Hyperspectral Imaging for Non-invasive Diagnostics of Melanocytic Lesions.

Malignant melanoma poses a clinical diagnostic problem, since a large number of benign lesions are excised to find a single melanoma. This study assessed the accuracy of a novel non-invasive diagnostic technology, hyperspectral imaging, for melanoma detection. Lesions were imaged prior to excision and histopathological analysis. A deep neural network algorithm was trained twice to distinguish between histopathologically verified malignant and benign melanocytic lesions and to classify the separate subgroups. Furthermore, 2 different approaches were used: a majority vote classification and a pixel-wise classification. The study included 325 lesions from 285 patients. Of these, 74 were invasive melanoma, 88 melanoma in situ, 115 dysplastic naevi, and 48 non-dysplastic naevi. The study included a training set of 358,800 pixels and a validation set of 7,313 pixels, which was then tested with a training set of 24,375 pixels. The majority vote classification achieved high overall sensitivity of 95% and a specificity of 92% (95% confidence interval (95% CI) 0.024-0.029) in differentiating malignant from benign lesions. In the pixel-wise classification, the overall sensitivity and specificity were both 82% (95% CI 0.005-0.005). When divided into 4 subgroups, the diagnostic accuracy was lower. Hyperspectral imaging provides high sensitivity and specificity in distinguishing between naevi and melanoma. This novel method still needs further validation.

Hyperspectral Imaging Reveals Spectral Differences and Can Distinguish Malignant Melanoma from Pigmented Basal Cell Carcinomas: A Pilot Study.

Pigmented basal cell carcinomas can be difficult to distinguish from melanocytic tumours. Hyperspectral imaging is a non-invasive imaging technique that measures the reflectance spectra of skin in vivo. The aim of this prospective pilot study was to use a convolutional neural network classifier in hyperspectral images for differential diagnosis between pigment-ed basal cell carcinomas and melanoma. A total of 26 pigmented lesions (10 pigmented basal cell carcinomas, 12 melanomas in situ, 4 invasive melanomas) were imaged with hyperspectral imaging and excised for histopatho-logical diagnosis. For 2-class classifier (melano-cytic tumours vs pigmented basal cell carcinomas) using the majority of the pixels to predict the class of the whole lesion, the results showed a sensitivity of 100% (95% confidence interval 81-100%), specificity of 90% (95% confidence interval 60-98%) and positive predictive value of 94% (95% confidence interval 73-99%). These results indicate that a convolutional neural network classifier can differentiate melanocytic tumours from pigmented basal cell carcinomas in hyperspectral images. Further studies are warranted in order to confirm these preliminary results, using larger samples and multiple tumour types, including all types of melanocytic lesions.

Digital Quantification of Melanocytic Density in Resection Margins of Lentigo Maligna Using SOX10 Versus Hematoxylin-Eosin Staining.

ABSTRACT: Lentigo maligna (LM) represents an overgrowth of atypical melanocytes at the dermal-epidermal junction of chronically sun-damaged skin. The presence of LM on sun-damaged skin poses a diagnostic challenge because the solar-induced melanocytic hyperplasia makes it difficult to assess the LM margins. Melanocytic density can be used to discriminate sun-damaged skin from LM. The aim of this study was to quantify the melanocytic density at the surgical margins of scanned whole-slide images of LM comparing sections stained with H&E and SOX10. Twenty-six surgically excised LM diagnosed at the Department of Pathology at Sahlgrenska University Hospital were collected. The slides that contained the closest surgical margin or harbored the highest density of melanocytes at the margin were selected for serial sectioning using H&E and SOX10. Whole-slide imaging at ×40 magnification was used, and a circular field with a diameter of 0.5 mm at the surgical margin was superimposed on the image. Five blinded pathologists reviewed the slides in a randomized order. In the majority of the cases (24/26), the pathologists identified more melanocytes on the SOX10 slides than those on the H&E slides. On average, 2.5 times more melanocytes were counted using SOX10 compared with H&E (P < 0.05). Furthermore, the average group SD on the H&E slides was 4.12 compared with 2.83 on the SOX10 slides (P = 0.004). Thus, the use of SOX10 staining leads to higher melanocytic density counts compared with H&E staining when assessing the surgical margins of LM. The use of SOX10 staining also significantly decreased the interobserver variability between pathologists.

Evaluation of the Diagnostic Accuracy of an Online Artificial Intelligence Application for Skin Disease Diagnosis.

Artificial intelligence (AI) algorithms for automated classification of skin diseases are available to the consumer market. Studies of their diagnostic accuracy are rare. We assessed the diagnostic accuracy of an open-access AI application (Skin Image Search™) for recognition of skin diseases. Clinical images including tumours, infective and inflammatory skin diseases were collected at the Department of Dermatology at the Sahlgrenska University Hospital and uploaded for classification by the online application. The AI algorithm classified the images giving 5 differential diagnoses, which were then compared to the diagnoses made clinically by the dermatologists and/or histologically. We included 521 images portraying 26 diagnoses. The diagnostic accuracy was 56.4% for the top 5 suggested diagnoses and 22.8% when only considering the most probable diagnosis. The level of diagnostic accuracy varied considerably for diagnostic groups. The online application demonstrated low diagnostic accuracy compared to a dermatologist evaluation and needs further development.

Deep learning model shows pathologist-level detection of sentinel node metastasis of melanoma and intra-nodal nevi on whole slide images.

Introduction: Nodal metastasis (NM) in sentinel node biopsies (SNB) is crucial for melanoma staging. However, an intra-nodal nevus (INN) may often be misclassified as NM, leading to potential misdiagnosis and incorrect staging. There is high discordance among pathologists in assessing SNB positivity, which may lead to false staging. Digital whole slide imaging offers the potential for implementing artificial intelligence (AI) in digital pathology. In this study, we assessed the capability of AI to detect NM and INN in SNBs. Methods: A total of 485 hematoxylin and eosin whole slide images (WSIs), including NM and INN from 196 SNBs, were collected and divided into training (279 WSIs), validation (89 WSIs), and test sets (117 WSIs). A deep learning model was trained with 5,956 manual pixel-wise annotations. The AI and three blinded dermatopathologists assessed the test set, with immunohistochemistry serving as the reference standard. Results: The AI model showed excellent performance with an area under the curve receiver operating characteristic (AUC) of 0.965 for detecting NM. In comparison, the AUC for NM detection among dermatopathologists ranged between 0.94 and 0.98. For the detection of INN, the AUC was lower for both AI (0.781) and dermatopathologists (range of 0.63-0.79). Discussion: In conclusion, the deep learning AI model showed excellent accuracy in detecting NM, achieving dermatopathologist-level performance in detecting both NM and INN. Importantly, the AI model showed the potential to differentiate between these two entities. However, further validation is warranted.

Utilizing deep learning model for assessing melanocytic density in resection margins of lentigo maligna.

BACKGROUND: Surgical excision with clear histopathological margins is the preferred treatment to prevent progression of lentigo maligna (LM) to invasive melanoma. However, the assessment of resection margins on sun-damaged skin is challenging. We developed a deep learning model for detection of melanocytes in resection margins of LM. METHODS: In total, 353 whole slide images (WSIs) were included. 295 WSIs were used for training and 58 for validation and testing. The algorithm was trained with 3,973 manual pixel-wise annotations. The AI analyses were compared to those of three blinded dermatopathologists and two pathology residents, who performed their evaluations without AI and AI-assisted. Immunohistochemistry (SOX10) served as the reference standard. We used a dichotomized cutoff for low and high risk of recurrence (≤ 25 melanocytes in an area of 0.5 mm for low risk and > 25 for high risk). RESULTS: The AI model achieved an area under the receiver operating characteristic curve (AUC) of 0.84 in discriminating margins with low and high recurrence risk. In comparison, the AUC for dermatopathologists ranged from 0.72 to 0.90 and for the residents in pathology, 0.68 to 0.80. Additionally, with aid of the AI model the performance of two pathologists significantly improved. CONCLUSIONS: The deep learning showed notable accuracy in detecting resection margins of LM with a high versus low risk of recurrence. Furthermore, the use of AI improved the performance of 2/5 pathologists. This automated tool could aid pathologists in the assessment or pre-screening of LM margins.

ToF-SIMS imaging reveals changes in tumor cell lipids during metastatic progression of melanoma.

Most melanomas progress from radial to vertical growth phase before spreading locoregionally and distally. Much is still unknown about the metabolic changes in the tumor cells and their microenvironment during this metastatic progression. We aimed to gain new insight into the molecular characteristics of melanoma in regard to spatial lipidomics to deliver new knowledge regarding tumor metastatic progression. We included 10 fresh tumor samples from 10 patients including two in situ melanomas, two invasive primary melanomas, and six metastatic melanomas (four in-transit metastases and two distant metastases). In addition, we analyzed four healthy skin controls from the same patients. Time-of-flight imaging secondary ion mass spectrometry (ToF-SIMS) enabled detailed spatial-lipidomics that could be directly correlated with conventional histopathological analysis of consecutive H&E-stained tissue sections. Significant differences in the lipid profiles were found in primary compared to metastatic melanomas, notably an increase in phosphatidylethanolamine lipids relative to phosphatidylinositol lipids and an increase in GM3 gangliosides in the metastatic samples. Furthermore, analysis of the data from in transit versus distant metastases samples highlighted that specific phospholipids, and a difference in the long versus shorter chain GM3 gangliosides, discriminated the metastatic routes. Further studies are warranted to verify these preliminary findings. Lipidomic changes could serve as a novel biomarker for tumor progression and even serve as a target for novel treatments. Furthermore, analyzing the lipid profiles could help to differentiate between primary and metastatic melanomas in challenging cases.

Weakly supervised detection and classification of basal cell carcinoma using graph-transformer on whole slide images.

The high incidence rates of basal cell carcinoma (BCC) cause a significant burden at pathology laboratories. The standard diagnostic process is time-consuming and prone to inter-pathologist variability. Despite the application of deep learning approaches in grading of other cancer types, there is limited literature on the application of vision transformers to BCC on whole slide images (WSIs). A total of 1832 WSIs from 479 BCCs, divided into training and validation (1435 WSIs from 369 BCCs) and testing (397 WSIs from 110 BCCs) sets, were weakly annotated into four aggressivity subtypes. We used a combination of a graph neural network and vision transformer to (1) detect the presence of tumor (two classes), (2) classify the tumor into low and high-risk subtypes (three classes), and (3) classify four aggressivity subtypes (five classes). Using an ensemble model comprised of the models from cross-validation, accuracies of 93.5%, 86.4%, and 72% were achieved on two, three, and five class classifications, respectively. These results show high accuracy in both tumor detection and grading of BCCs. The use of automated WSI analysis could increase workflow efficiency.

Exploring reflectance confocal microscopy as a non-invasive diagnostic tool for genital lichen sclerosus.

The diagnosis of genital lichen sclerosus (LS) is often confirmed by obtaining a skin biopsy, which can lead to unwanted complications and is uncomfortable in the sensitive genital area. Thus, there is a need of finding novel, non-invasive techniques that can rapidly and accurately diagnose LS. The present study investigated the potential for reflectance confocal microscopy (RCM) to diagnose LS compared with healthy penile skin and other common penile skin disorders in males. A total of 30 male patients, including patients with LS, nonspecific balanoposthitis, plasma cell balanitis and psoriasis, and healthy individuals were included and were subject to non-invasive RCM investigation. Prominent fiber-like structures, representing hyaline sclerosis, were observed in the RCM images for almost half of the patients. Differences between healthy penile skin and LS were confirmed by identifying the edged papillae on healthy skin and their absence or obscureness in patients with LS. Notably, RCM could detect the atypical honeycomb pattern referring to dysplasia in 1 patient with LS with penile intraepithelial neoplasia. In conclusion, the present study demonstrated that RCM can detect sclerosis in penile LS. RCM can potentially become a valuable tool for monitoring patients with LS for dysplasia providing a useful non-invasive diagnostic tool for genital disorders.

Generating Hyperspectral Skin Cancer Imagery using Generative Adversarial Neural Network.

In this study we develop a proof of concept of using generative adversarial neural networks in hyperspectral skin cancer imagery production. Generative adversarial neural network is a neural network, where two neural networks compete. The generator tries to produce data that is similar to the measured data, and the discriminator tries to correctly classify the data as fake or real. This is a reinforcement learning model, where both models get reinforcement based on their performance. In the training of the discriminator we use data measured from skin cancer patients. The aim for the study is to develop a generator for augmenting hyperspectral skin cancer imagery.

TOF-SIMS imaging reveals tumor heterogeneity and inflammatory response markers in the microenvironment of basal cell carcinoma.

Basal cell carcinoma (BCC) is the most common skin malignancy. In fact, it is as common as the sum of all other skin malignancies combined and the incidence is rising. In this focused and histology-guided study, tissue from a patient diagnosed with aggressive BCC was analyzed by imaging mass spectrometry in order to probe the chemistry of the complex tumor environment. Time-of-flight secondary ion mass spectrometry using a (CO2)6 k + gas cluster ion beam allowed a wide range of lipid species to be detected. Their distributions were then imaged in the tissue that contained small tumor islands that were histologically classified as more/less aggressive. Maximum autocorrelation factor (MAF) analysis highlighted chemical differences between the tumors and the surrounding stroma. A closer inspection of the distribution of individual ions, selected based on the MAF loadings, showed heterogeneity in signal between different microtumors, suggesting the potential of chemically grading the aggressiveness of each individual tumor island. Sphingomyelin lipids were found to be located in stroma containing inflammatory cells.

Attitudes Toward Artificial Intelligence Within Dermatopathology: An International Online Survey.

Background: Artificial intelligence (AI) has recently surfaced as a research topic in dermatology and dermatopathology. In a recent survey, dermatologists were overall positive toward a development with an increased use of AI, but little is known about the corresponding attitudes among pathologists working with dermatopathology. The objective of this investigation was to make an inventory of these attitudes. Participants and Methods: An anonymous and voluntary online survey was prepared and distributed to pathologists who regularly analyzed dermatopathology slides/images. The survey consisted of 39 question divided in five sections; (1) AI as a topic in pathology; (2) previous exposure to AI as a topic in general; (3) applications for AI in dermatopathology; (4) feelings and attitudes toward AI and (5) self-reported tech-savviness and demographics. The survey opened on March 13, 2020 and closed on May 5, 2020. Results: Overall, 718 responders (64.1% females) representing 91 countries were analyzed. While 81.5% of responders were aware of AI as an emerging topic in pathology, only 18.8% had either good or excellent knowledge about AI. In terms of diagnosis classification, 42.6% saw strong or very strong potential for automated suggestion of skin tumor diagnoses. The corresponding figure for inflammatory skin diseases was 23.0% (Padj < 0.0001). For specific applications, the highest potential was considered for automated detection of mitosis (79.2%), automated suggestion of tumor margins (62.1%) and immunostaining evaluation (62.7%). The potential for automated suggestion of immunostaining (37.6%) and genetic panels (48.3%) were lower. Age did not impact the overall attitudes toward AI. Only 6.0% of the responders agreed or strongly agreed that the human pathologist will be replaced by AI in the foreseeable future. For the entire group, 72.3% agreed or strongly agreed that AI will improve dermatopathology and 84.1% thought that AI should be a part of medical training. Conclusions: Pathologists are generally optimistic about the impact and potential benefit of AI in dermatopathology. The highest potential is expected for narrow specified tasks rather than a global automated suggestion of diagnoses. There is a strong need for education about AI and its use within dermatopathology.

Investigation of psoriasis skin tissue by label-free multi-modal imaging: a case study on a phototherapy-treated patient.

Background: Psoriasis is a systemic inflammatory disease characterized by epidermal proliferation in the skin. Altered lipid metabolism is considered to be a central factor in the psoriatic etiopathogenesis. Thus, it is necessary to visualize chemical specificity of the samples for better medical diagnosis and treatment. Here, we investigate its role in the development of psoriatic lesions, before and after ultraviolet phototherapy, in a case study. Methods: The distribution and morphology of different lipids and fibrous proteins in psoriatic (lesional) tissues were visualized by two complementary label-free imaging techniques: 1) non-linear microscopy (NLM), providing images of lipids/proteins throughout the skin layers at submicrometer resolution; and 2) mass spectrometry imaging (MSI), offering high chemical specificity and hence the detection of different lipid species in the epidermal and dermal regions. A conventional method of histological evaluation was performed on the tissues, with no direct comparison with NLM and MSI. Results: Psoriatic tissues had a higher lipid content, mainly in cholesterol, in both the epidermal and dermal regions, compared to healthy tissues. Moreover, the collagen and elastin fibers in the psoriatic tissues had a tendency to assemble as larger bundles, while healthy tissues showed smaller fibers more homogeneously spread. Although phototherapy significantly reduced the cholesterol content, it also increased the amounts of collagen in both lesional and non-lesional tissues. Conclusion: This study introduces NLM and MSI as two complementary techniques which are chemical specific and can be used to assess and visualize the distribution of lipids, collagen, and elastin in a non-invasive and label-free manner.

Chemical imaging of aggressive basal cell carcinoma using time-of-flight secondary ion mass spectrometry.

A set of basal cell carcinoma samples, removed by Mohs micrographic surgery and pathologically identified as having an aggressive subtype, have been analyzed using time-of-flight secondary ion mass spectrometry (SIMS). The SIMS analysis employed a gas cluster ion beam (GCIB) to increase the sensitivity of the technique for the detection of intact lipid species. The GCIB also allowed these intact molecular signals to be maintained while surface contamination and delocalized chemicals were removed from the upper tissue surface. Distinct mass spectral signals were detected from different regions of the tissue (epidermis, dermis, hair follicles, sebaceous glands, scar tissue, and cancerous tissue) allowing mass spectral pathology to be performed. The cancerous regions of the tissue showed a particular increase in sphingomyelin signals that were detected in both positive and negative ion mode along with increased specific phosphatidylserine and phosphatidylinositol signals observed in negative ion mode. Samples containing mixed more and less aggressive tumor regions showed increased phosphatidylcholine lipid content in the less aggressive areas similar to a punch biopsy sample of a nonaggressive nodular lesion.