[Skin lesion classification with multi-level fusion of Swin-T and ConvNeXt].

Skin cancer is a significant public health issue, and computer-aided diagnosis technology can effectively alleviate this burden. Accurate identification of skin lesion types is crucial when employing computer-aided diagnosis. This study proposes a multi-level attention cascaded fusion model based on Swin-T and ConvNeXt. It employed hierarchical Swin-T and ConvNeXt to extract global and local features, respectively, and introduced residual channel attention and spatial attention modules for further feature extraction. Multi-level attention mechanisms were utilized to process multi-scale global and local features. To address the problem of shallow features being lost due to their distance from the classifier, a hierarchical inverted residual fusion module was proposed to dynamically adjust the extracted feature information. Balanced sampling strategies and focal loss were employed to tackle the issue of imbalanced categories of skin lesions. Experimental testing on the ISIC2018 and ISIC2019 datasets yielded accuracy, precision, recall, and F1-Score of 96.01%, 93.67%, 92.65%, and 93.11%, respectively, and 92.79%, 91.52%, 88.90%, and 90.15%, respectively. Compared to Swin-T, the proposed method achieved an accuracy improvement of 3.60% and 1.66%, and compared to ConvNeXt, it achieved an accuracy improvement of 2.87% and 3.45%. The experiments demonstrate that the proposed method accurately classifies skin lesion images, providing a new solution for skin cancer diagnosis.

Understanding the Dermoscopic Patterns of Basal Cell Carcinoma Using Line-Field Confocal Tomography.

Basal cell carcinoma (BCC) is the most frequent malignancy in the general population. To date, dermoscopy is considered a key tool for the diagnosis of BCC; nevertheless, line-field confocal optical coherence tomography (LC-OCT), a new non-invasive optical technique, has become increasingly important in clinical practice, allowing for in vivo imaging at cellular resolution. The present study aimed to investigate the possible correlation between the dermoscopic features of BCC and their LC-OCT counterparts. In total, 100 histopathologically confirmed BCC cases were collected at the Dermatologic Clinic of the University of Siena, Italy. Predefined dermoscopic and LC-OCT criteria were retrospectively evaluated, and their frequencies were calculated. The mean (SD) age of our cohort was 65.46 (13.36) years. Overall, BCC lesions were mainly located on the head (49%), and they were predominantly dermoscopically pigmented (59%). Interestingly, all dermoscopic features considered had a statistically significant agreement with the LC-OCT criteria (all p < 0.05). In conclusion, our results showed that dermoscopic patterns may be associated with LC-OCT findings, potentially increasing accuracy in BCC diagnosis. However, further studies are needed in this field.

Diagnostic accuracy of high-frequency ultrasound for cutaneous neoplasms: a narrative review of the literature.

High-frequency ultrasound has been used to visualize depth and vascularization of cutaneous neoplasms, but little has been synthesized as a review for a robust level of evidence about the diagnostic accuracy of high-frequency ultrasound in dermatology. A narrative review of the PubMed database was performed to establish the correlation between ultrasound findings and histopathologic/dermoscopic findings for cutaneous neoplasms. Articles were divided into the following four categories: melanocytic, keratinocytic/epidermal, appendageal, and soft tissue/neural neoplasms. Review of the literature revealed that ultrasound findings and histopathology findings were strongly correlated regarding the depth of a cutaneous neoplasm. Morphological characteristics were correlated primarily in soft tissue/neural neoplasms. Overall, there is a paucity of literature on the correlation between high-frequency ultrasound and histopathology of cutaneous neoplasms. Further studies are needed to investigate this correlation in various dermatologic conditions.

Reflectance confocal microscopy versus dermoscopy for the diagnosis of cutaneous melanoma: a head-to-head comparative meta-analysis.

This meta-analysis aimed to evaluate the comparative diagnostic performance of reflectance confocal microscopy (RCM) and dermoscopy in detecting cutaneous melanoma patients. An extensive search was conducted in the PubMed and Embase databases to identify available publications up to December 2023. Studies were included if they evaluated the diagnostic performance of RCM and dermoscopy in patients with cutaneous melanoma. The quality of the included studies was assessed using the Quality Assessment of Diagnostic Performance Studies (QUADAS-2) tool. A total of 14 articles involving 2013 patients were included in the meta-analysis. The overall sensitivity of RCM was 0.94 [95% confidence interval (CI), 0.87-0.98], while the overall sensitivity of dermoscopy was 0.84 (95% CI, 0.71-0.95). These results suggested that RCM has a similar level of sensitivity compared with dermoscopy ( P  = 0.15). In contrast, the overall specificity of RCM was 0.76 (95% CI, 0.67-0.85), while the overall specificity of dermoscopy was 0.47 (95% CI, 0.31-0.63). The results indicated that RCM appears to have a higher specificity in comparison to dermoscopy ( P  < 0.01). Our meta-analysis indicates that RCM demonstrates superior specificity and similar sensitivity to dermoscopy in detecting cutaneous melanoma patients. The high heterogeneity, however, may impact the evidence of the current study, further larger sample prospective research is required to confirm these findings.

Pushing boundaries: Anterolateral thigh free flaps for extensive scalp defects beyond previous limits, leveraging imaging modalities with ultrasound and indocyanine green.

BACKGROUND: Scalp defect reconstruction poses considerable challenges, with ongoing debates regarding the most effective strategies. While the latissimus dorsi (LD) flap has traditionally been favored, the anterolateral thigh (ALT) flap has been well described as a versatile alternative for addressing extensive scalp defects. This study underscores the success of scalp reconstruction using ALT flaps, notably pushing the boundaries of previously reported flap sizes. Our approach leverages the use of indocyanine green (ICG) perfusion to guide precise preoperative planning and vascular modification, contributing to improved outcomes in challenging cases. METHODS: We performed 43 ALT flap reconstructions for scalp defects between 2016 and 2023. We collected patients' demographic and clinical data and evaluated flap size and recipient vessels and additional surgical techniques. Detailed preoperative plans with ultrasound and ICG use for intraoperative plans were performed to find perforators location. The cohort was divided into two, with or without complications on flaps, and analyzed depending on its surgical details. RESULTS: This study involved 38 patients with extensive scalp defects (mean age: 69.4 ± 11 years) who underwent ALT perforator flap transfers (mean flap size: 230.88 ± 145.6 cm2). There was only one case of unsuccessful flap transfer, and four cases had a few complications. The characteristics of the complication group included a large flap size (303.1 ± 170.9 vs. 214.9 ± 136.6 cm2, P = .211), few perforator numbers without pedicle manipulation, lack of intraoperative indocyanine green administration (75% vs. 25%, P = .607), and the use of superficial temporal vessels as recipient vessels. CONCLUSIONS: Scalp reconstruction using large ALT free flaps with the aid of imaging modalities facilitates the optimization of surgical techniques, such as pedicle manipulation, perforator numbers, and vein considerations, thereby contributing to successful reconstruction.

Prenatal recognition of cerebriform sebaceous naevus.

Cerebriform sebaceous naevus (CSN) is a rare morphological sebaceous naevus variant and challenging to diagnose prenatally due to its flat, smooth and waxy appearance and lack of association with extracutaneous manifestations.A multigravida was referred to our tertiary obstetric unit at 24 weeks of gestation for evaluation of fetal auricular lesions. We were able to further characterise the lesions via serial obstetric ultrasound imaging with the aid of three-dimensional (3D) technology. Although the precise diagnosis prenatally was uncertain, the use of 3D technology allowed the reconstruction of the fetal cutaneous lesions for multidisciplinary assessment to facilitate the development of a neonatal management plan. The diagnosis of CSN was made postnatally on biopsy.

Visualising cancer in 3D: 3-Dimensional Tissue Imaging for management of cutaneous basal cell carcinoma.

Surgical management of basal cell carcinoma (BCC) typically involves surgical excision with post-operative margin assessment using the bread-loafing technique; or gold-standard Mohs micrographic surgery (MMS), where margins are iteratively examined for residual cancer after tumour removal, with additional excisions performed upon detecting residual tumour at margins. There is limited sampling of resection margins with bread loafing, with detection of positive margins 44% of the time using 2 mm intervals. To resolve this, we have developed three-dimensional (3D) Tissue Imaging for: (1) complete examination of cancer margins and (2) detection of tumour proximity to nerves and blood vessels. 3D Tissue optical clearing with a light sheet imaging protocol was developed for margin assessment in two datasets assessed by two independent evaluators: (1) 48 samples from 29 patients with varied BCC subtypes, sizes and pigmentation levels; (2) 32 samples with matching Mohs' surgeon reading of tumour margins using two-dimensional haematoxylin & eosin-stained sections. The 3D Tissue Imaging protocol permits a complete examination of deeper and peripheral margins. Two independent evaluators achieved negative predictive values of 92.3% and 88.24% with 3D Tissue Imaging. Images obtained from 3D Tissue Imaging recapitulates histological features of BCC, such as nuclear crowding, palisading and retraction clefting and provides a 3D context for recognising normal skin adnexal structures. Concurrent immunofluorescence labelling of nerves and blood vessels allows visualisation of structures closer to tumour-positive regions, which may have a higher risk for neural and vascular infiltration. Together, this method provides more information in a 3D spatial context, enabling better cancer management by clinicians.

BCN20000: Dermoscopic Lesions in the Wild.

Advancements in dermatological artificial intelligence research require high-quality and comprehensive datasets that mirror real-world clinical scenarios. We introduce a collection of 18,946 dermoscopic images spanning from 2010 to 2016, collated at the Hospital Clínic in Barcelona, Spain. The BCN20000 dataset aims to address the problem of unconstrained classification of dermoscopic images of skin cancer, including lesions in hard-to-diagnose locations such as those found in nails and mucosa, large lesions which do not fit in the aperture of the dermoscopy device, and hypo-pigmented lesions. Our dataset covers eight key diagnostic categories in dermoscopy, providing a diverse range of lesions for artificial intelligence model training. Furthermore, a ninth out-of-distribution (OOD) class is also present on the test set, comprised of lesions which could not be distinctively classified as any of the others. By providing a comprehensive collection of varied images, BCN20000 helps bridge the gap between the training data for machine learning models and the day-to-day practice of medical practitioners. Additionally, we present a set of baseline classifiers based on state-of-the-art neural networks, which can be extended by other researchers for further experimentation.

Correlation between super-high magnification (400x) dermoscopy and reflectance confocal microscopy for the diagnosis of melanosis and other pigmented genital lesions.

The clinical diagnosis of pigmented genital lesions is challenging. Reflectance confocal microscopy (RCM) is effective for diagnosis but is limited in its application due to elevated costs. A more affordable dermatoscope with a 400x magnification (D400) has recently been brought to market. The aim of our study was to compare these two imaging techniques for the analysis of pigmented genital tumours. An observational, prospective and mono-centric study was carried out from October 2017 to May 2019, in which clinical, dermatoscopic (20x and 400x) and RCM data from 207 pigmented genital lesions were collected. The images generated via D400 and RCM were analysed by three expert investigators. Similarities between the criteria observed using D400 and RCM were evaluated by each investigator. In total, 207 lesions were included: 183 melanosis, 19 nevi, one basal cell carcinoma (BCC), two condylomas and two melanomas in situ. Our series correlates well with data found in the literature especially for the distribution of different lesions, their topography, and their aspect using x20 dermatoscopy and RCM. Pattern and cell criteria defined using RCM largely paralleled those observed with D400 for all three investigators. Correlation between D400 and RCM was moderate to strong with regards to the identification of the ring pattern and clustered round cells, strong for dendritic and plump cells, and perfect for isolated round cells and spindle cells. D400 is an easy-to-use, cost-effective alternative for the analysis of pigmented genital lesions, particularly for melanosis.

An atypical case of pediatric anaplastic lymphoma kinase-positive anaplastic large cell lymphoma limited to the skin on chest wall detected by ultrasound.

.

A novel deep learning framework for accurate melanoma diagnosis integrating imaging and genomic data for improved patient outcomes.

BACKGROUND: Melanoma is one of the most malignant forms of skin cancer, with a high mortality rate in the advanced stages. Therefore, early and accurate detection of melanoma plays an important role in improving patients' prognosis. Biopsy is the traditional method for melanoma diagnosis, but this method lacks reliability. Therefore, it is important to apply new methods to diagnose melanoma effectively. AIM: This study presents a new approach to classify melanoma using deep neural networks (DNNs) with combined multiple modal imaging and genomic data, which could potentially provide more reliable diagnosis than current medical methods for melanoma. METHOD: We built a dataset of dermoscopic images, histopathological slides and genomic profiles. We developed a custom framework composed of two widely established types of neural networks for analysing image data Convolutional Neural Networks (CNNs) and networks that can learn graph structure for analysing genomic data-Graph Neural Networks. We trained and evaluated the proposed framework on this dataset. RESULTS: The developed multi-modal DNN achieved higher accuracy than traditional medical approaches. The mean accuracy of the proposed model was 92.5% with an area under the receiver operating characteristic curve of 0.96, suggesting that the multi-modal DNN approach can detect critical morphologic and molecular features of melanoma beyond the limitations of traditional AI and traditional machine learning approaches. The combination of cutting-edge AI may allow access to a broader range of diagnostic data, which can allow dermatologists to make more accurate decisions and refine treatment strategies. However, the application of the framework will have to be validated at a larger scale and more clinical trials need to be conducted to establish whether this novel diagnostic approach will be more effective and feasible.

Ultrasound features of high-risk basal cell carcinoma: a systematic review.

Basal Cell Carcinoma (BCC) is the most prevalent skin cancer and continues to witness a surge in incidence rates. The categorization of BCC subtypes into low or high risk, guided by recurrence and invasiveness metrics, underscores the need for precise differentiation. While the punch biopsy remains the gold standard for diagnosis, its invasiveness prompts a need for non-invasive alternatives. Ultrasound (US) has emerged as a noteworthy candidate, gaining momentum in its potential to offer a less intrusive diagnostic approach. We conducted a systematic review regarding features of the high-risk subtypes of BCC on US. A thorough literature search of PubMed Medline, Embase, and CINAHL databases was conducted according to PRISMA guidelines and a total of nine studies meeting our inclusion criteria were included in this review. Evidence is still nascent but US features such as lesional shape, depth, hyperechoic spots, and color doppler may be helpful in differentiating high-risk BCC subtypes. However, further prospective studies with standardized interventions and outcome measures are required.

Empirical validation of Conformal Prediction for trustworthy skin lesions classification.

BACKGROUND AND OBJECTIVE: Uncertainty quantification is a pivotal field that contributes to realizing reliable and robust systems. It becomes instrumental in fortifying safe decisions by providing complementary information, particularly within high-risk applications. existing studies have explored various methods that often operate under specific assumptions or necessitate substantial modifications to the network architecture to effectively account for uncertainties. The objective of this paper is to study Conformal Prediction, an emerging distribution-free uncertainty quantification technique, and provide a comprehensive understanding of the advantages and limitations inherent in various methods within the medical imaging field. METHODS: In this study, we developed Conformal Prediction, Monte Carlo Dropout, and Evidential Deep Learning approaches to assess uncertainty quantification in deep neural networks. The effectiveness of these methods is evaluated using three public medical imaging datasets focused on detecting pigmented skin lesions and blood cell types. RESULTS: The experimental results demonstrate a significant enhancement in uncertainty quantification with the utilization of the Conformal Prediction method, surpassing the performance of the other two methods. Furthermore, the results present insights into the effectiveness of each uncertainty method in handling Out-of-Distribution samples from domain-shifted datasets. Our code is available at: github.com/jfayyad/ConformalDx. CONCLUSIONS: Our conclusion highlights a robust and consistent performance of conformal prediction across diverse testing conditions. This positions it as the preferred choice for decision-making in safety-critical applications.

A novel multi-task learning network for skin lesion classification based on multi-modal clues and label-level fusion.

In this paper, we propose a multi-task learning (MTL) network based on the label-level fusion of metadata and hand-crafted features by unsupervised clustering to generate new clustering labels as an optimization goal. We propose a MTL module (MTLM) that incorporates an attention mechanism to enable the model to learn more integrated, variable information. We propose a dynamic strategy to adjust the loss weights of different tasks, and trade off the contributions of multiple branches. Instead of feature-level fusion, we propose label-level fusion and combine the results of our proposed MTLM with the results of the image classification network to achieve better lesion prediction on multiple dermatological datasets. We verify the effectiveness of the proposed model by quantitative and qualitative measures. The MTL network using multi-modal clues and label-level fusion can yield the significant performance improvement for skin lesion classification.

Comprehensive analysis of clinical images contributions for melanoma classification using convolutional neural networks.

BACKGROUND: Timely diagnosis plays a critical role in determining melanoma prognosis, prompting the development of deep learning models to aid clinicians. Questions persist regarding the efficacy of clinical images alone or in conjunction with dermoscopy images for model training. This study aims to compare the classification performance for melanoma of three types of CNN models: those trained on clinical images, dermoscopy images, and a combination of paired clinical and dermoscopy images from the same lesion. MATERIALS AND METHODS: We divided 914 image pairs into training, validation, and test sets. Models were built using pre-trained Inception-ResNetV2 convolutional layers for feature extraction, followed by binary classification. Training comprised 20 models per CNN type using sets of random hyperparameters. Best models were chosen based on validation AUC-ROC. RESULTS: Significant AUC-ROC differences were found between clinical versus dermoscopy models (0.661 vs. 0.869, p < 0.001) and clinical versus clinical + dermoscopy models (0.661 vs. 0.822, p = 0.001). Significant sensitivity differences were found between clinical and dermoscopy models (0.513 vs. 0.799, p = 0.01), dermoscopy versus clinical + dermoscopy models (0.799 vs. 1.000, p = 0.02), and clinical versus clinical + dermoscopy models (0.513 vs. 1.000, p < 0.001). Significant specificity differences were found between dermoscopy versus clinical + dermoscopy models (0.800 vs. 0.288, p < 0.001) and clinical versus clinical + dermoscopy models (0.650 vs. 0.288, p < 0.001). CONCLUSION: CNN models trained on dermoscopy images outperformed those relying solely on clinical images under our study conditions. The potential advantages of incorporating paired clinical and dermoscopy images for CNN-based melanoma classification appear less clear based on our findings.

18F-FDG PET/CT findings in nevoid basal cell carcinoma syndrome: a systematic review and a new case report.

BACKGROUND: To demonstrate and analyze the 18F-FDG positron emission tomography/computed tomography (PET/CT) findings in this rare nevoid basal cell carcinoma syndrome (NBCCS). CASE PRESENTATION: A 71-year-old woman with the left invasive breast cancer was treated with hormone therapy for six months and underwent the 18F-FDG PET/CT examination for efficacy evaluation. 18F-FDG PET/CT revealed the improvement after treatment and other unexpected findings, including multiple nodules on the skin with 18F-FDG uptake, bone expansion of cystic lesions in the bilateral ribs, ectopic calcifications and dilated right ureter. She had no known family history. Then, the patient underwent surgical excision of the all skin nodules and the postoperative pathology were multiple basal cell carcinomas. Finally, the comprehensive diagnosis of NBCCS was made. The patient was still in follow-up. Additionally, we have summarized the reported cases (n = 3) with 18F-FDG PET/CT from the literature. CONCLUSIONS: It is important to recognize this syndrome on 18F-FDG PET/CT because of different diagnoses and therapeutic consequences.

Outcomes of the Transungual Approach in 56 Consecutive Digital Subungual Glomus Tumours.

Background: Glomus tumour is a painful small tumour of the glomus body commonly located under the nail bed. The aim of this study is to evaluate the correlation of clinical diagnosis with MRI findings, determine the prevalence of the tumour at different subungual locations and determine the differences in outcomes (if any) between a longitudinal and a transverse nail bed incision for excision of the tumour. Methods: This retrospective study of 56 subungual glomus tumour was conducted from May 2010 to December 2021. Data with regard to gender, age at presentation, digit involved, presenting symptoms, duration of symptoms, clinical signs, need for MRI, anatomical location, surgical approach (longitudinal versus transverse), histopathology result, period of follow-up and complications were recorded. Results: All 56 (100%) patients presented with classic triad of symptoms. The average duration of symptoms was 52.9 months (range: 3-204 months). Eleven (20%) tumours were in the sterile matrix, 38 (68%) at the junction of sterile and germinal matrix and 7 (12%) in the germinal matrix. The tumours were excised through the longitudinal incision in 31 (55.3%) patients and transverse incision in 25 (44.7%). One (1.8%) tumour was intraosseous that was diagnosed intraoperatively and excised successfully. Average follow-up was 35.4 months (range: 6-120 months). There was no difference in outcomes (pain or nail deformity) between the two incisions. One patient (1.8%) has persistent pain that was due to a missed satellite lesion in the same digit. This was excised later with resolution of symptoms. There were no recurrences and all patients were cured after excision of tumour. Conclusions: Diagnosis of glomus tumour is usually clinical, and most are located at junction of sterile and germinal matrix. Tumour can be excised either by longitudinal or transverse nail bed incisions without any change of treatment outcome. Level of Evidence: Level IV (Therapeutic).