Dataset

The zip files contain the following data: In the InSilico folder, 1. RawLnet: Raw binary mask of an examplar L-System 2. OpticalSimLnet: Intermediate L-Net post forward optical simulation 3. AcousticSimLnet: Reconstructed L-Net post acoustic simulation 4. FinalDenoisedLnet: Final denoised L-Net used for analysis 5. VesselnessFilteredLNet: (optional) Vesselness filtered final L-Net 6. SegmentedLNet: Final segmented L-Net using the four proposed methods (auto-thresholding, auto-thresholding + vesselness filtering, random forest, random forest + vesselness filtering) In the Phantom folder, 1. RawPhantom: Raw reconstructed string image exported from the RSOM 2. FinalDenoisedPhantom: Final denoised string image used for analysis 3. VesselnessFilteredPhantom: (optional) Vesselness filtered string image 4. SegmentedPhantom: Final segmented string image using the four proposed methods (auto-thresholding, auto-thresholding + vesselness filtering, random forest, random forest + vesselness filtering) In the InVivo folder, 1. RawRSOM: Raw reconstructed tumour image exported from the RSOM 2. FinalDenoisedRSOM: Final denoised tumour image used for analysis 3. VesselnessFilteredRSOM: (optional) Vesselness filtered tumour image 4. SegmentedRSOM: Final segmented tumour image using the four proposed methods (auto-thresholding, auto-thresholding + vesselness filtering, random forest, random forest + vesselness filtering)

The software associated with the deposited data is a series of custom codes that are hosted on publicly available repositories. Code to generate synthetic vascular trees (LNets) is available on GitHub (https://github.com/psweens/V-System). In silico photoacoustic simulations were performed using the SIMPA toolkit (https://github.com/CAMI-DKFZ/simpa). Both the trained 3D CNN to extract tumour ROIs from RSOM images (https://github.com/psweens/Predict-RSOM-ROI) and vascular TDA package are available on GitHub (https://github.com/psweens/Vascular-TDA).