This PDF file contains the front matter associated with SPIE Proceedings Volume 12628, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

A novel technique to treat different diseases from inflammation to poisonous bites from snakes on small animals is the hyperbaric chamber treatment [1]. Non-invasive and real-time hemodynamic monitoring of patient’s tissue could be useful to quantify the effect of oxygen therapy on the patient. In this pilot study, we explored the possibility of noninvasively detecting canine tissues optical properties by Time Domain Near-Infrared Spectroscopy (TD-NIRS) and then retrieving hemodynamic parameters (deoxygenated and oxygenated hemoglobin molar concentration and tissue oxygen saturation) on different tissues (Triceps Brachii, Biceps Femoralis, and Head) of dogs. Four dogs with different hair length and color undergoing to hyperbaric chamber treatment were measured before and after the treatment, on each of the three sites and on both sides of the animal. In Triceps Brachii and Biceps Femoralis we found an increase in the absorption coefficient for both wavelengths after the treatment, meaning that the total concentration of blood has increased. Different results were obtained in the head, where the total hemoglobin concentration is decreased. The use of TD-NIRS oximetry technology seems a clinically feasible means to assess tissue oxygenation in most of dogs, thanks to a sufficiently high signal-to-noise ratio that allows to evaluate the optical parameters and consequently the physiological parameters of the area under investigation. Moreover, the presence of hair and dark skin did not prevent the possibility of obtaining robust readings.

We present initial evidence of the SOLUS potential for the multimodal non-invasive diagnosis of breast cancer by describing the correlation between optical and standard radiological data and analyzing a case study.

This study aims at evaluating the effect of NeoAdjuvant Chemotherapy (NAC) on the contralateral tumor-free breast tissue through time domain diffuse optical spectroscopy. The breast tissue composition consisting of hemoglobin, water, lipid, and collagen concentrations is quantitatively derived using our seven-wavelength (635-1060 nm) optical mammograph. Preliminary analysis of ten patients' data shows compositional changes occurring in the non-tumor breast in addition to the tumor breast. This includes reduction in breast density and components’ concentrations through the course of the therapy. The final goal is to eventually identify if there is a correlation of these effects with pathological complete response.

Non-invasively monitoring tumors during their growth and disease progression could provide invaluable diagnostic information and improve our understanding of tumors and their microenvironment, especially blood vessels. Hyperspectral imaging (HSI) with integrated three-dimensional optical profilometry (3D OP) provides the necessary tools for non-invasive and contactless disease diagnosis by utilizing intrinsic tissue contrast of incoming visible and near-infrared light. Therefore, information about tissue, morphology, and pathology could be extracted from the images. In this study, we monitored six female BALB/c mice with a subcutaneously grown CT26 murine colon carcinoma over a period of 14 days, starting on the day of tumor cells injection. Blood vessels in the tumor and its surrounding healthy tissue were segmented from hyperspectral images, and physiological properties such as blood volume fraction and tissue oxygenation were extracted using the inverse adding-doubling (IAD) algorithm. The results indicate that oxygenation in blood vessels within the CT26 tumors and surrounding tissue peaks eight days after tumor cell injection at 35 %, a two-fold increase from the beginning of the study, and then gradually decreases to around 25 % 14 days after injection.

Colorectal cancer (CRC) is the third most common and second most deadly type of cancer worldwide. Late-stage detection of CRC occurs in most cases and leads to a large mortality due to poor prognosis. Mortality and poor prognosis are partially caused by cancer recurrence and postoperative complications, which should be improved for increased patient survival. Therefore, patient survival could be increased by using accurate surgical guidance tools based on diffuse reflectance spectroscopy (DRS). DRS enables real-time tissue identification for potential cancer margin delineation through determination of the circumferential resection margin (CRM), while also supporting non-invasive and label-free approaches for laparoscopic surgery to avoid short-term complications of open surgery as suitable. In this study, we have estimated the scattering properties and chromophore concentrations based on 2949 DRS measurements of freshly excised ex vivo specimens of 47 patients, and used this estimation to classify colorectal wall (CW), fat and tumor tissues. DRS measurements were performed with fiber-optic probes of 630-μm source detector distance (SDD; probe 1) and 2500-μm SDD (probe 2) to measure tissue layers from ~0.5-1mm and from ~0.5-2 mm deep, respectively. By using the 5-fold crossvalidation of machine learning models generated with the classification and regression tree (CART) algorithm, we achieved 95.9 ± 0.7% sensitivity, 98.9 ± 0.3% specificity, 90.2 ± 0.4% accuracy, and 95.5 ± 0.3% AUC for probe 1. Similarly, we achieved 96.9 ± 0.8 % sensitivity, 98.9 ± 0.2 % specificity, 94.0 ± 0.4 % accuracy, and 96.7 ± 0.4 % AUC for probe 2. To the best of our knowledge, this is the first study to evaluate the tissue chromophore concentrations and scattering properties in both superficial and deeper tissue layers based on tissue surface measurements for application in CRC detection during open surgery, laparoscopic surgery and/or robotic surgery. Aiming for the same application, this study is also the first study to use Monte Carlo look-up table methods to extract such values based on the ultraviolet, visible and near-infrared wavelength ranges (350-1920 nm).

Fetal hypoxic brain injury is the deprivation of oxygen during labour and is associated with up to 60% mortality. The gold standard of fetal monitoring during labour, the cardio tocograph (CTG) and fetal blood sampling are poor at diagnosing hypoxia continuously and non-invasively. Our research is towards developing a non-invasive, continuous hypoxia assessment system using long wavelength near-infrared spectroscopy through a fiber optic based reflectance. Lactate is a key biomarker for hypoxia determination in babies during birth. For successful implementation of this probe, it is required that it detects lactate in maternal environment and in presence of other spectroscopic interferences. In this paper we look at lactate sensing through a liquid phantom containing spectrally interfering components alongside lactate like glucose, urea, triacetin and albumin. Through these experiments we determine the relevant wavelengths and their combination for effective lactate sensing.

Optical biomarkers of neonatal hypoxic ischemic (HI) brain injury can offer the advantage of continuous, cot-side assessment of the degree of injury; research thus far has focused on examining different optical measured brain physiological signals and feature combinations to achieve this. To maximize the breadth of physiological characteristics being taken into consideration, a multimodal optical platform has been developed, allowing unique physiological insights into brain injury. In this paper we present an assessment of severity of injury using a state-of-the-art hybrid broadband Near Infrared Spectrometer (bNIRS) and Diffusion Correlation Spectrometer (DCS) instrument called FLORENCE with a machine learning pipeline. We demonstrate in the preclinical neonatal model (the newborn piglet) that our approach can identify different HI insult severity (controls, mild, severe). We show that a machine learning pipeline based on k-means clustering can be used to differentiate between the controls and the HI piglets with an accuracy of 78%, the mild severity insult piglets from the severe insult piglets with an accuracy of 90% and can also differentiate the 3 piglet groups with an accuracy of 80%. So, this analytics pipeline demonstrates how optical data from multiple instruments can be processed towards markers of brain health.

Critically ill neonates are subjected to severe and abrupt hemodynamic imbalances that could cause cerebral damage (e.g., hemorrhage, ischemic events). In this population, the need for cerebral monitoring tools to identify dangerous hemodynamic variations in real-time is paramount. Near-infrared spectroscopy (NIRS) is largely exploited in neonatal intensive care units (NICU) to monitor critically ill patients' cerebral oxygenation. However, the most used NIRS devices exploit continuous wave NIRS (CW NIRS) technique, which is affected by motion artifacts and has low penetration depth compared to other more complex NIRS techniques. Moreover, CW-NIRS does not allow the investigation of tissue perfusion. Thus, in this ongoing study, we tested a hybrid device that combines time-domain NIRS (TD-NIRS) and diffuse correlation spectroscopy (DCS) for monitoring absolute cerebral total hemoglobin concentration (tHb), cerebral tissue oxygen saturation (S_tO₂), and cerebral blood flow index (BFI) of piglets during induced hemodynamic variations. Cerebral hemodynamic variations were induced during extracorporeal membrane oxygenation (ECMO) procedure, simulating four common conditions affecting the cerebral hemodynamics of ill neonates: hypocapnia, hypercapnia, hypotension, and hypertension. We measured 4 piglets, and the preliminary results shown in this study are promising, obtaining hemodynamic variations in accordance with previous findings, and empowering the possibility to exploit hybrid TD-NIRS and DCS devices to assess cerebral health of ill neonates.

We present a simulation study to evaluate the feasibility of using Time Domain fNIRS to monitor hemodynamic oscillations in biological tissues like the cerebral cortex. Two geometries (slab and two-layer medium) were considered to define the optimal acquisition parameters and to assess the ability of the technique to detect and separate oscillations occurring at different depths within the probed medium by exploiting the time-gating of TD fNIRS signals.

This work aims to show the possibility to perform in-vivo acquisitions with high sampling rate (20 Hz) with Time Domain functional Near-Infrared Spectroscopy (TD fNIRS) for studying brain resting state oscillations. Based on numerical simulations, a protocol was designed for acquiring hemodynamics parameters on 13 healthy volunteers during normal and forced respiration. Both the experiments had a length of 15 minutes and during the forced respiration one, subjects were ask to breath at 5 breaths per minute (0.083 Hz) following a metronome. Systemic (UP) and cortical (DW) oxy- (O₂Hb) and deoxy- (HHb) hemoglobin concentrations (absolute values) were successfully retrieved with a single measure on the frontal lobe. Temporal series and Power Spectral Density (PSD) were calculated for: physiological signals (electrocardiogram, breath signal, blood volume pulse, skin conductance and temperature), total counts at the two wavelengths (RED = 689.5±0.5 nm and IR=828.5±0.5 nm), counts in temporal gates for RED and IR, absolute values of O₂Hb_UP, O₂Hb_DW, HHb_UP and HHb_DW. Specific characteristic peaks were evaluated in the cardiac, respiratory, low, and very low frequency bands. The behavior among the subjects was uniform and differences between the two experiments were found.

Functional near infrared spectroscopy (fNIRS) delivers a flexible, portable, and wearable technique for monitoring brain function in situations where fMRI is not feasible, not suitable, or inaccessible. However, variations in optode locations and head shapes and sizes throughout development lead to considerable challenges in group-based and longitudinal studies that generally use either channel-focused analyses or image reconstruction techniques that require strong participant-atlas correspondence. We present a scalp-based parcellation technique that compensates for variation in optode array placement and general head morphology and accounts for fNIRS spatial sampling with minimal assumptions about the underlying head and brain structure to support robust statistical analyses.

Anemia is a common problem in preterm neonates, and red blood cell transfusion (RBCT) is used to improve oxygen delivery. In order to limit the risk of possible complications new strategies to minimize the need for RBCTs are needed, as assessment of hemoglobin concentration in blood ([Hb]) alone appears to be an inadequate biomarker. In this study, we search for hemodynamic and metabolic thresholds to help define the need of RBCT in anemic newborns. The effect of RBCTs on cerebral tissue oxygen saturation (S_tO₂) and blood flow (measured as Blood Flow Index, BFI) was estimated using a non-invasive hybrid diffuse optical device that combines Time Domain NIRS (TD-NIRS) and Diffuse Correlation Spectroscopy (DCS) techniques (BabyLux device). We enrolled 18 clinically stable neonates receiving RBCT at Neonatal Intensive Care Unit (NICU) of Ospedale Maggiore Policlinico in Milan. Tissue oxygen extraction (TOE) and the cerebral metabolic rate of oxygen consumption index (CMRO₂I) were computed, the Wilkinson signed rank test for paired data was performed to compare data before and after RBCT. Preliminary results are in accordance with previous publications as regards cerebral oxygenation: a significant increase in S_tO₂ (from 56.62 ± 5.20% to 63.85 ± 4.95%, p<0.05) and reduction in TOE (from 41.35 ± 5.9 % to 31.04 ±5.41%, p<0.05) were observed. The response in cerebral blood flow was smaller (only 10%) but also more variable, so conclusions regarding the effect of transfusion on cerebral oxygen metabolism are still uncertain.

Optical methods are well-suited for non-invasive bedside brain imaging: near-infrared spectroscopy (NIRS) for measuring brain oxygenation and diffuse correlation spectroscopy (DCS) for measuring cerebral blood flow. However, data obtained with those optical techniques are prone to signal contamination from extracerebral tissue. This study aimed to evaluate extracerebral signal contamination in trNIRS/multidistance DCS data acquired during transient hypotension and assess suitable means of separating scalp and brain signals. An in-house built hybrid system was used to acquire oxygenation and blood flow data simultaneously during transient orthostatic hypotension induced by rapid-onset lower body negative pressure (LBNP). In nine healthy young adults, LBNP significantly decreased arterial blood pressure (-18 ± 14%), scalp blood flow (-36 ± 25%), and scalp tissue oxygenation (all p ≤ 0.04 vs baseline). In contrast, LBNP had no significant effect on cerebral blood flow or oxygenation. This finding was confirmed by transcranial Doppler ultrasound, which found no significant changes (-8 ± 16%) in middle cerebral artery blood velocity during LBNP. These results demonstrate the importance of using depth-enhanced methods when applying these optical technologies to physiological paradigms designed to test cerebral autoregulation that also affect systemic physiology.

Optical imaging is a non-invasive technique that is able to monitor hemodynamic and metabolic responses during neurosurgery. However, a robust quantification is complicated to perform. To overcome this issue, phantoms that mimic biological tissues are required for the development of imaging systems in order to reach a true standardization. In this work, we explore the possibility to use a combined liquid blood phantom with cytochrome contained yeast to evaluate the reliability of hyperspectral imaging to measure oxygenation and metabolic changes. This phantom can be used to verify the reliability of intraoperative optical setups before moving on to clinical application.

We propose a novel analytical time-domain model for migration of Raman scattered photons in inhomogeneous two-layer diffusive media. Based on this model, the methods for reconstruction of the Raman spectra of the two layers are developed, tested in simulations and validated on phantom measurements data.

In this contribution we show a novel application of Fourier space analysis to determine overall order of tissue architecture in samples of murine abdominal walls from an animal study of induced peritonitis. Such metrices present a necessary steppingstone in the development of digital histology, facilitating both longitudinal as well as large screening studies through introduction of quantitative and rapidly computable biomarkers. The samples were collected from a study including three different experimental groups receiving different intraperitoneal injections; a control group, which received a phosphate-buffered saline solution, a pure model group, receiving chlorhexidine gluconate to induce peritonitis, and a treated model group, which was additionally administered resolvin D1. We present an approach for analysis of hematoxylin and eosin stained histological slices from this study. First, hyperspectral microscopy images of the slides are acquired, and Beer-Lambert law is used to calculate relative volume fraction maps of both stains. Following this decomposition, a Fourier space representation of the eosin cellular image is calculated. Differences between healthy and diseased subjects, which are attributed to necrosis and edema, are then quantified in the Fourier space through the ellipticity of spatial frequency distribution. The proposed metric is shown to significantly discriminate between the healthy and diseased subject (p=0.02) and is additionally strongly linearly correlated to scattering parameter b on macroscopic scale observed in our previous analysis employing macroscopic imaging of whole abdominal walls (linear correlation coefficient of -0.9).

Infant head injuries and damage can be caused by various factors such as tumors or physical trauma. The treatment for a head injury will depend on the severity of the damage. Nevertheless, the infant’s head should be imaged before any treatment. High-density diffuse optical tomography (HD-DOT) is a non-invasive imaging technology that can be employed for subsurface imaging of the infant brain. However, there are problems with HD-DOT, such as low resolution, ill-posedness of the inverse problem, and high computational costs. In this study, to improve subsurface imaging of the infant head, an extreme gradient boosting (XGBoost) algorithm is combined with HD-DOT. The proposed method is then used to detect subsurface anomalies in the infant head. The proposed method achieves a similarity index greater than 0.97 in terms of cosine similarity and less than 0.12 in terms of the root mean square error, demonstrating its effectiveness. Moreover, the proposed method requires a minimal dataset compared to conventional deep learning methods and consumes significantly less time to train. The results of this study suggest that the proposed method can provide a promising alternative for subsurface imaging of the infant head, which could significantly impact the medical imaging field in the future.

Diffuse Optical Tomography (DOT) is an imaging method that utilizes near-infrared light to image the human body. The image is acquired by inverting a propagation model based on measuring the scattered fields. One of the most promising methods to solve DOT is deep learning, due to its high computational power and fast inference. However, the ill-posed nature of the problem and the non-uniqueness of its solution can pose challenges in learning the inverse transformation. To address these challenges, two main approaches are used: learning the forward mapping and then applying classical inversion techniques, or directly learning the inverse mapping. In this study, both approaches were applied on a realistic breast model to evaluate the impact of dataset size on the optimal technique for deep learning-based Diffuse Optical Tomography. The results showed that for the specific scenario the direct inversion approach was superior for large datasets (over 2K), improving the RMSE and CNR by 44% and 97% respectively. Conversely, the forward optimization approach was better for smaller datasets (150), improving both the RMSE and CNR by 10% and 67% respectively.

A high density speckle contrast optical tomography system is introduced for in vivo cerebral blood tomography of the adult head.

Near-infrared optical tomography (NIROT) is a non-invasive imaging technique based on diffuse propagation of light in turbid media. It has high sensitivity to tissue oxygenation, which is a vital biomarker. NIROT enables one to quantify and image oxygenation at a bed-side in clinics, thus complementing other imaging modalities. Time-domain (TD) NIROT systems benefit from time-of-flight (ToF) information, but are often affected by relatively low signal-to-noise ratio and long acquisition time. However, fast acquisition is needed for in vivo assessment of oxygenation. In this work we present a time-multiplexing approach which enables multiple-fold faster acquisition with high SNR, suitable for various ToF applications, including NIROT. We combine it with hybrid convolutional neural-network (hCNN)-enhanced reconstruction to achieve an impressive 11-fold increase in acquisition speed.

Quantitative depth of tumor invasion in early gastric cancer cases via the scattering of circularly polarized light was computationally and experimentally estimated. Invasion depth information of cancer in early stages, which is difficult to obtain accurately using current biological imaging technologies, is very beneficial for cancer diagnosis. The circularly polarized light scattering method, proposed as a novel biomedical evaluation technique, can provide depth profiles by varying the incident or detection angles as well as wavelengths of circularly polarized light. Incident circularly polarized light is gradually depolarized owing to multiple scattering against cell nuclei in biological tissues. Since the degree of depolarization significantly depends on the size and shape of the scattering particles, the enlargement of cell nuclei in the scattering volume due to cancerization can be investigated by assaying the polarization of scattered light. This paper reports the results of Monte Carlo simulations and experiments to demonstrate cancer depth estimation using this technique. The Monte Carlo simulation for bilayered pseudo-tissues with a cancerous layer on a healthy layer indicates that the degree of circular polarization of scattered light shows systematic changes depending on the thickness and depth of cancerous tissues. Additionally experiments with bilayered biological tissues exhibits the same tendency as the simulation results.

We present the application to time-domain diffuse optics of a device composed of 8x256 CMOS SPAD array with 256 7-bit time-to-digital converters. Thanks to its structure and despite the limitation on the maximum repetition rate of the laser (2 MHz), it has been demonstrated to be suitable for fast acquisitions (10 ms) provided that a high photon count-rate is used and pile-up distortion is corrected. We demonstrate that high penetration depth (>30 mm) and good linearity in absorption coefficient retrieval can be achieved. Finally, we were able to clearly record the heart beat in a resting state forehead measurement.

Time domain diffuse correlation spectroscopy (td-DCS), has been proposed as a method that can increase the sensitivity of DCS for detecting blood flow index (BFI) in deep tissue. Several important parameters including the instrument response function (IRF), gate start time, gate width, and source-detector separation (SDS) must be taken into consideration. In our study, we characterized td-DCS system at three different SDS values and assessed each SDS's ability to detect dynamic changes of blood flow caused by moving red blood cells during cuff occlusion.

Bioresorbable materials have gained interest for implantable optical components such as fibers for medical devices and have been demonstrated as suitable to perform diffuse optical measurements. In this work, we demonstrate interstitial, broadband, time-domain diffuse optical spectroscopy measurements using bioresorbable fibers, by employing a single-photon avalanche diode operated in an ultrafast time-gate mode for photon detection. Using tissue equivalent liquid phantoms, we test the system absorption linearity as per the MEDPHOT protocol and demonstrate the scattering independent absorption retrieval of the water spectrum in the 600-920 nm range. Consequently, we also attempt to distinguish the spectral changes due to the presence of optically denser speck inclusion in a tissue equivalent liquid phantom.

Optical imaging in the second near-infrared (NIR-II, 1000–1700 nm) is suitable for visualizing tumor vascular imaging and targeting bone structures in small animals. In this study, we developed a novel three-dimensional (3D) NIR-II fluorescence rotational stereo imaging system for conducting 3D tumor vascular imaging and bone imaging in mice with only one NIR-II camera and a rotational stereo vision technique. Our system utilizes a 3D blood vessel reconstruction algorithm to present high-resolution 3D blood vessel maps and skeletal systems through two camera views. This technique allows for the visualization of bone structure and the identification of metabolic diseases such as osteoporosis. We validated the system with custom-made 3D printing phantoms and 4T1 tumor-bearing mice, demonstrating the ability to accurately recover the tumor blood vessels and bones with an imaging depth of 5 mm, an image resolution of 0.15 mm, and a depth resolution of 0.35 mm. This pioneering development provides an effective tool for non-invasive real-time NIR-II fluorescence imaging and is instrumental in understanding the effects of cancer treatments on tumor vessels and bone structure in small animals.

We have successfully printed an anatomically shaped hand model with properties similar to in vivo investigations with a fluorescence hand imager using indocyanine green (ICG). By adding the fluorescent dye Lumogen IR 765, titanium dioxide powder and black ink to a 3D printing methacrylate photopolymer we realized phantoms with physiologically relevant absorption and reduced scattering coefficients as well as fluorescence properties similar to the widely applied contrast agent ICG. The phantoms show an excellent long-term photostability making them well suited for device performance monitoring and comparison. In contrast, fluorescence of phantoms printed with ICG continuously decreases over time.

We introduce a novel cerebral perfusion phantom to calibrate laser speckle imaging systems for small animal brain imaging. A method based on Hele-Shaw cell technique is used to create fractal like structures which serves as the mould for the phantom making. The structure is casted to create micro channels in PDMS phantoms which is found to closely mimic the structure of superficial cerebral blood vessels in small animals. The proposed method has the potential to fabricate optically and anatomically accurate cerebral perfusion phantom using a quick and inexpensive fabrication technique suitable for blood flow imaging studies.

Recent advancements in imaging technologies (MRI, PET, CT, among others) have significantly improved clinical localisation of lesions of the central nervous system (CNS) before surgery, making possible for neurosurgeons to plan and navigate away from functional brain locations when removing tumours, such as gliomas. However, neuronavigation in the surgical management of brain tumours remains a significant challenge, due to the inability to maintain accurate spatial information of pathological and healthy locations intraoperatively. To answer this challenge, the HyperProbe consortium have been put together, consisting of a team of engineers, physicists, data scientists and neurosurgeons, to develop an innovative, all-optical, intraoperative imaging system based on (i) hyperspectral imaging (HSI) for rapid, multiwavelength spectral acquisition, and (ii) artificial intelligence (AI) for image reconstruction, morpho-chemical characterisation and molecular fingerprint recognition. Our HyperProbe system will (1) map, monitor and quantify biomolecules of interest in cerebral physiology; (2) be handheld, cost-effective and user-friendly; (3) apply AI-based methods for the reconstruction of the hyperspectral images, the analysis of the spatio-spectral data and the development and quantification of novel biomarkers for identification of glioma and differentiation from functional brain tissue. HyperProbe will be validated and optimised with studies in optical phantoms, in vivo against gold standard modalities in neuronavigational imaging, and finally we will provide proof of principle of its performances during routine brain tumour surgery on patients. HyperProbe aims at providing functional and structural information on biomarkers of interest that is currently missing during neuro-oncological interventions.

We present a method for the computer aided focus tracking of circling and rotating samples. This makes the manual sample alignment, required for Scanning Laser Optical Tomography samples, obsolete. Furthermore, we present an approach to correct the residual jitter of the samples with a post-processing algorithm.

As an application of near-infrared spectroscopy, pulse oximetry is widely used to non-invasively measure the arterial oxygen saturation (SpO₂). To improve the maternal and fetal outcome during delivery, transabdominal fetal pulse oximetry can be used to measure the fetal SpO₂. The layered tissue structure above the fetus, however, is complex and thick. In order to understand the feasibility of transabdominal pulse oximetry, we simulated light propagation through the maternal abdomen. For realistic geometry, we segmented a magnetic resonance imaging (MRI) scan of a pregnant women to create a 3D anatomical model, from which a 3D tetrahedron mesh was generated. Using this mesh, we then simulated photon propagation for 5 wavelengths and a grid of 70 sources and detectors (35 each) on the maternal abdomen above the fetal head with NIRFAST. Finally, depth sensitives were examined with Jacobian (J) and flat field. For a fetal head at ~4 cm depth, we found the normalized J at this depth is ~0.1-1% for source-detector distances within 9-10 cm. We also observed that at the same depth, the normalized flat field sensitivity is ~5-10%, which is 1-2 orders of magnitude higher than the normalized J. These results indicate that enough light can reach the fetus when considering ~9 cm source detector distances.

Diffuse Optical Tomography (DOT) is a non-invasive medical imaging technique that utilizes near-infrared light to study the optical properties of tissues. Recently, deep learning has gained popularity as a reconstruction method to solve DOT. However, despite its success, previous studies only reconstructed semi-homogeneous breasts with an absorption coefficient resolution of 2e-3 1/mm. In this paper, we propose a novel preprocessing method that considers the spatial correlations between different measurements to improve the reconstruction accuracy. Our algorithm is applied on a non-homogeneous breast phantom with absorption coefficient resolution of 5e-7 1/mm to reconstruct its optical properties. We compare our algorithm performance with and without the preprocessing step and to a SOTA analytical inversion technique. The proposed method is able to reduce the RMSE by more than 70% (0.44 to 0.11) and increase the contrast ratio by almost an order of magnitude (0.09 to 0.79).

The spherical harmonics methods (classical and modified) can be used to solve the radiative transport equation in series form. However, the evaluation of the truncated series for the angular radiance near sources or boundaries is known to be a difficult or partly an impossible task. In this conference paper, we report on a hybrid P_N-method which enables the evaluation of the angle-resolved radiance at the boundary of an anisotropically scattering medium. The number of spherical harmonics required under the use of the hybrid model is reduced significantly compared with the classical or modified spherical harmonics method.

Analytical solutions to the radiative transport equation are used in spatial frequency domain imaging to analyze the light propagation through turbid media. In this work analytical solutions for the angular resolved single scattered radiance in a two-layered turbid medium are presented and compared to Monte Carlo simulations.

Phantoms that accurately simulate in-vivo tissue properties are essential for the advancement of medical imaging methods. In this paper, we propose a novel approach to develop a fast and tunable dynamical phantom that mimics in-vivo blood flow, based on stochastic differential equations (SDE) and piezoelectric actuators. We validate the phantom using in-vivo human blood flow studies.

In this work we introduce an analytical way of computing the photon measurement density functions in multilayered flat and spherical media. Comparisons with Monte Carlo simulations in the particular case of two-layered media show very good agreement (differences below 10%), with the additional advantage that the time taken by the theoretical calculations is several orders of magnitude (more than 6) lower than the corresponding Monte Carlo calculations.

Identifying vulnerable plaques at an early stage is critical to reducing patient mortality associated with cardiovascular diseases. Diffuse reflectance spectroscopy (DRS) can directly measure of absorption and scattering properties of tissue and can detect oxy-haemoglobin inside the plaque associated with intraplaque haemorrhage, which is a feature of vulnerable plaques. This study assesses the potential of using a diffuse reflectance spectroscopy combined with a machine learning algorithm to identify oxygenated haemoglobin in a chick embryo chorioallantoic membrane (CAM) assay model. A total of 88 diffuse reflectance spectra measurements were collected from five 12 to 14-day-old embryos. The first and second derivative of the reflectance spectra were calculated, followed by the use of partial least square-linear discriminant analysis (PLS-LDA) to identify oxygenated haemoglobin in chick embryo vessels. The model achieved a sensitivity and specificity of 96% and 72%, respectively, in differentiating arteries from veins (oxy-haemoglobin) using reflectance data. The sensitivity and specificity were 92% and 88% using the first derivative of reflectance data, and 100% and 92% using the second derivative of reflectance data in the wavelength range of 500-600 nm. Initial results indicate that derivative reflectance combined with multivariate analysis has advantages for detecting tissue oxygenated haemoglobin in CAM assay model. This approach shows promise as a way to identify and study the features of vulnerable plaques.

Photoplethysmography (PPG) is a promising optical measurement method for daily patient monitoring for a wide range of physiological parameters. Nevertheless, this measurement technique remains sensitive to external disturbances, inter/intra-subject physiology, and sensor configuration. In order to study the behavior of these optical signals, we report an original bench based on PPG measurement and Diffuse Reflectance Spectroscopy (DRS). Thanks to this setup, we are able to investigate several sensor configurations in terms of source-detector distances and wavelengths at kHz acquisition rate. These measurements are multi-wavelength in the visible and near infrared range, synchronized and performed on up to eight configurable source-detector distances (three reported in this paper). An optical calibration suited for comparative measurements is also implemented. We additionally present the first results obtained on a representative and dynamic medium. This study highlights the dependence between the optical signal measurement and the configuration of the sensor. These first results demonstrate the interest of the development of our optical bench to study in depth the complex processes occurring in heterogeneous scattering environments such as biological tissues. This work paves the way for robust, quantitative, accurate and continuous monitoring of vital signs based on PPG, and opens perspectives for physiological measurements on persons and sensor development.

We present a method to assess steatosis percentage and relative quantification of fat in the liver using near infrared diffuse reflectance spectroscopy. Measurements performed are in good agreement with the gold standard anatomopathological results.

In neurosurgical operations, such as cerebral aneurysm clipping, bypass surgery and arteriovenous malformations, monitoring of cerebral blood flow is critically important. Currently, surgeons do not have real-time noninvasive methods for intraoperative visualization of cerebral blood flow. Laser speckle contrast imaging (LSCI), widely used in the diagnosis of blood flow, may be promising for solving this problem. In the present study, the LSCI was demonstrated for the evaluation of acute cerebral blood flow abnormalities in laboratory animals. To visualize cerebral blood flow, the rats underwent cranial trepanation. The disruption of cerebral blood circulation was simulated by clamping both common carotid arteries through the neck approach. A specially designed LSCI system was used to assess blood flow. A distinct reduction in blood flow in the cerebral cortex after carotid artery clipping was shown, which lasted for 10 min continuously. At the same time, a more intensive blood flow decrease was observed in the right hemisphere, which can be related to the blood supply of the left hemisphere through the circle of Willis. After the clips were removed, blood flow in both hemispheres was restored to a level higher than the initial level. The results obtained show the prospects of using LSCI for the control of acute disorders of cerebral blood flow during neurosurgical operations in real-time.

The light propagation in three-dimensional scattering media was studied based on solutions of Maxwell’s equations and the radiative transfer equation. The light distribution of laterally infinitely extended slabs was calculated versus depth for an incident plane wave. The energy density obtained by a numerical solution of Maxwell’s equations was compared to the fluence rate calculated by an analytical solution of the radiative transfer equation for different concentrations of spherical particles in the slab. In general, a good agreement between the energy density and the fluence rate was found. A small systematic difference was obtained at large concentrations of the scattering particles, which can be attributed to dependent scattering effects.

In this work we derive general equations for the analytical calculation of photon mean partial pathlengths (MPPLs) in turbid media with an arbitrary number of layers. Comparisons with their Monte Carlo (MC) counterpart show excellent agreement. These quantities can now be used to retrieve haemoglobin concentrations changes in cerebral blood in real-time and with minimal computing requirements.

The laser speckle imaging technique with sub-pixel correlation analysis allows to identify changes in the sterile zone radius, and makes it possible to predict these changes significantly earlier than the disk diffusion method which is recommended by the European Committee on Antimicrobial susceptibility testing (EUCAST). Results are oriented towards speeding and facilitating epidemiological analysis.

The photoacoustic guided (PAG) technique can noninvasively increase light penetration and localization in deep tissue. The mechanism of the PAG technique is considered to be a pressure dependent refractive index change caused by acoustic wave that produces a virtual waveguide, but it is still a highly controversial issue. In this study, light propagation in the models with the PAG region was calculated by a Monte Carlo simulation. We assumed that the change in the refractive index or anisotropy factor of scattering in the PAG region contributes to the light guiding effect. Both the refractive index and anisotropy factor heterogeneities increase the photon density in the PAG region of the deep tissue part. The effect of localization of the photon density in the PAG region with the anisotropy factor heterogeneity is less significant compared to that with the refractive index heterogeneity.

Time-domain diffuse optical tomography (TD-DOT) systems use pulsed lasers and measure time-varying temporal point spread function to estimate spatially varying optical parameters in biological tissues. Generally, the TD-DOT systems have been based on picosecond light sources and measurements were performed using photon counting methods or time-gated detectors. In this work, we demonstrate the feasibility of TD-DOT using a nanosecond laser source and measurements using a standard digital oscilloscope. For this, we constructed a prototype TD-DOT experimental system utilising a high-energy nanosecond Nd:YAG laser combined with a high-bandwidth oscilloscope. The system was used to image an optical phantom. The experiment verified that both absorbing and scattering objects can be simultaneously reconstructed with nanosecond TD-DOT.

We aim to develop a LED-based precise method to determine the optical properties of scattering media between 850 and 1000 nm. For this purpose, we designed a portable device based on spatially resolved diffuse reflection spectroscopy. The approach uses a spectrometer and fiber-coupled NIR-LEDs. We apply an inverse Monte Carlo model to determine the optical properties. A set of 3x4 phantoms serves for validation. The optical properties determined with our sensor and with an standard integrating sphere device agree well. The accuracy of μ′_s and μ_a is around three percent. Furthermore we theoretically analyse the influence of temperature change of the LEDs on the optical properties.

RGB imaging is a non-invasive technique that is able to monitor hemodynamic brain responses following neuronal activation during neurosurgery. These cameras are often present in operating rooms, but a robust quantification is complicated to perform during neurosurgery. Liquid blood have been proposed, but it is not possible to model hemodynamic responses similar to those that occur in the brain. To overcome this issue, we propose a 3D brain model, including activated, non-activated grey matter and temporal hemodynamic fluctuations using Monte Carlo simulations. Several setups were modeled to evaluate their impact for identifying activated brain areas using statistical parametric mapping.

Hyperspectral imaging (HSI) is an optical technique that processes the electromagnetic spectrum at a multitude of monochromatic, adjacent frequency bands. The wide-bandwidth spectral signature of a target object’s reflectance allows fingerprinting its physical, biochemical, and physiological properties. HSI has been applied for various applications, such as remote sensing and biological tissue analysis. Recently, HSI was also used to differentiate between healthy and pathological tissue under operative conditions in a surgery room on patients diagnosed with brain tumors. In this article, we perform a statistical analysis of the brain tumor patients’ HSI scans from the HELICoiD dataset with the aim of identifying the correlation between reflectance spectra and absorption spectra of tissue chromophores. By using the principal component analysis (PCA), we determine the most relevant spectral features for intra- and inter-tissue class differentiation. Furthermore, we demonstrate that such spectral features are correlated with the spectra of cytochrome, i.e., the chromophore highly involved in (hyper) metabolic processes. Identifying such fingerprints of chromophores in reflectance spectra is a key step for automated molecular profiling and, eventually, expert-free biomarker discovery.

Laser speckle contrast imaging (LSCI) methods are extensively used in assessment of blood flow to detect various pathological conditions in different parts of human body. In contrast to LSCI being deployed for larger region of interest (few centimeters), we present a laser speckle imaging at microscopic level. Rather than utilizing the conventional speckle contrast, we use intensity auto-correlation using recently developed Multi-step Volterra Integral method(MVIM) to quantify the micro-perfusion. The proposed laser speckle correlation microscopy (LSCM) system is validated using microfluidic flow phantom experiments.

For the prediction of the scalp-cortex correlation (SCC), it is important to consider the influence of the extracerebral tissues on light propagation in the brain. The three-layered infant head models in which the scalp, skull, and cerebrospinal fluid (CSF) are combined as a homogeneous tissue are constructed in order to investigate the influence of the heterogeneity of the thin extracerebral tissues on the prediction of the SCC of infants. The spatial sensitivity profile in the brain calculated by the three-layered model is less broadened than that by the ordinary five-layered model due to the lack of the low scattering CSF layer. However, the influence of the heterogeneity of the extracerebral tissues on the SCC of infants is less significant than that of an adult. The three-layered head model is practically useful to predict the acceptable SCC of the functional near infrared spectroscopy measurements of infants.

In recent years, hyperspectral imaging (HSI) has demonstrated the capacity to non-invasively differentiate tumours from healthy tissues and identify cancerous regions during surgery, particularly for glioma resection. This is thanks to the use of a relatively large number of adjacent wavelength bands, in order to reconstruct full reflectance spectra of each pixel in the acquired images of the target, thus providing information about its morpho-chemical composition. However, current HSI analysis approaches seem not to fully exploit such advantage, since they mostly tend to focus on tissue features recognition and cancer identification based on supervised algorithm trained upon diagnostic evaluations made by the neurosurgeons or from other diagnostic tools (e.g., histopathology). There is indeed a lack of proper broad-range, optical characterisation of tumour tissue, specifically gliomas, which could provide a more objective, comprehensive and quantitative insight in the spectro-chemistry of the tumour itself and help identifying novel biomarkers for cancer imaging via HSI. For this purpose, we present a fully optical characterisation of fresh ex vivo samples of glioma from surgical biopsies using both a laboratory spectrophotometer and an in-house, high-spectral density HSI system. The latter is based on spectral scanning of the samples via supercontinuum laser (SCL) illumination filtered with acousto-optic tunable filters (AOTF). The results of the spectral characterisation are analysed and compared to extract optical signatures for potential glioma biomarkers in order to further aid neuronavigation via HSI during glioma resection, in particular in the framework of our recently started HyperProbe project.

We report on vein imaging simulation studies using Zemax OpticStudio simulation software. These simulations are done using a 9-layer skin model. All the parameters save the thickness are wavelength dependent, and the simulation was considered at three different wavelengths – 650nm, 780nm and 850nm. For the vein in a given layer and at a given incident wavelength, the source-lens-detector setup was placed at various angles. Simulation results show the angle-dependent penetration of near-infrared (NIR) light into the skin and provide design inputs on the optimum wavelength and angle for vein illumination.

The feasibility of new-generation miniature spectrometers for use in portable broadband NIRS (bNIRS) devices was explored in this investigation. The study outlines tests of varying integration time between 1000 ms, 500 ms and 250 ms and source-detector separations between 2 cm, 3cm and 4 cm, and their effect on the received signal through MEDPHOT tissue-mimicking phantoms, A2, B2, B3 and D7, using the Oceans Optics HL-2000-HP tungsten halogen lamp as a broadband light source. The spectra and SNR were then compared to two gold-standard bNIRS systems. It is found that two of the spectrometers give respectable SNR values for the detection range of 600 - 1000 nm in all regimes except when saturated or using phantom D7. It is demonstrated that these two devices can appropriately be used for source-detector separations of 3 cm and 4 cm at 500 ms and 1000 ms integration times, to determine absorption changes in tissue and thus chromophore concentrations. To use them for 2 cm separations, an additional attenuation component will be required.

In recent years, hyperspectral imaging (HSI) has demonstrated its capacity to non-invasively differentiate tumors from healthy tissues and identify cancerous regions during neurosurgery. Indeed, the spectral information contained in the HS images allows to identify more chromophores, refining the information provided by the imaging system, and allowing to identify the unique signature of each tissue types more accurately. Our HyperProbe project aims at developing a novel HSI system optimized for neurosurgery. As part of this project, we are developing a digital instrument simulator (DIS), based on Monte-Carlo (MC) simulations of the light propagation in tissues, in order to optimize both the hardware and data processing pipeline of our novel instrument. This framework allows us (1) to test the effect on the accuracy of the measurement of several hardware parameters, like the numerical aperture or sensitivity of the detector; (2) to be used as numerical phantoms to test various data processing algorithms; and (3) to generate generic data to develop and train machine learning (ML) algorithms. To do so, our framework is based on a 2-step method. Firstly, MC simulations are run to produce an ideal dataset of the photon transport in tissue. Then, the raw output parameters of the simulations, such as the exit positions and directions of the photons, are processed to take into account the physical parameters of an instrument in order to produce realistic images and test various scenarios. We present here the initial development of this DIS.

We investigate the potential of CW fNIRS with two NIR wavelengths to estimate tissue oxygen saturation by a model based on spatially resolved spectroscopy. Phantoms with optical properties similar to brain tissue are applied to study the influence of the optical parameters and the spatial probe configuration on effective attenuation, absorption, and oxygen saturation. Attenuation coefficients and saturation values can be estimated with a typical discrepancy of 5%, whereas absorption coefficients show larger deviations from the true values. The results are very sensitive against small errors in the source-detector separation.

Muscle aging is characterized by the loss of muscle mass and strength that starts mostly after 50yr and, according to the World Health Organization (2000), it is one of the major causes of independence loss and a risk factor for the development of morbidities at older age. Early diagnosis and treatment are paramount for the progression of the disease. The “Trajector-AGE” project focuses on the study of neuromuscular decline in middle-aged and old populations. Within this project, different techniques are exploited to investigate muscle health (e.g., biopsy, electromyography, Near Infrared Spectroscopy), and among them, Time Domain Near Infrared Spectroscopy (TD-NIRS) and diffuse correlation spectroscopy (DCS) are non-invasive optical techniques which enable to assess muscle oxidative metabolism and perfusion respectively. During the project life, we plan to recruit 100 individuals to evaluate differences among the hemodynamics and microcirculation responses of the vastus lateralis to arterial occlusion and incremental cycling in different age-groups (55–60yrs, middle-aged population; 75-80yrs, old population). The main parameters extrapolated will be the time courses for oxy- (HbO₂), deoxy- (HHb), total- hemoglobin (tHb), tissue oxygen saturation (S_tO₂) and blood flow index (BFI). From these, biomarkers for the neuromuscular decline will be defined. At the time of this work, 21 subjects were already acquired. Here, we present the preliminary results from 1 healthy volunteer.

Protoporphyrin IX (PpIX) is a fluorophore being currently used to localize tumoral tissues. The tissue is usually excited at one wavelength, e.g., 405 nm, and the fluorescence signal is used to estimate the amount of PpIX during surgery. However, other fluorophores (baseline) whose emission spectra are close to the one of PpIX impair the quantification of PpIX and consequently the tissue pathological status classification. An efficient multi-excitation wavelengths method, free from any a priori on the baseline shape, has been proposed to cope with this issue. This method requires decorrelated measurements in the range of PpIX emission at multiple excitation wavelengths. We investigated the influence of the source bandwith on this decorelation by comparing two experimental setups using either LED or laser diode sources. The experimental setup using laser diodes for excitation increases the decorrelation by 35.3 % compared to the one using LEDs in the spectral range of PpIX emission.

Recent findings in animal models and human tissue suggest that deficient menstrual endometrial hypoxia results in heavy menstrual bleeding (HMB)1. There is no current method available to directly measure endometrial oxygen concentrations during hysteroscopy investigation of the underlying cause of HMB. We propose that an optical fibre-based probe delivered through the working channel of a hysteroscope can accurately assess uterine oxygenation. This probe will be designed to transmit and collect broadband white / NIR (~550 nm – 1050 nm) light using multiple fibre optics co-packaged and modified for side emission / collection of light passing through different depths of tissue. Observed spectral variations reveal haemoglobin oxygen saturation, while the controlled depth measurement allows comparison of the endometrium layer to the deeper myometrium muscular outer layer to observe the expected (or defective) relative hypoxia in the menstrual endometrium. An initial co-packaged fibre probe has been prototyped, optimising diffuse emission and collection of light into surrounding tissue. An initial investigation shows real time spectral changes, observing the variation in haemoglobin oxygenation due to pulse and breathing. Fibre probes were placed onto tissue and the separation of the fibre probes was varied, allowing the spectral properties of different tissue types at varying depths to be observed. Introducing more fibres into the probe offers better resolving of obscured tissue layers, a customised multi-fibre spectroscopy measurement instrument is in development to enable this.

Usually, in biomedical optics, the average photon fluence rate, evaluated in a subvolume of a propagating medium, is obtained by Monte Carlo simulations by calculating the power deposited by photons absorbed in the subvolume. We propose an alternative method based on evaluating the average path length traveled by all photons injected within the subvolume. Application examples are given. This method also works for a zero absorption coefficient and for a nonconstant spatial distribution of the absorption coefficient within the subvolume. The proposed approach is a re-visitation of a well-known method applied to nuclear and radiation physics. The results obtained show that a potential advantage of the proposed method is that it can improve the convergence of Monte Carlo simulations. Indeed, when calculating the fluence in a region of interest with the proposed method, all photons passing through the region are considered. Whereas with the traditional approach, only absorbed" photons are considered. In the latter case, this can produce a poorer Monte Carlo statistic for the same number of photons launched.

The Monte Carlo (MC) method is a gold standard for "solving" the radiative transport equation even in complex geometries and distributions of optical properties. The exact analytical benchmark provided by the invariant total mean path length law spent by light injected with uniform Lambertian illumination within nonabsorbing scattering media is used to verify Monte Carlo codes developed for biomedical optics applications. The correctness of an MC code can be evaluated with a sample t-test. In addition, the invariance of the mean path length ensures that the expected value is known regardless of the complexity of the medium. The accuracy of the estimated mean path length can progressively increase as the number of simulated trajectories increases. The method can be used regardless of the scattering and geometric properties of the medium, as well as in the presence of refractive index mismatch between the medium and the outer region and between different regions of the medium. The proposed method is particularly reliable for detecting inaccuracies in the treatment of finite media boundaries. The results presented in this contribution, obtained with a standard computer, show a verification of our MC code to the sixth decimal place. This method can provide a fundamental tool for verification of Monte Carlo codes in the geometry of interest without resorting to simpler geometries and uniform distribution of the scattering properties.

Deep-learning Diffuse Optical Tomography (DL-DOT) is a non-invasive diagnostic method that uses near-infrared radiation and deep-learning algorithms to image soft tissues in the body, such as the breast. However, DL-DOT studies have limitations, such as using only homogeneous or semihomogeneous datasets for the forward problem, which can lead to predictions not being accurate when used on experimental measurements. Another limitation regarding DL-DOT is the severe overfitting of the prediction model observed when DL methods are employed for DOT image reconstruction. To overcome this challenge, a regularized nested UNet++ deep-learning algorithm is employed. The proposed method effectively solves the DOT inverse problem in inhomogeneous breasts by applying a regularization technique. This technique reduces overfitting and simplifies the prediction model. Results show that when the regularized neural network is used to detect tumors, a minimal mean square error (MSE) loss of 5.16 × 10⁻³ is achieved compared to a non-regularized MSE loss of 4.18 × 10⁻². The enhancement of close to one order of magnitude shown by the proposed method demonstrates the significance of regularization neural networks in breast tumor detection and improving the accuracy of DOT image reconstruction.

In this work, we aim to develop a virtual platform to compare the performance of the different manifestations of photon Time of Flight Spectroscopy namely Direct, Indirect and Interferometric photon Time of Flight Spectroscopy (pToFS). Extending the comparison over a range of scenarios, defined by a matrix of optical properties (dubbed here as Virtual Tissue), allows for the definition of different use cases for each of these techniques. The effect of parameters like temporal drift, exposure time and background noise will also be studied.

We present a novel approach that combines Monte Carlo simulations to propagate photons in a turbid media having the dynamics modelled using stochastic differential equations, resulting in simulating diffuse laser speckles for in-vivo blood flow imaging applications. The proposed method allows to model the tissue dynamics with a pre-defined probability density function and spatially varying autocorrelation.

Laser speckle based superficial and deep tissue blood flow imaging is gaining interest with the advent of high speed cameras. Multi-exposure speckle intensity images are often utilized for this purpose, owing to the better quantification of flow. However, any uncertainty in selecting the required exposure range apriori and the data acquisition time associated with multi-exposure intensity measurements limit the temporal resolution of these systems. To address these concerns, we propose a deep learning-based imputation using Generative Adversarial Imputation Network (GAIN) to generate additional temporal samples from coarsely acquired multi-exposure speckle data. The feasibility of the proposed method has been verified by using simulations where the trade-off between temporal resolution and the accuracy of flow measurement is minimized.