2 MRI and Ultrasound specifications[12] Abbreviations: AT = acquisition time, mpMRI = multiparametric magnetic resonance imaging, TR = Repetition Time, TE = Echo Time, FOV = Field of View. *(Invivo, Gainesville, FL, USA). Table S3 Synthetic data. Outputs comparing the detections and links made by 3 different versions of the tracking algorithm: Nearest Neighbour (NN), Forward Model (FM), Vascular Model (VM). Table S4 Comparison of detection and linking outputs from contrast-enhanced ultrasound video data from the sheep corpus luteum (CL) for nearest neighbour (NN) version and the Vascular model (VM) version. Table S5 Quantitative output from Nearest Neighbour (NN) and Vascular Model (VM) tracking models from participant 127 Mid-Apex (Fig.3 main document). Table S6 Processing outputs for P127 vascular model (VM) and nearest neighbour (NN) processing for cancer and non-cancer regions (ROI 1 and 2 in Fig 3). Table S7 Output for both nearest neighbour (NN) and vascular model (VM) tracking models for P134 (Fig 4 main document). Also included are the numerical outputs (VM) from smaller regions of interest for cancer and non-cancer regions of the mid-base plane. Table S8 Full summary of all clinical prostate data processing using super-resolution ultrasound imaging. Each data set that has been processed is represented, and those where unusual regions of low, high or fast flow were noted (x). Entries in bold are associated with a clinical diagnosis of Gleason 7 cancer by biopsy. The super-resolution method appears to have good sensitivity in that areas of faster flow can be identified and these are consistently associated with regions of cancer, further work is needed to fully assess the specificity and identify regions of the prostate where faster flow is normal.