Review of machine learning for optical imaging of burn wound severity assessment.

Significance: Over the past decade, machine learning (ML) algorithms have rapidly become much more widespread for numerous biomedical applications, including the diagnosis and categorization of disease and injury. Aim: Here, we seek to characterize the recent growth of ML techniques that use imaging data to classify burn wound severity and report on the accuracies of different approaches. Approach: To this end, we present a comprehensive literature review of preclinical and clinical studies using ML techniques to classify the severity of burn wounds. Results: The majority of these reports used digital color photographs as input data to the classification algorithms, but recently there has been an increasing prevalence of the use of ML approaches using input data from more advanced optical imaging modalities (e.g., multispectral and hyperspectral imaging, optical coherence tomography), in addition to multimodal techniques. The classification accuracy of the different methods is reported; it typically ranges from ∼70% to 90% relative to the current gold standard of clinical judgment. Conclusions: The field would benefit from systematic analysis of the effects of different input data modalities, training/testing sets, and ML classifiers on the reported accuracy. Despite this current limitation, ML-based algorithms show significant promise for assisting in objectively classifying burn wound severity.

Motion-resistant three-wavelength spatial frequency domain imaging system with ambient light suppression using an 8-tap CMOS image sensor.

Significance: We present a motion-resistant three-wavelength spatial frequency domain imaging (SFDI) system with ambient light suppression using an 8-tap complementary metal-oxide semiconductor (CMOS) image sensor (CIS) developed at Shizuoka University. The system addresses limitations in conventional SFDI systems, enabling reliable measurements in challenging imaging scenarios that are closer to real-world conditions. Aim: Our study demonstrates a three-wavelength SFDI system based on an 8-tap CIS. We demonstrate and evaluate the system's capability of mitigating motion artifacts and ambient light bias through tissue phantom reflectance experiments and in vivo volar forearm experiments. Approach: We incorporated the Hilbert transform to reduce the required number of projected patterns per wavelength from three to two per spatial frequency. The 8-tap image sensor has eight charge storage diodes per pixel; therefore, simultaneous image acquisition of eight images based on multi-exposure is possible. Taking advantage of this feature, the sensor simultaneously acquires images for planar illumination, sinusoidal pattern projection at three wavelengths, and ambient light. The ambient light bias is eliminated by subtracting the ambient light image from the others. Motion artifacts are suppressed by reducing the exposure and projection time for each pattern while maintaining sufficient signal levels by repeating the exposure. The system is compared to a conventional SFDI system in tissue phantom experiments and then in vivo measurements of human volar forearms. Results: The 8-tap image sensor-based SFDI system achieved an acquisition rate of 9.4 frame sets per second, with three repeated exposures during each accumulation period. The diffuse reflectance maps of three different tissue phantoms using the conventional SFDI system and the 8-tap image sensor-based SFDI system showed good agreement except for high scattering phantoms. For the in vivo volar forearm measurements, our system successfully measured total hemoglobin concentration, tissue oxygen saturation, and reduced scattering coefficient maps of the subject during motion (16.5 cm/s) and under ambient light (28.9 lx), exhibiting fewer motion artifacts compared with the conventional SFDI. Conclusions: We demonstrated the potential for motion-resistant three-wavelength SFDI system with ambient light suppression using an 8-tap CIS.

Quantitative hemodynamic imaging: a method to correct the effects of optical properties on laser speckle imaging.

Significance: Studying cerebral hemodynamics may provide diagnostic information on neurological conditions. Wide-field imaging techniques, such as laser speckle imaging (LSI) and optical intrinsic signal imaging, are commonly used to study cerebral hemodynamics. However, they often do not account appropriately for the optical properties of the brain that can vary among subjects and even during a single measurement. Here, we describe the combination of LSI and spatial-frequency domain imaging (SFDI) into a wide-field quantitative hemodynamic imaging (QHI) system that can correct the effects of optical properties on LSI measurements to achieve a quantitative measurement of cerebral blood flow (CBF). Aim: We describe the design, fabrication, and testing of QHI. Approach: The QHI hardware combines LSI and SFDI with spatial and temporal synchronization. We characterized system sensitivity, accuracy, and precision with tissue-mimicking phantoms. With SFDI optical property measurements, we describe a method derived from dynamic light scattering to obtain absolute CBF values from LSI and SFDI measurements. We illustrate the potential benefits of absolute CBF measurements in resting-state and dynamic experiments. Results: QHI achieved a 50-Hz raw acquisition frame rate with a 10×10 mm field of view and flow sensitivity up to ∼4 mm/s. The extracted SFDI optical properties agreed well with a commercial system (R2≥0.98). The system showed high stability with low coefficients of variations over multiple sessions within the same day (<1%) and over multiple days (<4%). When optical properties were considered, the in-vivo hypercapnia gas challenge showed a slight difference in CBF (-1.5% to 0.5% difference). The in-vivo resting-state experiment showed a change in CBF ranking for nine out of 13 animals when the correction method was applied to LSI CBF measurements. Conclusions: We developed a wide-field QHI system to account for the confounding effects of optical properties on CBF LSI measurements using the information obtained from SFDI.

Current and Emerging Imaging Techniques for Neurofibromatosis Type 1-Associated Cutaneous Neurofibromas.

A consistent set of measurement techniques must be applied to reliably and reproducibly evaluate the efficacy of treatments for cutaneous neurofibromas (cNFs) in people with neurofibromatosis type 1 (NF1). cNFs are neurocutaneous tumors that are the most common tumor in people with NF1 and represent an area of unmet clinical need. This review presents the available data regarding approaches in use or development to identify, measure, and track cNFs, including calipers, digital imaging, and high-frequency ultrasound sonography. We also describe emerging technologies such as spatial frequency domain imaging and the application of imaging modalities such as optical coherence tomography that may enable the detection of early cNFs and prevention of tumor-associated morbidity.

Quantifying the confounding effect of pigmentation on measured skin tissue optical properties: a comparison of colorimetry with spatial frequency domain imaging.

SIGNIFICANCE: Spatial frequency domain imaging (SFDI) is a wide-field diffuse optical imaging technique for separately quantifying tissue reduced scattering (µs ' ) and absorption (µa) coefficients at multiple wavelengths, providing wide potential utility for clinical applications such as burn wound characterization and cancer detection. However, measured µs ' and µa can be confounded by absorption from melanin in patients with highly pigmented skin. This issue arises because epidermal melanin is highly absorbing for visible wavelengths and standard homogeneous light-tissue interaction models do not properly account for this complexity. Tristimulus colorimetry (which quantifies pigmentation using the L * "lightness" parameter) can provide a point of comparison between µa, µs ' , and skin pigmentation. AIM: We systematically compare SFDI and colorimetry parameters to quantify confounding effects of pigmentation on measured skin µs ' and µa. We assess the correlation between SFDI and colorimetry parameters as a function of wavelength. APPROACH: µs ' and µa from the palm and ventral forearm were measured for 15 healthy subjects with a wide range of skin pigmentation levels (Fitzpatrick types I to VI) using a Reflect RS® (Modulim, Inc., Irvine, California) SFDI instrument (eight wavelengths, 471 to 851 nm). L * was measured using a Chroma Meter CR-400 (Konica Minolta Sensing, Inc., Tokyo). Linear correlation coefficients were calculated between L * and µs ' and between L * and µa at all wavelengths. RESULTS: For the ventral forearm, strong linear correlations between measured L * and µs ' values were observed at shorter wavelengths (R > 0.92 at ≤659 nm), where absorption from melanin confounded the measured µs ' . These correlations were weaker for the palm (R < 0.59 at ≤659 nm), which has less melanin than the forearm. Similar relationships were observed between L * and µa. CONCLUSIONS: We quantified the effects of epidermal melanin on skin µs ' and µa measured with SFDI. This information may help characterize and correct pigmentation-related inaccuracies in SFDI skin measurements.

Assessing multimodal optical imaging of perfusion in burn wounds.

A critical need exists for early, accurate diagnosis of burn wound severity to help identify the course of treatment and outcome of the wound. Laser speckle imaging (LSI) is a promising blood perfusion imaging approach, but it does not account for changes in tissue optical properties that can occur with burn wounds, which are highly dynamic environments. Here, we studied optical property dynamics following burn injury and debridement and the associated impact on interpretation of LSI measurements of skin perfusion. We used spatial frequency domain imaging (SFDI) measurements of tissue optical properties to study the impact of burn-induced changes in these properties on LSI measurements. An established preclinical porcine model of burn injury was used (n = 8). SFDI and LSI data were collected from burn wounds of varying severity. SFDI measurements demonstrate that optical properties change in response to burn injury in a porcine model. We then apply theoretical modeling to demonstrate that the measured range of optical property changes can affect the interpretation of LSI measurements of blood flow, but this effect is minimal for most of the measured data. Collectively, our results indicate that, even with a dynamic burn wound environment, blood-flow measurements with LSI can serve as an appropriate strategy for accurate assessment of burn severity.

Characterizing reduced scattering coefficient of normal human skin across different anatomic locations and Fitzpatrick skin types using spatial frequency domain imaging.

SIGNIFICANCE: Spatial frequency domain imaging (SFDI), a noncontact wide-field imaging technique using patterned illumination with multiple wavelengths, has been used to quantitatively measure structural and functional parameters of in vivo tissue. Using SFDI in a porcine model, we previously found that scattering changes in skin could potentially be used to noninvasively assess burn severity and monitor wound healing. Translating these findings to human subjects necessitates a better understanding of the variation in "baseline" human skin scattering properties across skin types and anatomical locations. AIM: Using SFDI, we aim to characterize the variation in the reduced scattering coefficient (µs') for skin across a range of pigmentation and anatomic sites (including common burn locations) for normal human subjects. These measurements are expected to characterize baseline human skin properties to inform our use of SFDI for clinical burn severity and wound healing assessments. APPROACH: SFDI was used to measure µs' in the visible- and near-infrared regime (471 to 851 nm) in 15 subjects at 10 anatomical locations. Subjects varied in age, gender, and Fitzpatrick skin type. RESULTS: For all anatomical locations, the coefficient of variation in measured µs' decreased with increasing wavelength. High intersubject variation in µs' at visible wavelengths coincided with large values of the melanin extinction coefficient at those wavelengths. At 851 nm, where intersubject variation in µs' was smallest for all anatomical locations and absorption from melanin is minimal, significant intrasubject differences in µs' were observed at the different anatomical locations. CONCLUSIONS: Our study is the first report of wide-field mapping of human skin scattering properties across multiple skin types and anatomical locations using SFDI. Measured µs' values varied notably between skin types at wavelengths where absorption from melanin was prominent. Additionally, µs' varied considerably across different anatomical locations at 851 nm, where the confounding effects from melanin absorption are minimized.

Spatial Frequency Domain Imaging (SFDI) of clinical burns: A case report.

While visual assessment by a clinician is the standard of care for burn severity evaluations, new technologies at various stages of development are attempting to add objectivity to this practice by quantifying burn severity. Assessment accuracy generally improves after the burn injury has progressed, but early assessments that correctly identify superficial partial and deep partial burns have the potential to lead to more prompt treatments and shorter recovery times. To date, Spatial Frequency Domain Imaging (SFDI) has only been used in animal models of burns, but has shown the potential to categorize burns accurately at earlier time points. Here we examine the potential for SFDI to assess burn severity in clinical patients. We also utilize Laser Speckle Imaging (LSI), an FDA cleared non-invasive imaging technology that typically measures blood perfusion in order to evaluate burns in clinical patients. We present a case series of two patients, both with partial thickness burns of varying severity. Partial thickness burns are often difficult for clinicians to categorize based on visual appearance alone. SFDI and LSI were both performed on each patient at approximately 24 and 72 h after their respective burn incidents. Each technique was able to render spatially resolved information that enabled improved assessment accuracy for each burn. This represents the first publication of SFDI applied to clinical burn patients after being successfully utilized in animal models, and highlights the potential for SFDI as a feasible tool for the timely categorization of burn severity.

Spatial frequency domain imaging: a quantitative, noninvasive tool for in vivo monitoring of burn wound and skin graft healing.

There is a need for noninvasive, quantitative methods to characterize wound healing in the context of longitudinal investigations related to regenerative medicine. Such tools have the potential to inform the assessment of wound status and healing progression and aid the development of new treatments. We employed spatial frequency domain imaging (SFDI) to characterize the changes in optical properties of tissue during wound healing progression in a porcine model of split-thickness skin grafts and also in a model of burn wound healing with no graft intervention. Changes in the reduced scattering coefficient measured using SFDI correlated with structural changes reported by histology of biopsies taken concurrently. SFDI was able to measure spatial inhomogeneity in the wounds and predicted heterogeneous healing. In addition, we were able to visualize differences in healing rate, depending on whether a wound was debrided and grafted, versus not debrided and left to heal without intervention apart from topical burn wound care. Changes in the concentration of oxy- and deoxyhemoglobin were also quantified, giving insight into hemodynamic changes during healing.

Burn wound classification model using spatial frequency-domain imaging and machine learning.

Accurate assessment of burn severity is critical for wound care and the course of treatment. Delays in classification translate to delays in burn management, increasing the risk of scarring and infection. To this end, numerous imaging techniques have been used to examine tissue properties to infer burn severity. Spatial frequency-domain imaging (SFDI) has also been used to characterize burns based on the relationships between histologic observations and changes in tissue properties. Recently, machine learning has been used to classify burns by combining optical features from multispectral or hyperspectral imaging. Rather than employ models of light propagation to deduce tissue optical properties, we investigated the feasibility of using SFDI reflectance data at multiple spatial frequencies, with a support vector machine (SVM) classifier, to predict severity in a porcine model of graded burns. Calibrated reflectance images were collected using SFDI at eight wavelengths (471 to 851 nm) and five spatial frequencies (0 to 0.2 mm - 1). Three models were built from subsets of this initial dataset. The first subset included data taken at all wavelengths with the planar (0 mm - 1) spatial frequency, the second comprised data at all wavelengths and spatial frequencies, and the third used all collected data at values relative to unburned tissue. These data subsets were used to train and test cubic SVM models, and compared against burn status 28 days after injury. Model accuracy was established through leave-one-out cross-validation testing. The model based on images obtained at all wavelengths and spatial frequencies predicted burn severity at 24 h with 92.5% accuracy. The model composed of all values relative to unburned skin was 94.4% accurate. By comparison, the model that employed only planar illumination was 88.8% accurate. This investigation suggests that the combination of SFDI with machine learning has potential for accurately predicting burn severity.

Impact of hemoglobin breakdown products in the spectral analysis of burn wounds using spatial frequency domain spectroscopy.

Burn wounds and wound healing invoke several biological processes that may complicate the interpretation of spectral imaging data. Through analysis of spatial frequency domain spectroscopy data (450 to 1000 nm) obtained from longitudinal investigations using a graded porcine burn wound healing model, we have identified features in the absorption spectrum that appear to suggest the presence of hemoglobin breakdown products, e.g., methemoglobin. Our results show that the calculated concentrations of methemoglobin directly correlate with burn severity, 24 h after the injury. In addition, tissue parameters such as oxygenation (StO2) and water fraction may be underestimated by 20% and 78%, respectively, if methemoglobin is not included in the spectral analysis.

Solid tissue simulating phantoms having absorption at 970 nm for diffuse optics.

Tissue simulating phantoms can provide a valuable platform for quantitative evaluation of the performance of diffuse optical devices. While solid phantoms have been developed for applications related to characterizing exogenous fluorescence and intrinsic chromophores such as hemoglobin and melanin, we report the development of a poly(dimethylsiloxane) (PDMS) tissue phantom that mimics the spectral characteristics of tissue water. We have developed these phantoms to mimic different water fractions in tissue, with the purpose of testing new devices within the context of clinical applications such as burn wound triage. Compared to liquid phantoms, cured PDMS phantoms are easier to transport and use and have a longer usable life than gelatin-based phantoms. As silicone is hydrophobic, 9606 dye was used to mimic the optical absorption feature of water in the vicinity of 970 nm. Scattering properties are determined by adding titanium dioxide, which yields a wavelength-dependent scattering coefficient similar to that observed in tissue in the near-infrared. Phantom properties were characterized and validated using the techniques of inverse adding-doubling and spatial frequency domain imaging. Results presented here demonstrate that we can fabricate solid phantoms that can be used to simulate different water fractions

Quantitative long-term measurements of burns in a rat model using Spatial Frequency Domain Imaging (SFDI) and Laser Speckle Imaging (LSI).

BACKGROUND AND OJECTIVES: The current standard for diagnosis of burn severity and subsequent wound healing is through clinical examination, which is highly subjective. Several new technologies are shifting focus to burn care in an attempt to help quantify not only burn depth but also the progress of healing. While accurate early assessment of partial thickness burns is critical for dictating the course of treatment, the ability to quantitatively monitor wound status over time is critical for understanding treatment efficacy. SFDI and LSI are both non-invasive imaging modalities that have been shown to have great diagnostic value for burn severity, but have yet to be tested over the course of wound healing. METHODS: In this study, a hairless rat model (n = 6, 300-450 g) was used with a four pronged comb to create four identical partial thickness burns (superficial n = 3 and deep n = 3) that were used to monitor wound healing over a 28 days period. Weekly biopsies were taken for histological analysis to verify wound progression. Both SFDI and LSI were performed weekly to track the evolution of hemodynamic (blood flow and oxygen saturation) and structural (reduced scattering coefficient) properties for the burns. RESULTS: LSI showed significant changes in blood flow from baseline to 220% in superficial and 165% in deep burns by day 7. In superficial burns, blood flow returned to baseline levels by day 28, but not for deep burns where blood flow remained elevated. Smaller increases in blood flow were also observed in the surrounding tissue over the same time period. Oxygen saturation values measured with SFDI showed a progressive increase from baseline values of 66-74% in superficial burns and 72% in deep burns by day 28. Additionally, SFDI showed significant decreases in the reduced scattering coefficient shortly after the burns were created. The scattering coefficient progressively decreased in the wound area, but returned towards baseline conditions at the end of the 28 days period. Scattering changes in the surrounding tissue remained constant despite the presence of hemodynamic changes. CONCLUSIONS: Here, we show that LSI and SFDI are capable of monitoring changes in hemodynamic and scattering properties in burn wounds over a 28 days period. These results highlight the potential insights that can be gained by using non-invasive imaging technologies to study wound healing. Further development of these technologies could be revolutionary for wound monitoring and studying the efficacy of different treatments. Lasers Surg. Med. 49:293-304, 2017. © 2017 Wiley Periodicals, Inc.

Fluorescence lifetime spectroscopy of tissue autofluorescence in normal and diseased colon measured ex vivo using a fiber-optic probe.

We present an ex vivo study of temporally and spectrally resolved autofluorescence in a total of 47 endoscopic excision biopsy/resection specimens from colon, using pulsed excitation laser sources operating at wavelengths of 375 nm and 435 nm. A paired analysis of normal and neoplastic (adenomatous polyp) tissue specimens obtained from the same patient yielded a significant difference in the mean spectrally averaged autofluorescence lifetime -570 ± 740 ps (p = 0.021, n = 12). We also investigated the fluorescence signature of non-neoplastic polyps (n = 6) and inflammatory bowel disease (n = 4) compared to normal tissue in a small number of specimens.

Multi-site optical excitation using ChR2 and micro-LED array.

Studying neuronal processes such as synaptic summation, dendritic physiology and neural network dynamics requires complex spatiotemporal control over neuronal activities. The recent development of neural photosensitization tools, such as channelrhodopsin-2 (ChR2), offers new opportunities for non-invasive, flexible and cell-specific neuronal stimulation. Previously, complex spatiotemporal control of photosensitized neurons has been limited by the lack of appropriate optical devices which can provide 2D stimulation with sufficient irradiance. Here we present a simple and powerful solution that is based on an array of high-power micro light-emitting diodes (micro-LEDs) that can generate arbitrary optical excitation patterns on a neuronal sample with micrometre and millisecond resolution. We first describe the design and fabrication of the system and characterize its capabilities. We then demonstrate its capacity to elicit precise electrophysiological responses in cultured and slice neurons expressing ChR2.

A fluorescence lifetime imaging scanning confocal endomicroscope.

We describe a fluorescence lifetime imaging endomicroscope employing a fibre bundle probe and time correlated single photon counting. Preliminary images of stained pollen grains, eGFP-labelled cells exhibiting Förster resonant energy transfer and tissue autofluorescence are presented.

Improved sectioning in a slit scanning confocal microscope.

We describe a simple implementation of a slit scanning confocal microscope to obtain an axial resolution better than that of a point-scanning confocal microscope. Under slit illumination, images of a fluorescent object are captured using an array detector instead of a line detector so that out-of-focus light is recorded and then subtracted from the adjacent images. Axial resolution after background subtraction is 2.2 times better than the slit confocal resolution, and out-of-focus image suppression is calculated to attenuate with defocus faster by 1 order of magnitude than in the point confocal case.

A compact, multidimensional spectrofluorometer exploiting supercontinuum generation.

We report a novel, compact and automated multidimensional spectrofluorometer that exploits a fibre-laser-pumped ultrafast supercontinuum source to provide resolution with respect to intensity, excitation and emission wavelength, decay time and polarisation. This instrument has been applied to study the photophysics of the phase-sensitive membrane probe di-4-ANEPPDHQ and to characterise protein-protein interactions via Förster resonance energy transfer. It can be applied to in situ measurements via a fibre-optic probe in medical and other contexts and is demonstrated here to provide a comprehensive characterisation of tissue autofluorescence.