Simulating photodynamic therapy for the treatment of glioblastoma using Monte Carlo radiative transport.

Significance: Glioblastoma (GBM) is a rare but deadly form of brain tumor with a low median survival rate of 14.6 months, due to its resistance to treatment. An independent simulation of the INtraoperative photoDYnamic therapy for GliOblastoma (INDYGO) trial, a clinical trial aiming to treat the GBM resection cavity with photodynamic therapy (PDT) via a laser coupled balloon device, is demonstrated. Aim: To develop a framework providing increased understanding for the PDT treatment, its parameters, and their impact on the clinical outcome. Approach: We use Monte Carlo radiative transport techniques within a computational brain model containing a GBM to simulate light path and PDT effects. Treatment parameters (laser power, photosensitizer concentration, and irradiation time) are considered, as well as PDT's impact on brain tissue temperature. Results: The simulation suggests that 39% of post-resection GBM cells are killed at the end of treatment when using the standard INDYGO trial protocol (light fluence = 200 J/cm2 at balloon wall) and assuming an initial photosensitizer concentration of 5 µM. Increases in treatment time and light power (light fluence = 400 J/cm2 at balloon wall) result in further cell kill but increase brain cell temperature, which potentially affects treatment safety. Increasing the p hotosensitizer concentration produces the most significant increase in cell kill, with 61% of GBM cells killed when doubling concentration to 10 µM and keeping the treatment time and power the same. According to these simulations, the standard trial protocol is reasonably well optimized with improvements in cell kill difficult to achieve without potentially dangerous increases in temperature. To improve treatment outcome, focus should be placed on improving the photosensitizer. Conclusions: With further development and optimization, the simulation could have potential clinical benefit and be used to help plan and optimize intraoperative PDT treatment for GBM.

A divide and conquer approach to maximise deep learning mammography classification accuracies.

Breast cancer claims 11,400 lives on average every year in the UK, making it one of the deadliest diseases. Mammography is the gold standard for detecting early signs of breast cancer, which can help cure the disease during its early stages. However, incorrect mammography diagnoses are common and may harm patients through unnecessary treatments and operations (or a lack of treatment). Therefore, systems that can learn to detect breast cancer on their own could help reduce the number of incorrect interpretations and missed cases. Various deep learning techniques, which can be used to implement a system that learns how to detect instances of breast cancer in mammograms, are explored throughout this paper. Convolution Neural Networks (CNNs) are used as part of a pipeline based on deep learning techniques. A divide and conquer approach is followed to analyse the effects on performance and efficiency when utilising diverse deep learning techniques such as varying network architectures (VGG19, ResNet50, InceptionV3, DenseNet121, MobileNetV2), class weights, input sizes, image ratios, pre-processing techniques, transfer learning, dropout rates, and types of mammogram projections. This approach serves as a starting point for model development of mammography classification tasks. Practitioners can benefit from this work by using the divide and conquer results to select the most suitable deep learning techniques for their case out-of-the-box, thus reducing the need for extensive exploratory experimentation. Multiple techniques are found to provide accuracy gains relative to a general baseline (VGG19 model using uncropped 512 × 512 pixels input images with a dropout rate of 0.2 and a learning rate of 1 × 10-3) on the Curated Breast Imaging Subset of DDSM (CBIS-DDSM) dataset. These techniques involve transfer learning pre-trained ImagetNet weights to a MobileNetV2 architecture, with pre-trained weights from a binarised version of the mini Mammography Image Analysis Society (mini-MIAS) dataset applied to the fully connected layers of the model, coupled with using weights to alleviate class imbalance, and splitting CBIS-DDSM samples between images of masses and calcifications. Using these techniques, a 5.6% gain in accuracy over the baseline model was accomplished. Other deep learning techniques from the divide and conquer approach, such as larger image sizes, do not yield increased accuracies without the use of image pre-processing techniques such as Gaussian filtering, histogram equalisation and input cropping.

Meshless Monte Carlo radiation transfer method for curved geometries using signed distance functions.

SIGNIFICANCE: Monte Carlo radiation transfer (MCRT) is the gold standard for modeling light transport in turbid media. Typical MCRT models use voxels or meshes to approximate experimental geometry. A voxel-based geometry does not allow for the precise modeling of smooth curved surfaces, such as may be found in biological systems or food and drink packaging. Mesh-based geometry allows arbitrary complex shapes with smooth curved surfaces to be modeled. However, mesh-based models also suffer from issues such as the computational cost of generating meshes and inaccuracies in how meshes handle reflections and refractions. AIM: We present our algorithm, which we term signedMCRT (sMCRT), a geometry-based method that uses signed distance functions (SDF) to represent the geometry of the model. SDFs are capable of modeling smooth curved surfaces precisely while also modeling complex geometries. APPROACH: We show that using SDFs to represent the problem's geometry is more precise than voxel and mesh-based methods. RESULTS: sMCRT is validated against theoretical expressions, and voxel and mesh-based MCRT codes. We show that sMCRT can precisely model arbitrary complex geometries such as microvascular vessel network using SDFs. In comparison with the current state-of-the-art in MCRT methods specifically for curved surfaces, sMCRT is more precise for cases where the geometry can be defined using combinations of shapes. CONCLUSIONS: We believe that SDF-based MCRT models are a complementary method to voxel and mesh models in terms of being able to model complex geometries and accurately treat curved surfaces, with a focus on precise simulation of reflections and refractions. sMCRT is publicly available at https://github.com/lewisfish/signedMCRT.

To focus-match or not to focus-match inverse spatially offset Raman spectroscopy: a question of light penetration.

The ability to identify the contents of a sealed container, without the need to extract a sample, is desirable in applications ranging from forensics to product quality control. One technique suited to this is inverse spatially offset Raman spectroscopy (ISORS) which illuminates a sample of interest with an annular beam of light and collects Raman scattering from the center of the ring, thereby retrieving the chemical signature of the contents while suppressing signal from the container. Here we explore in detail the relative benefits of a recently developed variant of ISORS, called focus-matched ISORS. In this variant, the Fourier relationship between the annular beam and a tightly focused Bessel beam is exploited to focus the excitation light inside the sample and to match the focal point of excitation and collection optics to increase the signal from the contents without compromising the suppression of the container signal. Using a flexible experimental setup which can realize both traditional and focus-matched ISORS, and Monte-Carlo simulations, we elucidate the relative advantages of the two techniques for a range of optical properties of sample and container.

Depth Penetration of Light into Skin as a Function of Wavelength from 200 to 1000 nm.

An increase in the use of light-based technology and medical devices has created a demand for informative and accessible data showing the depth that light penetrates into skin and how this varies with wavelength. These data would be particularly beneficial in many areas of medical research and would support the use and development of disease-targeted light-based therapies for specific skin diseases, based on increased understanding of wavelength-dependency of cutaneous penetration effects. We have used Monte Carlo radiative transport (MCRT) to simulate light propagation through a multi-layered skin model for the wavelength range of 200-1000 nm. We further adapted the simulation to compare the effect of direct and diffuse light sources, varying incident angles and stratum corneum thickness. The lateral spread of light in skin was also investigated. As anticipated, we found that the penetration depth of light into skin varies with wavelength in accordance with the optical properties of skin. Penetration depth of ultraviolet radiation was also increased when the stratum corneum was thinner. These observations enhance understanding of the wavelength-dependency and characteristics of light penetration of skin, which has potential for clinical impact regarding optimizing light-based diagnostic and therapeutic approaches for skin disease.

Imaging in thick samples, a phased Monte Carlo radiation transfer algorithm.

SIGNIFICANCE: Optical microscopy is characterized by the ability to get high resolution, below 1 µm, high contrast, functional and quantitative images. The use of shaped illumination, such as with lightsheet microscopy, has led to greater three-dimensional isotropic resolution with low phototoxicity. However, in most complex samples and tissues, optical imaging is limited by scattering. Many solutions to this issue have been proposed, from using passive approaches such as Bessel beam illumination to active methods incorporating aberration correction, but making fair comparisons between different approaches has proven to be challenging. AIM: We present a phase-encoded Monte Carlo radiation transfer algorithm (φMC) capable of comparing the merits of different illumination strategies or predicting the performance of an individual approach. APPROACH: We show that φMC is capable of modeling interference phenomena such as Gaussian or Bessel beams and compare the model with experiment. RESULTS: Using this verified model, we show that, for a sample with homogeneously distributed scatterers, there is no inherent advantage to illuminating a sample with a conical wave (Bessel beam) instead of a spherical wave (Gaussian beam), except for maintaining a greater depth of focus. CONCLUSION: φMC is adaptable to any illumination geometry, sample property, or beam type (such as fractal or layered scatterer distribution) and as such provides a powerful predictive tool for optical imaging in thick samples.

Development of a Predictive Monte Carlo Radiative Transfer Model for Ablative Fractional Skin Lasers.

It is possible to enhance topical drug delivery by pretreatment of the skin with ablative fractional lasers (AFLs). However, the parameters to use for a given AFL to achieve the desired depth of ablation or the desired therapeutic or cosmetic outcome are hard to predict. This leaves open the real possibility of overapplication or underapplication of laser energy to the skin. In this study, we developed a numerical model consisting of a Monte Carlo radiative transfer (MCRT) code coupled to a heat transfer and tissue damage algorithm. The simulation is designed to predict the depth effects of AFL on the skin, verified with in vitro experiments in porcine skin via optical coherence tomography (OCT) imaging. Ex vivo porcine skin is irradiated with increasing energies (50-400 mJ/pixel) from a CO2 AFL. The depth of microscopic treatment zones is measured and compared with our numerical model. The data from the OCT images and MCRT model complement each other well. Nonablative thermal effects on surrounding tissue are also discussed. This model, therefore, provides an initial step toward a predictive determination of the effects of AFL on the skin. Lasers Surg. Med. © 2020 The Authors. Lasers in Surgery and Medicine published by Wiley Periodicals LLC.

Quantifying Direct DNA Damage in the Basal Layer of Skin Exposed to UV Radiation from Sunbeds.

Nonmelanoma and melanoma skin cancers are attributable to DNA damage caused by ultraviolet (UV) radiation exposure. One DNA photoproduct, the cyclobutane pyrimidine dimer (CPD), is believed to lead to DNA mutations caused by UV radiation. Using radiative transfer simulations, we compare the number of CPDs directly induced by UV irradiation from artificial and natural UV sources (a standard sunbed and the midday summer Mediterranean sun) for skin types I and II on the Fitzpatrick scale. We use Monte Carlo radiative transfer (MCRT) modeling to track the progression of UV photons through a multilayered three dimensional (3D) grid that simulates the upper layers of the skin. By recording the energy deposited in the DNA-containing cells of the basal layer, the number of CPDs formed can be quantified. The aim of this work was to compare the number of CPDs formed in the basal layer of the skin and by implication the risk of developing cancer, as a consequence of irradiation by artificial and natural sources. Our simulations show that the number of CPDs formed per second during sunbed irradiation is almost three times that formed during solar irradiation.