The role of artificial intelligence and convolutional neural networks in the management of melanoma: a clinical, pathological, and radiological perspective.

Clinical dermatoscopy and pathological slide assessment are essential in the diagnosis and management of patients with cutaneous melanoma. For those presenting with stage IIC disease and beyond, radiological investigations are often considered. The dermatoscopic, whole slide and radiological images used during clinical care are often stored digitally, enabling artificial intelligence (AI) and convolutional neural networks (CNN) to learn, analyse and contribute to the clinical decision-making. A keyword search of the Medline database was performed to assess the progression, capabilities and limitations of AI and CNN and its use in diagnosis and management of cutaneous melanoma. Full-text articles were reviewed if they related to dermatoscopy, pathological slide assessment or radiology. Through analysis of 95 studies, we demonstrate that diagnostic accuracy of AI/CNN can be superior (or at least equal) to clinicians. However, variability in image acquisition, pre-processing, segmentation, and feature extraction remains challenging. With current technological abilities, AI/CNN and clinicians synergistically working together are better than one another in all subspecialty domains relating to cutaneous melanoma. AI has the potential to enhance the diagnostic capabilities of junior dermatology trainees, primary care skin cancer clinicians and general practitioners. For experienced clinicians, AI provides a cost-efficient second opinion. From a pathological and radiological perspective, CNN has the potential to improve workflow efficiency, allowing clinicians to achieve more in a finite amount of time. Until the challenges of AI/CNN are reliably met, however, they can only remain an adjunct to clinical decision-making.

A reinforcement learning model for AI-based decision support in skin cancer.

We investigated whether human preferences hold the potential to improve diagnostic artificial intelligence (AI)-based decision support using skin cancer diagnosis as a use case. We utilized nonuniform rewards and penalties based on expert-generated tables, balancing the benefits and harms of various diagnostic errors, which were applied using reinforcement learning. Compared with supervised learning, the reinforcement learning model improved the sensitivity for melanoma from 61.4% to 79.5% (95% confidence interval (CI): 73.5-85.6%) and for basal cell carcinoma from 79.4% to 87.1% (95% CI: 80.3-93.9%). AI overconfidence was also reduced while simultaneously maintaining accuracy. Reinforcement learning increased the rate of correct diagnoses made by dermatologists by 12.0% (95% CI: 8.8-15.1%) and improved the rate of optimal management decisions from 57.4% to 65.3% (95% CI: 61.7-68.9%). We further demonstrated that the reward-adjusted reinforcement learning model and a threshold-based model outperformed naïve supervised learning in various clinical scenarios. Our findings suggest the potential for incorporating human preferences into image-based diagnostic algorithms.

Practitioner characteristics, diagnostic accuracy metrics and discovering-individual with respect to 637 melanomas documented by 27 general practitioners on the Skin Cancer Audit Research Database.

BACKGROUND AND OBJECTIVE: Knowledge of accuracy for melanoma diagnosis and melanoma discovering-individual in primary care is limited. We describe general practitioner (GP) characteristics and analyse defined diagnostic accuracy metrics for GPs in the current study comparing this with a previous study for GPs common to both, and we analyse the individual first discovering each melanoma as a lesion of concern. METHODS: The characteristics and diagnostic accuracy of 27 Australasian GPs documenting 637 melanomas on the Skin Cancer Audit Research Database (SCARD) in 2013 were described and analysed. The number needed to treat (NNT) and percentage of melanomas that were in situ (percentage in situ) were analysed as surrogates for specificity and sensitivity, respectively. The discovering-individual was analysed according to patient age and sex and lesion Breslow thickness. RESULTS: The average NNT and percentage in situ were 5.73% and 65.07%, respectively. For 21 GPs in both a 2008-2010 study and the current study, the NNT was 10.78 and 5.56, respectively (p = 0.0037). A consistent trend of decreasing NNT and increasing percentage in situ through increasingly subspecialised GP categories did not reach statistical significance. NNT trended high at ages and sites for which melanoma was rare. While the patient or family member was more likely to discover thick melanomas and melanomas in patients under 40 years, GPs discovered 73.9% of the melanomas as lesions of concern. CONCLUSIONS: GPs were the discovering-individuals for the majority of melanomas in the current study and their accuracy metrics compared favourably with published figures for dermatologists and GPs.

Dermatoscopic finding of white circles in an invasive pigmented melanoma.

Dermatoscopic white circles, seen with both non-polarized and polarized dermatoscopy, are a known clue to actinic keratosis in a flat lesion and invasive squamous cell carcinoma in a raised lesion. We have not discovered a previous published example of this clue in a melanoma. We present a case report of a 70-year-old Australian male with a pigmented superficial spreading melanoma on the face, Breslow thickness 1 mm, with dermatoscopic white circles displayed with both polarized and non-polarized dermatoscopy, and with dermatopathological correlation.

Vascular Diameter as Clue for the Diagnosis of Clinically and/or Dermoscopically Equivocal Pigmented and Non-Pigmented Basal Cell Carcinomas and Nodular Melanomas.

Background and objectives: Dermoscopy is a useful tool for the early and non-invasive diagnosis of skin malignancies. Besides many progresses, heavily pigmented and amelanotic skin tumors remain still a challenge. We aimed to investigate by dermoscopy if distinctive morphologic characteristics of vessels may help the diagnosis of equivocal nodular lesions. Materials and Methods: A collage of 16 challenging clinical and dermoscopic images of 8 amelanotic and 8 heavily pigmented nodular melanomas and basal cell carcinomas was sent via e-mail to 8 expert dermoscopists. Results: Dermoscopy improved diagnostic accuracy in 40 cases. Vessels were considered the best clue in 71 cases. Focusing on the diameter of vessels improved diagnosis in 5 cases. Conclusions: vascular diameter in addition to morphology and arrangement may be a useful dermoscopic clue for the differential diagnosis of clinically equivocal nodular malignant tumors.

The UQ Flap: A Novel Modification of the Keystone Flap.

Wound closure following excisions on the leg (between the knee and ankle), including the distal leg, is challenged by limited skin laxity. The keystone flap, first described by Behan in 2003, was proposed as one solution, but with a significant complication rate on the distal leg. This pilot study introduces a novel modification of the keystone flap, named the UQ flap, differing from other modifications, with an un-incised portion on one flap border and a unique curved leading-edge to absorb tension and distribute shearing forces in different directions, providing improved flap security and vascularization. The UQ flap was performed on 10 patients in two formats of "U" and "Q" also with two different orientations as base-proximal and base-distal. Other variations including minor deviation from the longitudinal axis, and double flap, were also performed. Except for one case with minor infection, there were no complications, and the results were favorable. No fasciotomy or undermining was required. The UQ flap proved to be a safe and convenient method of wound closure on the leg, including the distal leg. Compared with the keystone flap, there were reduced incisions leading to improved vascularity and less healthy tissue trimming. Its unique shape provided flap flexibility facilitating easy adjustment to the defects. The order and direction of wound closure after the excision of the lesion and incision of the flap are critical.

Characteristics, treatment and outcomes of 589 melanoma patients documented by 27 general practitioners on the Skin Cancer Audit Research Database.

BACKGROUND AND OBJECTIVE: General practitioners manage more melanomas than dermatologists or surgeons in Australia. Previously undescribed, the management and outcomes of melanoma patients treated by multiple Australasian general practitioners are examined. METHODS: The characteristics, management and outcomes of 589 melanoma patients, managed by 27 Australasian general practitioners and documented on the Skin Cancer Audit Research Database (SCARD), were analysed. RESULTS: Most patients (58.9%) were males with mean age at diagnosis of 62.7 years (range 18-96), and most melanomas were in situ or thin-invasive. Patients aged under 40 years had fewer melanomas, but a higher proportion (the majority) were invasive, compared with older patients (P < 0.0001). Most (55.9%) melanomas were diagnosed following elliptical excision biopsy, the rate of unintended involved margins being eightfold higher for shave biopsies. Wide re-excision was performed by the treating general practitioner for most (74.9%) melanomas, with thick melanomas preferentially referred to surgeons. The average Breslow thickness of invasive melanomas re-excised by general practitioners was 0.67 mm compared with 1.99 mm for those referred to other specialists (P < 0.0001). Of 205 patients with invasive melanoma, 14 progressed to metastatic disease, 50% of these being associated with nodular melanoma. Nine patients progressed to melanoma-specific death. The 5-year survival rate for patients with invasive melanoma was 95.2% (95% CI: 91.2-98.5%). CONCLUSIONS: Diagnostic and therapeutic management of a series of melanoma patients by Australasian general practitioners were closely aligned with current guidelines and 5-year survival with respect to invasive melanoma was at least as favourable as national population-based metrics.

Dermatoscopic and clinical features of congenital or congenital-type nail matrix nevi: A multicenter prospective cohort study by the International Dermoscopy Society.

BACKGROUND: Congenital nail matrix nevi (NMN) are difficult to diagnose because they feature clinical characteristics suggestive of adult subungual melanoma. Nail matrix biopsy is difficult to perform, especially in children. OBJECTIVE: To describe the initial clinical and dermatoscopic features of NMN appearing at birth (congenital) or after birth but before the age of 5 years (congenital-type). METHODS: We conducted a prospective, international, and consecutive data collection in 102 hospitals or private medical offices across 30 countries from 2009 to 2019. RESULTS: There were 69 congenital and 161 congenital-type NMNs. Congenital and congenital-type NMN predominantly displayed an irregular pattern of longitudinal microlines (n = 146, 64%), reminiscent of subungual melanoma in adults. The distal fibrillar ("brush-like") pattern, present in 63 patients (27.8%), was more frequently encountered in congenital NMN than in congenital-type NMN (P = .012). Moreover, congenital NMN more frequently displayed a periungual pigmentation (P = .029) and Hutchinson's sign (P = .027) than did congenital-type NMN. LIMITATIONS: Lack of systematic biopsy-proven diagnosis and heterogeneity of clinical and dermatoscopic photographs. CONCLUSION: Congenital and congenital-type NMN showed worrisome clinical and dermatoscopic features similar to those observed in adulthood subungual melanoma. The distal fibrillar ("brush-like") pattern is a suggestive feature of congenital and congenital-type NMN.

Characteristics of 637 melanomas documented by 27 general practitioners on the Skin Cancer Audit Research Database.

BACKGROUND AND OBJECTIVE: Most melanomas (including melanomas in situ), in Australasia, are treated by general practitioners (GPs). Previously undescribed, the characteristics of a series of melanomas treated by multiple GPs are examined. PATIENTS AND METHODS: Six hundred and thirty-seven melanomas treated by 27 Australasian GPs during 2013 and documented on the Skin Cancer Audit Research Database (SCARD) were analysed by anatomical site, subtype, Breslow thickness, diameter, associated naevi and linked adverse outcomes. RESULTS: Most melanomas (59.7%) were on males, mean age at diagnosis being 62.7 years (range 18-96). Most (65.0%) were in situ, with a high incidence of lentiginous melanoma (LM) (38.8%) and 32% were naevus associated. Most LM (86.4%) were in situ, compared to 55% of superficial spreading melanoma (SSM) (P < 0.0001). There was male predominance on the head, neck and trunk and female predominance on extremities. There was no significant association between Breslow thickness and diameter, with small melanomas as likely to be thick as large melanomas, and melanomas ≤3 mm diameter, on average, more likely to be invasive than larger melanomas. There was a positive correlation between age and both melanoma diameter and Breslow thickness. Seven cases progressed to melanoma-specific death: Five nodular melanoma (NM) and two SSM, one of which was thin (Breslow thickness 0.5 mm). CONCLUSIONS: A large series of melanomas treated by Australasian GPs were predominantly in situ, with a high proportion of LM subtype. With implications for GP training, NM linked to death was over-represented and there was a novel finding that older patients had larger diameter melanomas.

Dermatoscopic findings and dermatopathological correlates in clinical variants of actinic keratosis, Bowen's disease, keratoacanthoma, and squamous cell carcinoma.

Non-melanoma skin cancer (NMSC), predominantly squamous cell carcinoma (SCC) and basal cell carcinoma, is increasing worldwide. Dermatoscopy, which is one of the non-invasive diagnostic techniques, is important for early diagnosis of NMSC. In this study we aimed to determine dermatoscopic features of keratinocyte derived tumors including actinic keratosis (AK), Bowen's disease (BD), keratoacanthoma (KA), and SCC and correlate the dermatoscopic findings with pathology. A total of 242 lesions from 169 patients were included in the study and dermatoscopic and dermatopathological findings of the lesions were retrospectively studied. Revised pattern analysis was used for the dermatoscopic evaluation. Among 242 lesions, 145 were clinically flat (86 AK, 30 BD, and 29 SCC). Presence of vessels, ulceration, fiber sign, keratin mass, and blood spots decreased the probability of a lesion being AK. When the differential diagnosis was considered between KA and SCC vs AK and BD; vessel presence, ulceration, fiber sign, blood spots, white structureless, keratin, and centred vessels favored the diagnosis of KA and SCC. Our results may contribute to the determination of the lesions to be biopsied in patients with multiple AK on chronically sun damaged skin. In non-pigmented lesions when a final diagnosis cannot be established, knowledge of dermatopathologic and dermatoscopic correlation may significantly assist interpretation of dermatoscopic patterns and clues.

Skin lesions of face and scalp - Classification by a market-approved convolutional neural network in comparison with 64 dermatologists.

BACKGROUND: The clinical differentiation of face and scalp lesions (FSLs) is challenging even for trained dermatologists. Studies comparing the diagnostic performance of a convolutional neural network (CNN) with dermatologists in FSL are lacking. METHODS: A market-approved CNN (Moleanalyzer-Pro, FotoFinder Systems) was used for binary classifications of 100 dermoscopic images of FSL. The same lesions were used in a two-level reader study including 64 dermatologists (level I: dermoscopy only; level II: dermoscopy, clinical close-up images, textual information). Primary endpoints were the CNN's sensitivity and specificity in comparison with the dermatologists' management decisions in level II. Generalizability of the CNN results was tested by using four additional external data sets. RESULTS: The CNN's sensitivity, specificity and ROC AUC were 96.2% [87.0%-98.9%], 68.8% [54.7%-80.1%] and 0.929 [0.880-0.978], respectively. In level II, the dermatologists' management decisions showed a mean sensitivity of 84.2% [82.2%-86.2%] and specificity of 69.4% [66.0%-72.8%]. When fixing the CNN's specificity at the dermatologists' mean specificity (69.4%), the CNN's sensitivity (96.2% [87.0%-98.9%]) was significantly higher than that of dermatologists (84.2% [82.2%-86.2%]; p < 0.001). Dermatologists of all training levels were outperformed by the CNN (all p < 0.001). In confirmation, the CNN's accuracy (83.0%) was significantly higher than dermatologists' accuracies in level II management decisions (all p < 0.001). The CNN's performance was largely confirmed in three additional external data sets but particularly showed a reduced specificity in one Australian data set including FSL on severely sun-damaged skin. CONCLUSIONS: When applied as an assistant system, the CNN's higher sensitivity at an equivalent specificity may result in an improved early detection of face and scalp skin cancers.

Clinical and dermatoscopic characteristics of lichen planus-like keratosis in a West-Asian population.

BACKGROUND/OBJECTIVE: The clinical and dermatoscopic features of lichen planus-like keratosis have been described but the characteristics of this entity in a West-Asian population are not known. METHODS: We retrospectively analysed 82 histopathologically verified cases of lichen planus-like keratosis from 81 patients from Iran and Turkey. RESULTS: The majority of lichen planus-like keratoses were macules (61% n = 50), clinically pigmented (67.1% n = 55) and dermatoscopically multi-coloured (91.5% n = 75). The majority (63.4%) had a single dermatoscopic pattern, most frequently: structureless (35.4%), dots (14.6%) and angulated lines (8.5%). Of the lesions with more than one pattern (n = 30), the majority (n = 21) had asymmetry of pattern, the most common combinations being structureless plus dots (n = 8) and structureless plus angulated lines (n = 5). The most common structure was pigmented dots, most frequently grey and present in 70.7% of cases. Vessels were seen in 30.5% of lesions, being significantly more prevalent in non-pigmented, than pigmented, lichen planus-like keratoses (83.3% vs. 21.4% P < 0.001). When we compared lichen planus-like keratosis in the current study to that entity in a large North American study, the statistically significant differences in a West-Asian population included a greater frequency of pigmented variants, a lower incidence in females and a lower prevalence on the torso, in favour of the face. CONCLUSIONS: Lichen planus-like keratosis in a West-Asian population has clinical and dermatoscopic similarities to that entity in another studied population. The significant differences in gender association and anatomical site may be secondary to cultural factors.

Dermatoscopic features of pigmented intraepidermal carcinoma on the head and neck.

BACKGROUND AND OBJECTIVE: Pigmented intraepidermal carcinoma is characterized by dermatoscopic dots and structureless areas, including dots in linear arrangement and by coiled vessels. There are no studies describing the dermatoscopic features of pigmented intraepidermal carcinoma on the head and neck. We aim to characterize the clinical and dermatoscopic appearance of this entity. PATIENTS AND METHODS: We retrospectively analyzed 79 cases of pigmented intraepidermal carcinoma on the head and neck. RESULTS: Pigmented intraepidermal carcinoma on the head and neck was characterized dermatoscopically by multiple colors (98.7 %, n = 78), pigmented circles (48.1 %, n = 38), white circles (17.7 %, n = 14), angulated lines (41.8 %, n = 33) and structureless areas (86.1 %, n = 68). Dots in linear arrangement were present in 13.9 % (n = 11). Coiled vessels were present in 7.6 % (n = 6), the dominant vessel type being prominent serpentine vessels (29.2 %, n = 23), thicker and/or redder in color than surrounding vessels, most being in the angular arrangement of the dermal plexus (24.1 %, n = 19). CONCLUSIONS: Pigmented intraepidermal carcinoma on the head and neck differs from current published descriptions of pigmented intraepidermal carcinoma, reaching statistical significance with a lower incidence of coiled vessels and a higher incidence of pigmented circles, with evident similarities to pigmented actinic keratosis at that location.