Computer-aided classification of suspicious pigmented lesions using wide-field images

BACKGROUND AND OBJECTIVE

Early identification of melanoma is conducted through whole-body visual examinations to detect suspicious pigmented lesions, a situation that fluctuates in accuracy depending on the experience and time of the examiner. Computer-aided diagnosis tools for skin lesions are typically trained using pre-selected single-lesion images, taken under controlled conditions, which limits their use in wide-field scenes. Here, we propose a computer-aided classifier system with such input conditions to aid in the rapid identification of suspicious pigmented lesions at the primary care level.

METHODS

133 patients with a multitude of skin lesions were recruited for this study. All lesions were examined by a board-certified dermatologist and classified into "suspicious" and "non-suspicious". A new clinical database was acquired and created by taking Wide-Field images of all major body parts with a consumer-grade camera under natural illumination condition and with a consistent source of image variability. 3-8 images were acquired per patient on different sites of the body, and a total of 1759 pigmented lesions were extracted. A machine learning classifier was optimized and build into a computer aided classification system to binary classify each lesion using a suspiciousness score.

RESULTS

In a testing set, our computer-aided classification system achieved a sensitivity of 100% for suspicious pigmented lesions that were later confirmed by dermoscopy examination ("SPL_A") and 83.2% for suspicious pigmented lesions that were not confirmed after examination ("SPL_B"). Sensitivity for non-suspicious lesions was 72.1%, and accuracy was 75.9%. With these results we defined a suspiciousness score that is aligned with common macro-screening (naked eye) practices.

CONCLUSIONS

This work demonstrates that wide-field photography combined with computer-aided classification systems can distinguish suspicious from non-suspicious pigmented lesions, and might be effective to assess the severity of a suspicious pigmented lesions. We believe this approach could be useful to support skin screenings at a population-level.

H-EM: An algorithm for simultaneous cell diameter and intensity quantification in low-resolution imaging cytometry

Fluorescent cytometry refers to the quantification of cell physical properties and surface biomarkers using fluorescently-tagged antibodies. The generally preferred techniques to perform such measurements are flow cytometry, which performs rapid single cell analysis by flowing cells one-by-one through a channel, and microscopy, which eliminates the complexity of the flow channel, offering multi-cell analysis at a lesser throughput. Low-magnification image-based cytometers, also called "cell astronomy" systems, hold promise of simultaneously achieving both instrumental simplicity and high throughput. In this magnification regime, a single cell is mapped to a handful of pixels in the image. While very attractive, this idea has, so far, not been proven to yield quantitative results of cell-labeling, mainly due to the poor signal-to-noise ratio present in those images and to partial volume effects. In this work we present a cell astronomy system that, when coupled with custom-developed algorithms, is able to quantify cell intensities and diameters reliably. We showcase the system using calibrated MESF beads and fluorescently stained leukocytes, achieving good population identification in both cases. The main contribution of the proposed system is in the development of a novel algorithm, H-EM, that enables inter-cluster separation at a very low magnification regime (2x). Such algorithm provides more accurate brightness estimates than DAOSTORM when compared to manual analysis, while fitting cell location, brightness, diameter, and background level concurrently. The algorithm first performs Fisher discriminant analysis to detect bright spots. From each spot an expectation-maximization algorithm is initialized over a heterogeneous mixture model (H-EM), this algorithm recovers both the cell fluorescence and diameter with sub-pixel accuracy while discriminating the background noise. Finally, a recursive splitting procedure is applied to discern individual cells in cell clusters.

Non-invasive detection of severe neutropenia in chemotherapy patients by optical imaging of nailfold microcirculation

White-blood-cell (WBC) assessment is employed for innumerable clinical procedures as one indicator of immune status. Currently, WBC determinations are obtained by clinical laboratory analysis of whole blood samples. Both the extraction of blood and its analysis limit the accessibility and frequency of the measurement. In this study, we demonstrate the feasibility of a non-invasive device to perform point-of-care WBC analysis without the need for blood draws, focusing on a chemotherapy setting where patients’ neutrophils—the most common type of WBC—become very low. In particular, we built a portable optical prototype, and used it to collect 22 microcirculatory-video datasets from 11 chemotherapy patients. Based on these videos, we identified moving optical absorption gaps in the flow of red cells, using them as proxies to WBC movement through nailfold capillaries. We then showed that counting these gaps allows discriminating cases of severe neutropenia (<500 neutrophils per µL), associated with increased risks of life-threatening infections, from non-neutropenic cases (>1,500 neutrophils per µL). This result suggests that the integration of optical imaging, consumer electronics, and data analysis can make non-invasive screening for severe neutropenia accessible to patients. More generally, this work provides a first step towards a long-term objective of non-invasive WBC counting.

Depth-resolved analytical model and correction algorithm for photothermal optical coherence tomography

Photothermal OCT (PT-OCT) is an emerging molecular imaging technique that occupies a spatial imaging regime between microscopy and whole body imaging. PT-OCT would benefit from a theoretical model to optimize imaging parameters and test image processing algorithms. We propose the first analytical PT-OCT model to replicate an experimental A-scan in homogeneous and layered samples. We also propose the PT-CLEAN algorithm to reduce phase-accumulation and shadowing, two artifacts found in PT-OCT images, and demonstrate it on phantoms and in vivo mouse tumors.

Photothermal optical lock-in optical coherence tomography for in vivo imaging

Photothermal OCT (PTOCT) provides high sensitivity to molecular targets in tissue, and occupies a spatial imaging regime that is attractive for small animal imaging. However, current implementations of PTOCT require extensive temporal sampling, resulting in slow frame rates and a large data burden that limit its in vivo utility. To address these limitations, we have implemented optical lock-in techniques for photothermal optical lock-in OCT (poli-OCT), and demonstrated the in vivo imaging capabilities of this approach. The poli-OCT signal was assessed in tissue-mimicking phantoms containing indocyanine green (ICG), an FDA approved small molecule that has not been previously imaged in vivo with PTOCT. Then, the effects of in vivo blood flow and motion artifact were assessed and attenuated, and in vivo poli-OCT was demonstrated with both ICG and gold nanorods as contrast agents. Experiments revealed that poli-OCT signals agreed with optical lock-in theory and the bio-heat equation, and the system exhibited shot noise limited performance. In phantoms containing biologically relevant concentrations of ICG (1 µg/ml), the poli-OCT signal was significantly greater than control phantoms (p<0.05), demonstrating sensitivity to small molecules. Finally, in vivo poli-OCT of ICG identified the lymphatic vessels in a mouse ear, and also identified low concentrations (200 pM) of gold nanorods in subcutaneous injections at frame rates ten times faster than previously reported. This work illustrates that future in vivo molecular imaging studies could benefit from the improved acquisition and analysis times enabled by poli-OCT.

Uncoupling angiogenesis and inflammation in peripheral artery disease with therapeutic peptide-loaded microgels

Peripheral artery disease (PAD) is characterized by vessel occlusion and ischemia in the limbs. Treatment for PAD with surgical interventions has been showing limited success. Moreover, recent clinical trials with treatment of angiogenic growth factors proved ineffective as increased angiogenesis triggered severe inflammation in a proportionally coupled fashion. Hence, the overarching goal of this research was to address this issue by developing a biomaterial system that enables controlled, dual delivery of pro-angiogenic C16 and anti-inflammatory Ac-SDKP peptides in a minimally-invasive way. To achieve the goal, a peptide-loaded injectable microgel system was developed and tested in a mouse model of PAD. When delivered through multiple, low volume injections, the combination of C16 and Ac-SDKP peptides promoted angiogenesis, muscle regeneration, and perfusion recovery, while minimizing detrimental inflammation. Additionally, this peptide combination regulated inflammatory TNF-α pathways independently of MMP-9 mediated pathways of angiogenesis in vitro, suggesting a potential mechanism by which angiogenic and inflammatory responses can be uncoupled in the context of PAD. This study demonstrates a translatable potential of the dual peptide-loaded injectable microgel system for PAD treatment.

In vivo imaging of nanoparticle delivery and tumor microvasculature with multimodal optical coherence tomography

Current imaging techniques capable of tracking nanoparticles in vivo supply either a large field of view or cellular resolution, but not both. Here, we demonstrate a multimodality imaging platform of optical coherence tomography (OCT) techniques for high resolution, wide field of view in vivo imaging of nanoparticles. This platform includes the first in vivo images of nanoparticle pharmacokinetics acquired with photothermal OCT (PTOCT), along with overlaying images of microvascular and tissue morphology. Gold nanorods (51.8 ± 8.1 nm by 15.2 ± 3.3 nm) were intravenously injected into mice, and their accumulation into mammary tumors was non-invasively imaged in vivo in three dimensions over 24 hours using PTOCT. Spatial frequency analysis of PTOCT images indicated that gold nanorods reached peak distribution throughout the tumors by 16 hours, and remained well-dispersed up to 24 hours post-injection. In contrast, the overall accumulation of gold nanorods within the tumors peaked around 16 hours post-injection. The accumulation of gold nanorods within the tumors was validated post-mortem with multiphoton microscopy. This shows the utility of PTOCT as part of a powerful multimodality imaging platform for the development of nanomedicines and drug delivery technologies.

Quantitative optical imaging of vascular response in vivo in a model of peripheral arterial disease

The mouse hind limb ischemia (HLI) model is well established for studying collateral vessel formation and testing therapies for peripheral arterial disease, but there is a lack of quantitative techniques for intravitally analyzing blood vessel structure and function. To address this need, non-invasive, quantitative optical imaging techniques were developed to assess the time-course of recovery in the mouse HLI model. Hyperspectral imaging and optical coherence tomography (OCT) were used to non-invasively image hemoglobin oxygen saturation and microvessel morphology plus blood flow, respectively, in the anesthetized mouse after induction of HLI. Hyperspectral imaging detected significant increases in hemoglobin saturation in the ischemic paw as early as 3 days after femoral artery ligation (P < 0.01), and significant increases in distal blood flow were first detected with OCT 14 days postsurgery (P < 0.01). Intravital OCT images of the adductor muscle vasculature revealed corkscrew collateral vessels characteristic of the arteriogenic response to HLI. The hyperspectral imaging and OCT data significantly correlated with each other and with laser Doppler perfusion imaging (LDPI) and tissue oxygenation sensor data (P < 0.01). However, OCT measurements acquired depth-resolved information and revealed more sustained flow deficits following surgery that may be masked by more superficial measurements (LDPI, hyperspectral imaging). Therefore, intravital OCT may provide a robust biomarker for the late stages of ischemic limb recovery. This work validates non-invasive acquisition of both functional and morphological data with hyperspectral imaging and OCT. Together, these techniques provide cardiovascular researchers an unprecedented and comprehensive view of the temporal dynamics of HLI recovery in living mice.

Cross-sectional tracking of particle motion in evaporating drops: flow fields and interfacial accumulation

The lack of an effective technique for three-dimensional flow visualization has limited experimental exploration of the "coffee ring effect" to the two-dimensional, top-down viewpoint. In this report, high-speed, cross-sectional imaging of the flow fields was obtained using optical coherence tomography to track particle motion in an evaporating colloidal water drop. This approach enables z-dimensional mapping of primary and secondary flow fields and changes in these fields over time. These sectional images show that 1 μm diameter polystyrene particles have a highly nonuniform vertical distribution with particles accumulating at both the air-water interface and the water-glass interface during drop evaporation. Particle density and relative humidity are shown to influence interfacial entrapment, which suggests that both sedimentation rate and evaporation rate affect the dynamic changes in the cross-sectional distribution of particles. Furthermore, entrapment at the air-water interface delays the time at which particles reach the ring structure. These results suggest that the organization of the ring structure can be controlled based on the ratio of different density particles in a colloidal solution.

Non-invasive estimation of jugular venous oxygen saturation: a comparison between near infrared spectroscopy and transcutaneous venous oximetry

The ability of practitioners to assess the adequacy of global oxygen delivery is dependent on an accurate measurement of central venous saturation. Traditional techniques require the placement of invasive central venous access devices. This study aimed to compare two non-invasive technologies for the estimation of regional venous saturation (reflectance plethysmography and near infrared spectroscopy [NIRS]), using venous blood gas analysis as gold standard. Forty patients undergoing cardiac surgery were recruited in two groups. In the first group a reflectance pulse oximeter probe was placed on the skin overlying the internal jugular vein. In the second group, a Somanetics INVOS oximeter patch was placed on the skin overlying the internal jugular vein and overlying the ipsilateral cerebral hemisphere. Central venous catheters were placed in all patients. Oxygen saturation estimates from both groups were compared with measured saturation from venous blood. Twenty patients participated in each group.Data were analyzed by the limits of agreement technique suggested by Bland and Altman and by linear regression analysis. In the reflectance plethysmography group, the mean bias was 4.27% and the limits of agreement were 58.3 to -49.8% (r(2) = 0.00, p = 0.98). In the NIRS group the mean biases were 10.8% and 2.0% for the sensors attached over the cerebral hemisphere and over the internal jugular vein, respectively, and the limits of agreement were 33.1 to -11.4 and 19.5 to -15.5% (r(2) = 0.22, 0.28;p = 0.04, 0.03) for the cerebral hemisphere and internal jugular sites, respectively. While transcutaneous regional oximetry and NIRS have both been used to estimate venous and tissue oxygen saturation non-invasively, the correlation between estimates of ScvO(2) and SxvO(2) were statistically significant for near infrared spectroscopy, but not for transcutaneous regional oximetry. Placement of cerebral oximetry patches directly over the internal jugular vein (as opposed to on the forehead) appeared to approximate internal jugular venous saturation better (lower mean bias and tighter limits of agreement), which suggests this modality may with refinement offer the practitioner additional clinically useful information regarding global cerebral oxygen supply and demand matching.

Improved pressure-frequency sensing subxiphoid pericardial access system: performance characteristics during in vivo testing

We have designed, synthesized, and tested an improved version of our original subxiphoid access system intended to facilitate epicardial electrophysiology. The new version of the system incorporates a precision fiber-optic pressure sensor and a novel signal analysis algorithm for identifying pressure-frequency signatures which, in the clinical setting, may allow for safer access to the pericardial space. Following in vivo studies on ten adult canine models, we analyzed 215 pressure-frequency measurements made at the distal tip of the access needle, of which 98 were from nonpericardial, 112 were from pericardial, and five were from ventricular locations. The needle locations as identified by the algorithm were significantly different from each other (p &lt; 0.01), and the algorithm had improved performance when compared to a standard fast Fourier transform (FFT) analysis of the same data. Moreover, the structure of the algorithm can potentially overcome the time lags intrinsic to FFT analysis such that the needle's location can be determined in near-real time. Hydrodynamic pressure-frequency measurements made during traversal of the pericardial membrane revealed a distinct change in signal structure between the pericardial and nonpericardial anatomy. We present and discuss the design principles, details of construction, and performance characteristics of this system.

Anthropomorphic simulator for minimally invasive epicardial access procedures

We have developed a prototype, in vitro anthropomorphic model for simulating pressure-guided, subxiphoid access procedures done to enable minimally invasive epicardial cardiac procedures including treatments for ventricular tachycardia. Life-size replicas of the heart and lungs were modelled using anatomically accurate surrogates. The dynamic pressure-frequency profiles of simulated pericardial fluid surrounding the water-pumped replica heart were measured and validated against previously acquired human intrapericardial pressure observations (Pearson's r = 0.88, p < 0.001). In replicating access procedures for approaching and entering the pericardial space, the system produced physiologically appropriate pressure measurements at each intermediate point along the needle's insertion pathway. Details of construction and performance are presented and discussed.