Melanoma Imaging Research Articles

Comparing artificial intelligence algorithms to 157 German dermatologists: the melanoma classification benchmark

BackgroundSeveral recent publications have demonstrated the use of convolutional neural networks to classify images of melanoma at par with board-certified dermatologists. However, the non-availability of a public human benchmark restricts the comparability of the performance of these algorithms and thereby the technical progress in this field. MethodsAn electronic questionnaire was sent to dermatologists at 12 German university hospitals. Each questionnaire comprised 100 dermoscopic and 100 clinical images (80 nevi images and 20 biopsy-verified melanoma images, each), all open-source. The questionnaire recorded factors such as the years of experience in dermatology, performed skin checks, age, sex and the rank within the university hospital or the status as resident physician. For each image, the dermatologists were asked to provide a management decision (treat/biopsy lesion or reassure the patient). Main outcome measures were sensitivity, specificity and the receiver operating characteristics (ROC). ResultsTotal 157 dermatologists assessed all 100 dermoscopic images with an overall sensitivity of 74.1%, specificity of 60.0% and an ROC of 0.67 (range = 0.538–0.769); 145 dermatologists assessed all 100 clinical images with an overall sensitivity of 89.4%, specificity of 64.4% and an ROC of 0.769 (range = 0.613–0.9). Results between test-sets were significantly different (P < 0.05) confirming the need for a standardised benchmark. ConclusionsWe present the first public melanoma classification benchmark for both non-dermoscopic and dermoscopic images for comparing artificial intelligence algorithms with diagnostic performance of 145 or 157 dermatologists. Melanoma Classification Benchmark should be considered as a reference standard for white-skinned Western populations in the field of binary algorithmic melanoma classification.

Susceptibility‐weighted imaging in malignant melanoma brain metastasis

The value of cerebral susceptibility-weighted imaging (SWI) in malignant melanoma (MM) patients remains controversial and the effect of melanin on SWI is not well understood. To systematically analyze the spectrum of intracerebral findings in MM brain metastases (BM) on SWI and to determine the diagnostic value of SWI. Retrospective. In all, 100 patients with melanoma BM (69 having received radiotherapy [RT] and 31 RT-naïve) and a control group of 100 melanoma patients without BM were included. For detailed analysis of signal characteristics, 175 metastases were studied. Gradient echo SWI sequence at 1.5, 3.0, and 9.4 T. Signal characteristics from melanotic and amelanotic BMs on SWI with a focus on blooming artifacts were analyzed, as well as the presence and longitudinal dynamics of isolated SWI blooming artifacts in patients with and without BM. Chi-squared and Student's t-test were used for contingency table measures and group data of signal and clinical characteristics, respectively. Melanotic and amelanotic metastases did not show significant differences of SWI blooming artifacts (38% vs. 43%, P = 0.61). Most metastases without an initial SWI artifact developed a signal dropout during follow-up (80%; 65/81). Isolated SWI artifacts were detected more frequently in patients with BM (20 vs. 9, P = 0.03), of which the majority were found in patients who had received RT (17 vs. 3, P = 0.08). None of these isolated SWI blooming artifacts turned into overt metastases over time (median follow-up: 8.5 months). Similar findings persisted as remnants of successfully treated metastases (88%; 7/8). We conclude that SWI provides little additional diagnostic benefit over standard T1 -weighted imaging, as melanin content alone does not cause diagnostically relevant SWI blooming. Signal transition of SWI may rather indicate secondary phenomena like microbleeding and/or metal scavenging. Our results suggest that isolated SWI artifacts do not constitute vital tumor tissue but represent unspecific microbleedings, RT-related parenchymal changes or posttherapeutic remnants of former metastatic lesions. 3 Technical Efficacy Stage: 5 J. Magn. Reson. Imaging 2019;50:1251-1259.

Population-based validation of the National Cancer Comprehensive Network recommendations for baseline imaging workup of cutaneous melanoma

The aim of the current study is to assess the performance of some of the imaging scans recommended in the National Comprehensive Cancer Network Guidelines as part of baseline staging for cutaneous melanoma, regarding the detection of lung, brain, bone, and liver metastases. Surveillance, Epidemiology and End Results database (2010-2015) was used to extract the data, and cases with cutaneous melanoma and complete information about TN stages and sites of distant metastases were explored. Performance parameters assessed in the current study included positive predictive value (PPV), negative predictive value, sensitivity, specificity, number needed to investigate (NNI), and accuracy. A total of 109 971 patients were included in the analysis. If all stage III patients in the study cohort are to be staged through routine imaging, PPV (for the recognition of lung metastases) will be 2.9% and NNI to detect one case of lung metastasis will be 34. Likewise, PPV (for the recognition of bone metastases) will be 1.8% and NNI to detect one case of bone metastasis will be 55. Moreover, PPV (for the recognition of liver metastases) will be 1.8% and NNI to detect one case of liver metastasis will be 55. Excluding stage III patients with clinically node-negative/sentinel node-positive disease would improve PPV and decrease NNI for the three metastatic sites. Adherence to current National Comprehensive Cancer Network guidelines for cutaneous melanoma imaging for baseline staging results in low rates of failure to detect asymptomatic lung, liver, brain, or bone metastases.

Hyperspectral imaging in automated digital dermoscopy screening for melanoma

ObjectivesEarly melanoma detection decreases morbidity and mortality. Early detection classically involves dermoscopy to identify suspicious lesions for which biopsy is indicated. Biopsy and histological examination then diagnose benign nevi, atypical nevi, or cancerous growths. With current methods, a considerable number of unnecessary biopsies are performed as only 11% of all biopsied, suspicious lesions are actually melanomas. Thus, there is a need for more advanced noninvasive diagnostics to guide the decision of whether or not to biopsy. Artificial intelligence can generate screening algorithms that transform a set of imaging biomarkers into a risk score that can be used to classify a lesion as a melanoma or a nevus by comparing the score to a classification threshold. Melanoma imaging biomarkers have been shown to be spectrally dependent in Red, Green, Blue (RGB) color channels, and hyperspectral imaging may further enhance diagnostic power. The purpose of this study was to use the same melanoma imaging biomarkers previously described, but over a wider range of wavelengths to determine if, in combination with machine learning algorithms, this could result in enhanced melanoma detection.MethodsWe used the melanoma advanced imaging dermatoscope (mAID) to image pigmented lesions assessed by dermatologists as requiring a biopsy. The mAID is a 21‐wavelength imaging device in the 350–950 nm range. We then generated imaging biomarkers from these hyperspectral dermoscopy images, and, with the help of artificial intelligence algorithms, generated a melanoma Q‐score for each lesion (0 = nevus, 1 = melanoma). The Q‐score was then compared to the histopathologic diagnosis.ResultsThe overall sensitivity and specificity of hyperspectral dermoscopy in detecting melanoma when evaluated in a set of lesions selected by dermatologists as requiring biopsy was 100% and 36%, respectively.ConclusionWith widespread application, and if validated in larger clinical trials, this non‐invasive methodology could decrease unnecessary biopsies and potentially increase life‐saving early detection events. Lasers Surg. Med. 51:214–222, 2019. © 2019 The Authors. Lasers in Surgery and Medicine Published by Wiley Periodicals, Inc.

In vivo melanoma imaging based on dynamic nuclear polarization enhancement in melanin pigment of living mice using in vivo dynamic nuclear polarization magnetic resonance imaging

Melanin is a pigment that includes free radicals and is widely distributed in living animals. Malignant melanoma is one of the most progressive tumors in humans with increasing incidence worldwide, and has shown resistance to chemotherapy, resulting in high mortality at the metastatic stage. In general, melanoma involves the abnormal accumulation of melanin pigment produced by malignant melanocytes. Electron paramagnetic resonance (EPR) spectroscopy and imaging is a powerful technique to directly visualize melanomas using endogenous free radicals in the melanin pigment. Because melanin radicals have a large linewidth, the low spatial resolution of EPR imaging results in blurred images and a lack of anatomical information. Dynamic nuclear polarization (DNP)-MRI is a noninvasive imaging method to obtain the spatio-temporal information of free radicals with MRI anatomical resolution. Proton signals in tissues, including free radicals, can be dramatically enhanced by EPR irradiation at the resonance frequency of the free radical prior to applying the MRI pulse sequence. However, the DNP effects of free radicals in the pigment of living organisms is unclear. Therefore, if endogenous free radicals in melanin pigment could be utilized as a bio-probe for DNP-MRI, this will be an advantage for the specific enhancement of melanoma tissues and might allow the separate noninvasive visualization of melanoma tissues without the need for probe administration. Here, we report that biological melanin pigment induced a in vivo DNP effect by interacting with water molecules. In addition, we demonstrated in vivo melanoma imaging based on the DNP effects of endogenous free radicals in the melanin pigment of living mice.

111In-Labeled Glycoprotein Nonmetastatic b (GPNMB) Targeted Gemini Surfactant-Based Nanoparticles against Melanoma: In Vitro Characterization and in Vivo Evaluation in Melanoma Mouse Xenograft Model.

Melanoma is a devastating form of skin cancer with high tendency to metastasis. This work addresses the development of new targeted nanoparticles that can be used for single-photon emission computed tomography (SPECT) imaging of melanoma. Melanoma-specific glycoprotein nonmetastatic b (GPNMB) antigen targeted and nontargeted gemini nanoparticles were prepared, characterized, and radiolabeled with 111In. 111In-labeled nanoparticles were composed of gemini surfactant grafted with monoclonal antibody Fab fragment that targeted GPNMB. Specific uptake of GPNMB-Fab was studied in six melanoma cell lines using flow cytometry. In vitro cellular uptake and internalization were studied using flow cytometry, confocal laser scanning microscopy, and radiometric techniques. Specific uptake of anti-GPNMB targeted nanoparticles was observed in GPNMB expressing cells, which was higher than low expressing or control cells. In vitro studies showed that conjugation of GPNMB targeted nanoparticles led to enhanced intracellular uptake of the nanodelivery system, which is critical for drug delivery. In vivo distribution of the nanoparticles was studied by microSPECT/CT imaging and ex vivo biodistribution. Tumor uptake was significantly higher ( p < 0.05) in nontargeted nanoparticles (5.47 ± 0.46%IA/cc) compared to GPNMB targeted nanoparticles (1.87 ± 0.27% ID/cc), which might be attributed to the high spleen uptake of the targeted formulation. These findings demonstrated that the radiolabeled gemini nanoparticles are promising for image-guided radiotherapy of melanoma. Formulation optimization is needed to improved tumor uptake and in vivo intracellular delivery for radiotherapeutic applications.

Bioimaging: Microwave‐Assisted Synthesis of Quasi‐Pyramidal CuInS2–ZnS Nanocrystals for Enhanced Near‐Infrared Targeted Fluorescent Imaging of Subcutaneous Melanoma (Adv. Biosys. 1/2019)

Bioimaging: Microwave‐Assisted Synthesis of Quasi‐Pyramidal CuInS₂–ZnS Nanocrystals for Enhanced Near‐Infrared Targeted Fluorescent Imaging of Subcutaneous Melanoma (Adv. Biosys. 1/2019)

Cyclic RGD-Functionalized and Disulfide-Crosslinked Iodine-Rich Polymersomes as a Robust and Smart Theranostic Agent for Targeted CT Imaging and Chemotherapy of Tumor.

There is tremendous interest in integrating CT imaging with chemotherapy; however, reported iodine-based nanosystems such as nanogels and nano-emulsions display typically reduced contrast coefficient, low drug loading and stability, and poor targetability. Here, cRGD-functionalized disulfide-crosslinked iodine-rich polymersomes (cRGD-XIPs) were designed as a novel, robust and smart theranostic agent and investigated for targeted CT imaging and chemotherapy of malignant tumors.Methods: cRGD-XIPs were prepared from co-self-assembly of poly(ethylene glycol)-b-poly(dithiolane trimethylene carbonate-co-iodinated trimethylene carbonate) (PEG-P(DTC-IC)) and cRGD-PEG-P(DTC-IC) block copolymers. In vitro and in vivo CT contrast effect of cRGD-XIPs was studied using αvβ3-overexpressing B16 melanoma as a tumor model in comparison with clinical agent iohexol. The therapeutic efficacy of doxorubicin-loaded cRGD-XIPs (cRGD-XIPs-Dox) to B16 melanoma was investigated and compared with XIPs-Dox (non-targeted), cRGD-IPs-Dox (non-crosslinked) and free Dox.Results: cRGD-XIPs were formed with 55.5 wt.% iodine and ca. 90 nm in diameter. cRGD-XIPs-Dox with a Dox loading of 15.3 wt.% bared superior colloidal stability and reduction-responsive drug release. Notably, blank cRGD-XIPs showed a maximum-tolerated dose (MTD) > 400 mg iodine equiv./kg while cRGD-XIPs-Dox had an MTD > 150 mg Dox equiv./kg, ca. 15-fold improvement over free Dox. cRGD-XIPs revealed superior CT contrast effect and achieved 46.5- and 24.0-fold better enhancement of CT imaging of B16 melanoma than iohexol at 4 h following intratumoral and intravenous injection, respectively. cRGD-XIPs-Dox displayed an elimination half-life of 6.5 h and an elevated accumulation of 6.68% ID/g in the tumors. Furthermore, cRGD-XIPs-Dox was significantly more effective than XIPs-Dox and cRGD-XPs-Dox in inhibiting growth of B16 melanoma model.Conclusion: This proof-of-concept study demonstrates that cRGD-XIPs are a robust, non-toxic and smart polymeric theranostic agent that can not only significantly enhance CT imaging of tumors but also mediate efficient tumor-targeted chemotherapy. XIPs offer a unique and safe platform for theranostic polymersomes that pre-select patients using CT imaging prior to targeted chemotherapy with the same system.

On the spectral signature of melanoma: a non-parametric classification framework for cancer detection in hyperspectral imaging of melanocytic lesions.

Early detection and diagnosis is a must in secondary prevention of melanoma and other cancerous lesions of the skin. In this work, we present an online, reservoir-based, non-parametric estimation and classification model that allows for this functionality on pigmented lesions, such that detection thresholding can be tuned to maximize accuracy and/or minimize overall false negative rates. This system has been tested in a dataset consisting of 116 patients and a total of 124 hyperspectral images of nevi, raised nevi and melanomas, detecting up to 100% of the suspicious lesions at the expense of some false positives.

68Ga-DOTA-GGNle-CycMSHhex targets the melanocortin-1 receptor for melanoma imaging.

Melanocortin-1 receptor (MC1R) is a molecular target for melanoma imaging and therapy because of its overexpression on rodent and human melanoma cells. Here, we evaluated the MC1R targeting and specificity of 68Ga-DOTA-GGNle-CycMSHhex and Cy5.5-GGNle-CycMSHhex using murine and human melanoma cells, and murine and xenografted tumors. 68Ga-DOTA-GGNle-CycMSHhex was used first in human as an imaging probe to evaluate the possibility of radionuclide therapy in patients with advanced-stage melanoma. 68Ga-DOTA-GGNle-CycMSHhex and Cy5.5-GGNle-CycMSHhex displayed MC1R-specific targeting properties in murine and human melanoma cells, as well as in murine melanoma and human melanoma-xenografted tumors. Both B16/F10 and M21 melanoma lesions could be easily imaged by positron emission tomography using 68Ga-DOTA-GGNle-CycMSHhex The first-in-human images of melanoma brain metastases in patients demonstrated the clinical relevance of MC1R as a molecular target for melanoma imaging, highlighting the potential of 68Ga-DOTA-GGNle-CycMSHhex as an MC1R-targeting melanoma imaging probe and underscoring the need to develop MC1R-targeting therapeutic agents for treating patients with metastatic melanoma.

Microwave‐Assisted Synthesis of Quasi‐Pyramidal CuInS2–ZnS Nanocrystals for Enhanced Near‐Infrared Targeted Fluorescent Imaging of Subcutaneous Melanoma

Near-infrared (NIR) fluorescent CuInS2 -ZnS nanocrystals (CIZS NCs) are synthesized via an ultra-fast, non-injection microwave (MW)-assisted nanoalloying process at 230 ºC within 5 min using 1-dodecanethiol (DDT) as both the sulfur source and solvent under solvothermal (ST) condition. The structural and surface analyses reveal that DDT-functionalized CIZS NCs exhibit quasi-pyramids of tetragonal-phase with well-defined facets. The DDT-functionalized CIZS NCs present a photoluminescence quantum yield (PLQY) of 76% and a long-lived fluorescence lifetime of ≈0.6 µs in organic-phase. Subsequently, DDT-functionalized CIZS NCs are phase-transferred via ligand-exchange using 11-mercaptoundecanoic acid (MUA) into water-soluble MUA-CIZS NCs that exhibit a substantial PLQY of 55%. In addition, the NIR-fluorescent MUA-functionalized CIZS NCs in conjugation with folic acid (FA), as a tumor-targeting ligand, demonstrates enhanced tumor-targeted imaging ability. The FA-MUA-CIZS NC conjugates exhibit a cell viability of ≈75% even at the highest concentration of 1 mg mL-1 and a labeling efficiency of 95.4%. The in vivo imaging results corroborate that FA-MUA-CIZS NCs conjugates are actively targeted to folate receptor-positive B16F10 tumor-bearing C57BL/6 mice in 2 h. The histopathological and hematological studies confirm no significant changes in tissue architecture and blood biochemical parameters. The confocal microscopy studies reveal deep penetration and uniform distribution of FA-MUA-CIZS NCs conjugates in subcutaneous melanoma.

Line-field confocal optical coherence tomography for high-resolution noninvasive imaging of skin tumors.

An optical technique called line-field confocal optical coherence tomography (LC-OCT) is introduced for high-resolution, noninvasive imaging of human skin in vivo. LC-OCT combines the principles of time-domain optical coherence tomography and confocal microscopy with line illumination and detection using a broadband laser and a line-scan camera. LC-OCT measures the echo-time delay and amplitude of light backscattered from cutaneous microstructures through low-coherence interferometry associated with confocal spatial filtering. Multiple A-scans are acquired simultaneously while dynamically adjusting the focus. The resulting cross-sectional B-scan image is produced in real time at 10 frame / s. With an isotropic spatial resolution of ∼1 μm, the LC-OCT images reveal a comprehensive structural mapping of skin at the cellular level down to a depth of ∼500 μm. LC-OCT has been applied to the imaging of various skin lesions, in vivo, including carcinomas and melanomas. LC-OCT images are found to strongly correlate with conventional histopathological images. The use of LC-OCT as an adjunct tool in medical practice could significantly improve clinical diagnostic accuracy while reducing the number of biopsies of benign lesions.

The easy-to-hard training advantage with real-world medical images

Many medical professions require practitioners to perform visual categorizations in domains such as radiology, dermatology, and neurology. However, acquiring visual expertise is tedious and time-consuming and the perceptual strategies mediating visual categorization skills are poorly understood. In this paper, the Ease algorithm was developed to predict an item’s categorization difficulty (Ease value) based on the item’s perceptual similarity to all within-category items versus between-category items in the dataset. In this study, Ease values were used to construct an easy-to-hard and hard-to-easy training schedule for teaching melanoma diagnosis. Whereas previous visual training studies suggest that an easy-to-hard schedule benefits learning outcomes, no studies to date have demonstrated the easy-to-hard advantage with complex, real-world images. In our study, 237 melanoma and benign images were collected for training and testing purposes. The diagnostic accuracy of images was verified by an expert dermatologist. Based on their Ease values, the items were grouped into easy, medium, and hard categories, each containing an equal number of melanoma and benign lesions. During training, participants categorized images of skin lesions as either benign or melanoma and were given corrective feedback after each trial. In the easy-to-hard training condition, participants learned to categorize all the easy items first, followed by the medium items, and finally the hard items. Participants in the hard-to-easy training condition learned items in the reverse order. Post-training results showed that training in both conditions transferred to the classification of new melanoma and benign images. Participants in the easy-to-hard condition showed modest advantages both in the acquisition and retention of the melanoma diagnosis skills, but neither scheduling condition exhibited a gross advantage. The Ease values of the items predicted categorization accuracy after, but not before training, suggesting that the Ease algorithm is a promising tool for optimizing medical training in visual categorization.

Synthesis and Evaluation of New Bifunctional Chelators with Phosphonic Acid Arms for Gallium-68 Based PET Imaging in Melanoma.

Due to the increasing use of generator-produced radiometal Gallium-68 (68Ga) in positron-emission tomography/computed tomography (PET/CT), reliable bifunctional chelators that can efficiently incorporate 68Ga3+ into biomolecules are highly desirable. In this study, we synthesized two new bifunctional chelators bearing one or two phosphonic acid functional groups, named p-SCN-PhPr-NE2A1P and p-SCN-PhPr-NE2P1A, with the aim of enabling facile production of 68Ga-based radiopharmaceuticals. Both chelators were successfully conjugated to LLP2A-PEG4, a very late antigen-4 (VLA-4) targeting peptidomimetic ligand, to evaluate their application in 68Ga-based PET imaging. NE2P1A-PEG4-LLP2A exhibited the highest 68Ga3+ binding ability with molar activity of 37 MBq/nmol under mild temperature and neutral pH. Excellent serum stability of 68Ga-NE2P1A-PEG4-LLP2A was observed, which was consistent with the result obtained from density functional theory calculation. The in vitro cell study showed that 68Ga-NE2P1A-PEG4-LLP2A had significantly longer retention in B16F10 cells comparing to the reported retention of 64Cu-NE3TA-PEG4-LLP2A, although the uptake was relatively lower. In the biodistribution and micro-PET/CT imaging studies, high tumor uptake and low background were observed after 68Ga-NE2P1A-PEG4-LLP2A was injected into mice bearing B16F10 tumor xenografts, making it a highly promising radiotracer for noninvasive imaging of VLA-4 receptors overexpressed in melanoma.

Visual Cancer Communication on Social Media: An Examination of Content and Effects of #Melanomasucks.

BackgroundInstagram is increasingly becoming a platform on which visual communication of cancer takes place, but few studies have investigated the content and effects. In particular, a paucity of research has evaluated the effects of visual communication of cancer on participative engagement outcomes.ObjectiveThe objective of our study was to investigate cancer-related beliefs and emotions shared on Instagram and to examine their effects on participative engagement outcomes including likes, comments, and social support.MethodsThis study analyzed the content of 441 posts of #melanomasucks on Instagram and assessed the effects of the content characteristics on outcomes, including the number of likes and comments and types of social support using group least absolute shrinkage and selection operator logistic regression.ResultsPosts about controlling melanoma were most frequent (271/441, 61.5%), followed by 240 (54.4%) posts about outcomes of having melanoma. Ninety posts (20.4%) were about the causes of melanoma. A greater number of posts expressed positive (159/441, 36.1%) than negative emotions (100/441, 22.7%). Eighty posts (18.1%) expressed hope, making it the most frequently expressed emotion; 49 posts expressed fear (11.1%), 46 were humorous (10.4%), and 46 showed sadness (10.4%). Posts about self behavior as a cause of melanoma decreased likes (P<.001) and social support comments (P=.048). Posts about physical consequences of melanoma decreased likes (P=.02) but increased comments (P<.001) and emotional social support (P<.001); posts about melanoma treatment experience increased comments (P=.03) and emotional social support (P<.001). None of the expressions of positive emotions increased likes, comments, or social support. Expression of anger increased the number of likes (P<.001) but those about fear (P<.001) and joy (P=.006) decreased the number of likes. Posts about fear (P=.003) and sadness (P=.003) increased emotional social support. Posts showing images of melanoma or its treatment on the face or body parts made up 21.8% (96/441) of total posts. Inclusion of images increased the number of comments (P=.001).ConclusionsTo our knowledge, this is the first investigation of the content and effects of user-generated visual cancer communication on social media. The findings show where the self-expressive and social engagement functions of #melanomasucks converge and diverge, providing implications for extending research on the commonsense model of illness and for developing conceptual frameworks explaining participative engagement on social media.

Near-infrared fluorescence lymphatic imaging in vascular endothelial growth factor-C overexpressing murine melanoma.

In this study we employ a near-infrared fluorescence lymphatic imaging (NIRFLI) technique to longitudinally image spatial and temporal changes in the lymphatics in mice bearing vascular endothelial growth factor (VEGF)-C overexpressing B16F10 (VEGF-C-B16F10) or mock-transduced B16F10 (mock-B16F10) melanoma tumors. Our NIRFLI data show that ICG-laden lymph accumulates into a VEGF-C-B16F10 tumor compared to mock-B16F10 at 3 days post implantation, presumably due to increased lymphatic vessel permeability. Quantification shows a significantly greater percentage of ICG-perfused area in VEGF-C-B16F10 (7.6 ± 2) as compared to MOCK-B16F10 (1 ± 0.5; p = 0.02), which is also confirmed by quantification of the lymphatic leakage of evans blue dye (optical density at 610nm; VEGF-C-B16F10, 10.5 ± 2; mock-B16F10, 5.1 ± 0.5; p = 0.009); thereafter, lymphatic leakage is visualized only in the peritumoral region. Our imaging data also show that anti-VEGF-C treatment in VEGF-C-B16F10 restores normal lymphatic vessel integrity and reduces dye extravasation. Because NIRFLI technology can be used to non-invasively detect lymphatic changes associated with cancer, it may provide a new diagnostic to assess the lack of lymphatic vessel integrity that promotes lymphovascular invasion and to assess therapies that could arrest invasion through normalization of the lymphatic vasculature.

Melanoma Recognition in Dermoscopy Images via Aggregated Deep Convolutional Features.

In this paper, we present a novel framework for dermoscopy image recognition via both a deep learning method and a local descriptor encoding strategy. Specifically, deep representations of a rescaled dermoscopy image are first extracted via a very deep residual neural network pretrained on a large natural image dataset. Then these local deep descriptors are aggregated by orderless visual statistic features based on Fisher vector (FV) encoding to build a global image representation. Finally, the FV encoded representations are used to classify melanoma images using a support vector machine with a Chi-squared kernel. Our proposed method is capable of generating more discriminative features to deal with large variations within melanoma classes, as well as small variations between melanoma and nonmelanoma classes with limited training data. Extensive experiments are performed to demonstrate the effectiveness of our proposed method. Comparisons with state-of-the-art methods show the superiority of our method using the publicly available ISBI 2016 Skin lesion challenge dataset.

The role of routine imaging in pediatric cutaneous melanoma.

Optimal imaging for children with pediatric malignant melanoma (MM) is unknown. We reviewed clinical and imaging findings of patients with American Joint Commission on Cancer (AJCC) stage IIC-IV MM treated on our institutional MEL06 trial. All patients had baseline brain magnetic resonance imaging/computed tomography (MRI/CT), positron emission tomography/computed tomography (PET/CT), CT chest, abdomen, and pelvis (CTCAP). Patients on stratum A (PEG-interferon, where PEG is pegylated; AJCC IIC, IIIA, IIIB; n=16) had imaging every 6months; stratum B1 (PEG-interferon and temozolomide; unresectable measurable disease, metastatic, or recurrent; n=2) had PET/CT scans every 2months and brain imaging studies every 4months; stratum B2 patients (PEG-interferon and temozolomide; unresectable nonmeasurable, metastatic, or recurrent, n=3) had imaging every 4months. Off-therapy imaging was done every 6 months for 3 years. There were 21 patients (11 females, 11 spitzoid, median age 14years, head/neck [6], trunk [7], extremities [8]). Patients with spitzoid melanoma underwent 236 imaging studies in total (86 PET/CT, 81 CTCAP, 11 CT chest, 10 CT brain, 48 MRI brain) at a median cost per patient of $32,718. Thirteen studies (5.8%) had findings that led to two biopsies (one positive). For conventional MM, 162 studies (61 PET/CT, 57 CTCAP, 8 CT chest, 7 CT brain, and 29 MRI brain) were performed with a median cost per patient of $23,420. Twenty (14%) had findings leading to six biopsies (four positive). At 6.3years (range 0.4-9.2), 17 patients remain disease-free. Children with spitzoid melanoma require minimal imaging at diagnosis and follow-up. Patients with conventional MM should be imaged according to adult guidelines.

Label-Free Imaging of Melanoma with Confocal Photothermal Microscopy: Differentiation between Malignant and Benign Tissue

Label-free confocal photothermal (CPT) microscopy was utilized for the first time to investigate malignancy in mouse skin cells. Laser diodes (LDs) with 405 nm or 488 nm wavelengths were used as pumps, and a 638 nm LD was used as a probe for the CPT microscope. A Grey Level Cooccurrence Matrix (GLCM) for texture analysis was applied to the CPT images. Nine GLCM parameters were calculated with definite definitions for the intracellular super-resolved CPT images, and the parameters Entropy, Contrast, and Variance were found to be most suited among the nine parameters to discriminate clearly between healthy cells and malignant cells when a 405 nm pump was used. Prominence, Variance, and Shade were most suited when a pump wavelength of 488 nm was used.

Surface-Coil MRI for Small Peripheral Choroidal Melanoma: Imaging in a Rabbit Eye Model

Introduction: Imaging of small peripheral choroidal melanomas remains challenging in clinical and experimental settings, requiring wide-field ophthalmoscopy modalities. In this study, we applied surface-coil MRI to assess imaging feasibility for small peripherally located choroidal melanoma in a rabbit eye model. Methods: A female albino rabbit with a body weight of 2.8 kg was immunosuppressed with intramuscular cyclosporine 15 mg/kg daily for 3 days prior to inoculation of 14 million 92.1 line melanoma cells into the peripheral suprachoroidal space of the right eye. The experimental animal continued to receive the same dosage of cyclosporine daily for 26 days postinoculation until the day of sacrifice. On that interval, it underwent surface-coil MRI scanning using various protocols and models subsequently sacrificed to allow histological evaluation. A comparative examination of various MRI sequences for optimal lesion imaging was performed, based on semiquantitative criteria. Results: Imaging quality and superimposed artifacts varied across different sequences and the optimal tumor delineation from adjacent tissues was achieved in T1-weighted sequences. Conclusions: MRI of small peripheral choroidal melanomas is feasible in a rabbit model using surface coils, particularly in T1-weighted sequences. Findings may be used as a basis for further development of this technique for experimental and clinical applications.