Accurate early prediction of tumour response to PDT using optical coherence angiography

Abstract

Prediction of tumour treatment response may play a crucial role in therapy selection and optimization of its delivery parameters. Here we use optical coherence angiography (OCA) as a minimally-invasive, label-free, real-time bioimaging method to visualize normal and pathological perfused vessels and monitor treatment response following vascular-targeted photodynamic therapy (PDT). Preclinical results are reported in a convenient experimental model (CT-26 colon tumour inoculated in murine ear), enabling controlled PDT and post-treatment OCA monitoring. To accurately predict long-term treatment outcome, a robust and simple microvascular metric is proposed. It is based on perfused vessels density (PVD) at t = 24 hours post PDT, calculated for both tumour and peri-tumour regions. Histological validation in the examined experimental cohort (n = 31 animals) enabled further insight into the excellent predictive power of the derived early-response OCA microvascular metric. The results underscore the key role of peri-tumour microvasculature in determining the long-term PDT response.

Figure 1

Perfused microvascular network architecture in a mouse ear visualized by optical coherence angiography (OCA). (a) 2D projection of 3D OCA data showing typical normal ear microcirculation; (b) OCA of CT-26 tumour growing in a mouse ear (two weeks after tumour cells inoculation); (c,d) are the corresponding white-light photographs of the mouse ear. Black dotted lines indicate the 2.5 × 2.5 mm² regions of OCA imaging.

Figure 2

OCA images of tumour and peri-tumourous perfused microvascular reaction at t = 24 hrs post PDT. (a) before PDT; (b) OCA example of a responder (no visible perfused vessels in the tumour, and extremely low PVD in peri-tumourous tissue); (c) OCA example of a non-responder (no perfused vessels inside the tumour, but many perfused vessels in peri-tumourous tissue). The tumour borders (indicated by dashed contours) were automatically segmented using machine learning pixel classification technique.

Figure 3

OCA-derived PVD changes after treatment for predicting PDT success. (a) PVD temporal evolution in the tumour, and (b) in the peri-tumourous tissue; error bars signify standard deviation. (c) percentage of correctly predicted outcomes (CPO) based on PVD values at the three post-treatment time points. Dots indicate values calculated from the experimental data, bars indicate the 95% Wilson score confidence intervals (based on the sample size (n = 31) and the number of correctly predicted outcomes).

Figure 4

Graphical summary of the OCA-based PDT success assessment scheme, based on tumour and peri-tumourous microvascular reaction at t = 24 hrs after treatment.

Figure 5

Histological assessment of treatment outcomes at t = 7 days post PDT. Left column’s two panels exhibit an illustrative responder. (a) Total tumour necrosis (>95%) in the central part of tumour (blue arrows). (b) Total necrosis of ‘normal’ ear tissue ~2 mm distant from the tumour edge (but still within the PDT treatment field). Right column’s two panels exhibit an illustrative non-responder. (c) Partial tumour necrosis (~75%) and viable tumour region. (d) Undamaged ‘normal’ ear tissue ~2 mm away from the tumour edge (but still within the PDT treatment field). Central white-light photograph shows the colour-coded orientation of the two histological sections displayed for each illustrative case.

Figure 6

OCT/OCA data acquisition and signal processing overview. (a) white light image of the tumour-bearing murine ear, with superimposed scanning pattern; colored points represent individual positions of the scanning beam, scanning locations corresponding to the opposite scanning directions are represented in different colors (blue and red); only data acquired during one-directional scanning (from left to right, blue dots on the schematic) were used for OCA processing. Note the absence of repeated scanning at the same position, unlike many OCA methods. Inset shows the details of scanning beam cross sections overlap; (b) 3D OCT volume 2.4 × 2.4 × 1.5 mm; c – 3D OCA volume constructed using high-frequency finite-impulse response filtering (FIR-filter) of the 3D image along the slow axis; d – en face Maximum Intensity Projection of the volume represented in (c); e – en face microvascular image after tumour segmentation: vessels inside the tumour are red and peri-tumourous vessels are green. In (d,e), scale bars are 0.5 mm.