In Vivo Quantitative Vasculature Segmentation and Assessment for Photodynamic Therapy Process Monitoring Using Photoacoustic Microscopy

Thi Thao Mai, Su Woong Yoo, Suhyun Park, Jin Young Kim, Kang-Ho Choi, Chulhong Kim, Seong Young Kwon, Jung-Joon Min, Changho Lee2021 · DOI

TL;DR

This paper introduces a novel photoacoustic microscopy (PAM) approach for real-time, label-free monitoring of blood vessel changes during photodynamic therapy (PDT) in melanoma tumors. By combining advanced image segmentation with quantitative analysis of vascular parameters, the method offers a powerful tool for assessing PDT efficacy and optimizing treatment strategies.

What Problem Does This Paper Address?

Photodynamic therapy (PDT) is a promising cancer treatment that relies on damaging tumor vasculature to starve cancer cells. However, effectively monitoring the subtle, real-time changes in these micro-vessels during PDT has been a challenge. Existing imaging techniques often have limitations such as radiation exposure (CT, PET), toxicity (MRI contrast agents), low resolution, or shallow penetration depth. This gap hinders the ability to precisely assess PDT's impact on tumor vasculature and tailor treatments for optimal outcomes. The authors aim to address this by developing a high-resolution, label-free imaging method that can quantitatively evaluate vascular damage in vivo.

Methods

The core of this work lies in the integration of optical-resolution photoacoustic microscopy (OR-PAM) with sophisticated image analysis techniques.

  • OR-PAM System: A lab-built OR-PAM system was employed, utilizing a 532 nm pulsed laser for excitation and a high-frequency transducer (50 MHz) for detecting photoacoustic signals. This setup provides high spatial resolution (27 µm axial, 14 µm lateral) and a penetration depth of approximately 1 mm, suitable for imaging micro-vessels in the mouse ear.
  • Quantitative Image Analysis: A multi-stage process was developed for quantitative assessment:
  • Maximum Amplitude Projection (MAP): 3D OR-PAM data was processed to create 2D MAP images, representing the maximum PA signal intensity.
  • Adaptive Hessian Filter Segmentation: A multi-scale Hessian filter was applied to enhance blood vessel structures of various sizes, followed by adaptive thresholding to generate a binary image where blood vessels are highlighted.
  • Skeletonization: The binary image was skeletonized to obtain a single-pixel-wide representation of the vessel centerline.
  • Quantitative Parameter Extraction: Five key vascular parameters were calculated:
  • PA Signal: Overall signal intensity, indicative of material density.
  • Vessel Diameter (VDI): Determined using Euclidean distance transform on the skeletonized image.
  • Vessel Density (VD): The ratio of yellow pixels (vessels) to total pixels in the binary image.
  • Perfused Vessel Density (PVD): The ratio of red skeleton pixels to total skeleton pixels, representing the extent of functional vasculature.
  • Vessel Complexity (VC): Calculated using the box-counting method to quantify the tortuosity and branching of the vascular network.

Datasets

The study utilized a mouse ear melanoma tumor model for in vivo experiments. The dataset consists of OR-PAM images acquired before, during, and after photodynamic therapy (PDT) treatments. Specifically, images were collected at various time points: before rose bengal (RB) injection, after RB injection, after the first PDT, and after the second PDT. The data was analyzed from both whole-image perspectives and from four selected regions of interest (ROIs) representing control vessels, normal tissue, and the two melanoma tumors. No new benchmark datasets were introduced; rather, the study generated novel in vivo data for evaluating the proposed method.

Key Insights and Ideas

  • Label-Free Vascular Monitoring: The ability to visualize and quantify vascular changes without exogenous contrast agents is a significant advantage, reducing potential toxicity and simplifying the experimental procedure.
  • Quantitative Assessment of PDT Efficacy: Moving beyond qualitative observations, the paper demonstrates that quantifying parameters like vessel diameter, density, and perfusion provides a more objective and sensitive measure of PDT's impact on tumor vasculature.
  • Differential Response of Tumors and Normal Vessels: The study highlights that PDT can selectively damage tumor vasculature while minimizing harm to surrounding normal vessels, a key goal of effective cancer therapy. However, it also reveals that normal vessels adjacent to tumors can be inadvertently affected.
  • Multi-Parameter Analysis is Crucial: Relying on a single vascular parameter is insufficient. The combination of PA signal, VDI, VD, PVD, and VC provides a comprehensive picture of vascular health and damage, capturing different aspects of the therapeutic response.
  • Hessian Filter for Robust Segmentation: The adaptive Hessian filter proves effective in enhancing and segmenting micro-vessels, even in complex tumor environments, forming the foundation for reliable quantitative analysis.

Key Results

  • Tumor Destruction: The study successfully demonstrated the destruction of both melanoma tumors after two PDT treatments, visualized through the disappearance of PA signals and significant reduction in vascular parameters.
  • Vascular Parameter Changes:
    • PA Signal: Showed a significant decrease in tumor ROIs after PDT.
  • Vessel Diameter (VDI): Decreased significantly, particularly after the first PDT, indicating vasoconstriction and vessel damage.
  • Vessel Density (VD): Exhibited the largest change, with substantial decreases in tumor areas, reflecting the elimination of blood vessels.
  • Perfused Vessel Density (PVD): Also showed significant reductions, indicating a loss of functional blood flow.
  • Vessel Complexity (VC): Showed the least variation, suggesting that while vessel structure is altered, the overall complexity metric is more robust to minor changes.
  • Selective Damage: While tumor vessels were severely impacted, normal vessels in ROI A (control) showed minimal damage. However, normal vessels adjacent to melanoma 1 (ROI B) were also affected, with a 43% decrease in density and 26% decrease in PVD after PDT.
  • Melanoma 2 (ROI D) Impact: This tumor experienced the most severe impact, with complete obliteration observed after the second PDT.

Limitations and Future Work

  • Penetration Depth: The ~1 mm penetration depth of OR-PAM limits the evaluation of larger or deeper tumors. The authors suggest exploring AR-PAM (acoustic-resolution PAM) or integrated OR-AR-PAM systems for deeper imaging.
  • 2D Hessian Filter Limitations: The 2D Hessian filter can sometimes misclassify melanoma regions as vessels. Future work could involve 3D Hessian filtering or separating vessel and tumor regions before filtering for more accurate analysis.
  • Lack of Functional Parameters: The current system, using a single laser wavelength, cannot monitor oxygen saturation (SO2) or blood flow. Incorporating multiple laser wavelengths would enable the acquisition of these crucial functional parameters for a more comprehensive PDT assessment.
  • Pilot Study: The study was conducted on a single mouse, limiting the generalizability of the findings. Future work should involve larger cohorts of mice to statistically validate the results and assess PDT effectiveness across different tumor sizes and types.
  • Cost and Portability: While PAM offers advantages, the cost and size of current laser sources can be a barrier to clinical translation. The authors suggest exploring laser diodes or LEDs as more cost-effective and compact alternatives.

Summary updated

BibTeX
@misc{mai2021vivo,
  author = {Thi Thao Mai and Su Woong Yoo and Suhyun Park and Jin Young Kim and Kang-Ho Choi and Chulhong Kim and Seong Young Kwon and Jung-Joon Min and Changho Lee},
  title = {{In Vivo Quantitative Vasculature Segmentation and Assessment for Photodynamic Therapy Process Monitoring Using Photoacoustic Microscopy}},
  year = {2021},
  doi = {10.3390/s21051776},
}
Figure