ORCID Profile
0000-0001-9423-3435
Current Organisation
Alfred Health
Does something not look right? The information on this page has been harvested from data sources that may not be up to date. We continue to work with information providers to improve coverage and quality. To report an issue, use the Feedback Form.
Publisher: Ovid Technologies (Wolters Kluwer Health)
Date: 2018
Publisher: BMJ
Date: 02-2020
DOI: 10.1136/BMJOPEN-2019-032636
Abstract: Sentinel lymph node biopsy (SLNB) is a diagnostic procedure developed in the 1990s. It is currently used to stage patients with primary cutaneous melanoma, provide prognostic information and guide management. The Australian Clinical Practice Guidelines state that SLNB should be considered for patients with cutaneous melanoma mm in thickness (or .8 mm with high-risk pathology features). Until recently, sentinel lymph node (SLN) status was used to identify patients who might benefit from a completion lymph node dissection, a procedure that is no longer routinely recommended. SLN status is now also being used to identify patients who might benefit from systemic adjuvant therapies such as anti-programmed cell death 1 (PD1) checkpoint inhibitor immunotherapy or BRAF-directed molecular targeted therapy, treatments that have significantly improved relapse-free survival for patients with resected stage III melanoma and improved overall survival of patients with unresectable stage III and stage IV melanoma. Australian and international data indicate that approximately half of eligible patients receive an SLNB. This mixed-methods study seeks to understand the structural, contextual and cultural factors affecting implementation of the SLNB guidelines. Data collection will include: (1) cross-sectional questionnaires and semistructured interviews with general practitioners and dermatologists (2) semistructured interviews with other healthcare professionals involved in the diagnosis and early definitive care of melanoma patients and key stakeholders including researchers, representatives of professional colleges, training organisations and consumer melanoma groups and (3) documentary analysis of documents from government, health services and non-government organisations. Descriptive analyses and multivariable regression models will be used to examine factors related to SLNB practices and attitudes. Qualitative data will be analysed using thematic analysis. Ethics approval has been granted by the University of Sydney. Results will be disseminated through publications and presentations to clinicians, patients, policymakers and researchers and will inform the development of strategies for implementing SLNB guidelines in Australia.
Publisher: JMIR Publications Inc.
Date: 03-2022
DOI: 10.2196/36902
Abstract: A number of studies have shown promising performance of artificial intelligence (AI) algorithms for diagnosis of lesions in skin cancer. To date, none of these have assessed algorithm performance in the real-world setting. The aim of this project is to evaluate practical issues of implementing a convolutional neural network developed by MoleMap Ltd and Monash University eResearch in the clinical setting. Participants were recruited from the Alfred Hospital and Skin Health Institute, Melbourne, Australia, from November 1, 2019, to May 30, 2021. Any skin lesions of concern and at least two additional lesions were imaged using a proprietary dermoscopic camera. Images were uploaded directly to the study database by the research nurse via a custom interface installed on a clinic laptop. Doctors recorded their diagnosis and management plan for each lesion in real time. A pre-post study design was used. In the preintervention period, participating doctors were blinded to AI lesion assessment. An interim safety analysis for AI accuracy was then performed. In the postintervention period, the AI algorithm classified lesions as benign, malignant, or uncertain after the doctors’ initial assessment had been made. Doctors then had the opportunity to record an updated diagnosis and management plan. After discussing the AI diagnosis with the patient, a final management plan was agreed upon. Participants at both sites were high risk (for ex le, having a history of melanoma or being transplant recipients). 743 lesions were imaged in 214 participants. In total, 28 dermatology trainees and 17 consultant dermatologists provided diagnoses and management decisions, and 3 experienced teledermatologists provided remote assessments. A dedicated research nurse was essential to oversee study processes, maintain study documents, and assist with clinical workflow. In cases where AI algorithm and consultant dermatologist diagnoses were discordant, participant anxiety was an important factor in the final agreed management plan to biopsy or not. Although AI algorithms are likely to be of most use in the primary care setting, higher event rates in specialist settings are important for the initial assessment of algorithm safety and accuracy. This study highlighted the importance of considering workflow issues and doctor-patient-AI interactions prior to larger-scale trials in community-based practices. This research was supported by the Victorian Medical Research Acceleration Fund, with 1:1 contribution from MoleMap Ltd. VM is supported by the National Health and Medical Research Council Early Career Fellowship. CF is supported by the Monash University Research Training Program Scholarship. SM is head of clinical research and regulatory affairs at Kahu.ai Ltd, a subsidiary of MoleMap Ltd. MH was the chief medical officer and a director of MoleMap Ltd, and holds shares in MoleMap Ltd. ClinicalTrials.gov NCT04040114 t2/show/NCT04040114
Publisher: BMJ
Date: 2022
DOI: 10.1136/BMJOPEN-2021-050203
Abstract: Convolutional neural networks (CNNs) can diagnose skin cancers with impressive accuracy in experimental settings, however, their performance in the real-world clinical setting, including comparison to teledermatology services, has not been validated in prospective clinical studies. Participants will be recruited from dermatology clinics at the Alfred Hospital and Skin Health Institute, Melbourne. Skin lesions will be imaged using a proprietary dermoscopic camera. The artificial intelligence (AI) algorithm, a CNN developed by MoleMap Ltd and Monash eResearch, classifies lesions as benign, malignant or uncertain. This is a preintervention ostintervention study. In the preintervention period, treating doctors are blinded to AI lesion assessment. In the postintervention period, treating doctors review the AI lesion assessment in real time, and have the opportunity to then change their diagnosis and management. Any skin lesions of concern and at least two benign lesions will be selected for imaging. Each participant’s lesions will be examined by a registrar, the treating consultant dermatologist and later by a teledermatologist. At the conclusion of the preintervention period, the safety of the AI algorithm will be evaluated in a primary analysis by measuring its sensitivity, specificity and agreement with histopathology where available, or the treating consultant dermatologists’ classification. At trial completion, AI classifications will be compared with those of the teledermatologist, registrar, treating dermatologist and histopathology. The impact of the AI algorithm on diagnostic and management decisions will be evaluated by: (1) comparing the initial management decision of the registrar with their AI-assisted decision and (2) comparing the benign to malignant ratio (for lesions biopsied) between the preintervention and postintervention periods. Human Research Ethics Committee (HREC) approval received from the Alfred Hospital Ethics Committee on 14 February 2019 (HREC/48865/Alfred-2018). Findings from this study will be disseminated through peer-reviewed publications, non-peer reviewed media and conferences. NCT04040114 .
Publisher: Oxford University Press (OUP)
Date: 28-09-2023
DOI: 10.1093/BJD/LJAD365
Publisher: The Royal Australian College of General Practitioners
Date: 06-2019
Publisher: American Medical Association (AMA)
Date: 11-2019
Publisher: Wiley
Date: 15-04-2018
DOI: 10.1111/PCMR.12702
Abstract: This study aimed to determine the frequency and concordance of BRAF and NRAS mutation in tumours arising in patients with multiple primary melanoma (MPM). Patients with MPM managed at one of three tertiary referral centres in Melbourne, Australia, from 2010 to 2015 were included. Incident and subsequent melanomas underwent mutation testing. Cohen's kappa (κ) coefficient assessed agreement between incident and subsequent primary melanomas for both BRAF and NRAS mutation status (mutant versus wild-type). Mutation testing of at least two primary tumours from 64 patients was conducted. There was poor agreement for both BRAF and NRAS mutation status between incident and subsequent melanomas (κ = 0.10, 95% CI -0.10 to 0.42 κ = 0.06, 95% CI -0.10 to 0.57, respectively). In view of the low concordance in BRAF mutation status between incident and subsequent melanomas, mutational analysis of metastatic tissue, rather than of a primary melanoma, in patients with MPM should be used to guide targeted therapy.
Publisher: Oxford University Press (OUP)
Date: 29-12-2018
DOI: 10.1111/BJD.15855
Publisher: Elsevier BV
Date: 04-2020
DOI: 10.1016/J.JID.2019.07.725
Abstract: Lentigo maligna (LM) is a common subtype of in situ melanoma on chronically sun-exposed skin, particularly the head and neck of older patients. Although surgery is the standard treatment, there is associated morbidity, and options such as imiquimod cream or radiotherapy may be used if surgery is refused or inappropriate. Complete response rates following imiquimod treatment are variable in the literature. The aim of this study was to evaluate the host immune response both before and following treatment with imiquimod to better identify likely responders. Paired pre- and post-imiquimod treatment specimens were available for 27 patients. Patients were treated with imiquimod 5 days per week for 12 weeks at 16 weeks, lesions were excised for histological assessment. Of the 27 patients, 16 were responders and 11 failed to clear the disease. PDL1 protein expression was increased, accompanied by a unique gene signature in lesions from patients that subsequently histologically cleared LM by 16 weeks. This comprised 57 upregulated immune genes in signaling networks for antigen presentation, type I interferon signaling, and T-cell activation. This may represent an early responder group to imiquimod, and this unique gene signature potentially can be used as a biomarker of LM response to imiquimod.
Publisher: Wiley
Date: 17-07-2017
DOI: 10.1111/AJD.12662
Publisher: Springer Science and Business Media LLC
Date: 08-08-2017
DOI: 10.1038/BJC.2017.254
Publisher: Wiley
Date: 29-03-2011
DOI: 10.1111/J.1440-0960.2011.00747.X
Abstract: A 35-year-old woman presented with skin fragility and photosensitivity with blisters affecting her face and hands. Other symptoms included intermittent headache, fatigue, abdominal pain and nausea. Porphyrin studies were markedly raised, with features consistent with hereditary coproporphyria (HCP). Despite strict precautions, symptoms remained significantly problematic. Regular haem arginate infusions of 3 mg/kg per day over 4 days on a monthly basis were commenced and resulted in significant improvement of the patient's symptoms and a reduction in urinary porphobilinogen. Although haem arginate infusion is known as a treatment for severe acute attacks of HCP, the effectiveness of regular infusions as maintenance therapy has not been established. This is the first report of effective symptom control correlating with normalization of biochemical markers in a patient receiving regular haem arginate infusions for the treatment of HCP.
Publisher: Wiley
Date: 25-05-2023
DOI: 10.1111/AJD.14087
Publisher: Oxford University Press (OUP)
Date: 13-06-2015
DOI: 10.1111/BJD.13756
Abstract: The clinical behaviour and prognosis of primary melanomas harbouring BRAF mutations is not fully understood. To investigate the effect of mutation status on primary melanoma growth rate and melanoma-specific survival (MSS). A prospective cohort of 196 patients with stage I-III primary cutaneous melanoma were followed for a median of 92 months, pre-dating the institution of BRAF inhibitor therapy. Clinicopathological variables were correlated with mutation status and hazard ratios (HRs) estimated for MSS. Of 196 tumours, 77 (39.2%) were BRAF V600E, 10 (5.1%) BRAF V600K and 33 (16.8%) were NRAS mutant. BRAF V600E mutant melanomas were associated with favourable clinical characteristics and tended to be slower growing compared with BRAF V600K, NRAS mutant or BRAF/NRAS wild-type tumours (0.12 mm per month, 0.61 mm per month, 0.36 mm per month and 0.23 mm per month, respectively P = 0.05). There were 39 melanoma deaths, and BRAF mutant melanomas were associated with poorer MSS in stage I-III disease [HR 2.60, 95% confidence interval (CI) 1.20-5.63 P = 0.02] and stage I-II disease (HR 3.39, 95% CI 1.12-10.22 P = 0.03) after adjusting for other prognostic variables. Considered separately, BRAF V600E mutant melanomas were strongly associated with MSS independently of thickness and nodal status (HR 3.89, 95% CI 1.67-9.09 P < 0.01) but BRAF V600K mutant tumours were not (HR 1.19, 95% CI 0.36-3.92 P = 0.77). The presence of a BRAF mutation does not necessarily 'drive' more rapid tumour growth but is associated with poorer MSS in patients with early-stage disease.
Publisher: JMIR Publications Inc.
Date: 02-12-2021
Abstract: onvolutional neural networks (CNNs) are a type of artificial intelligence that show promise as a diagnostic aid for skin cancer. However, the majority are trained using retrospective image data sets of varying quality and image capture standardization. he aim of our study is to use CNN models with the same architecture, but different training image sets, and test variability in performance when classifying skin cancer images in different populations, acquired with different devices. Additionally, we wanted to assess the performance of the models against Danish teledermatologists when tested on images acquired from Denmark. hree CNNs with the same architecture were trained. CNN-NS was trained on 25,331 nonstandardized images taken from the International Skin Imaging Collaboration using different image capture devices. CNN-S was trained on 235,268 standardized images, and CNN-S2 was trained on 25,331 standardized images (matched for number and classes of training images to CNN-NS). Both standardized data sets (CNN-S and CNN-S2) were provided by Molemap using the same image capture device. A total of 495 Danish patients with 569 images of skin lesions predominantly involving Fitzpatrick skin types II and III were used to test the performance of the models. Four teledermatologists independently diagnosed and assessed the images taken of the lesions. Primary outcome measures were sensitivity, specificity, and area under the curve of the receiver operating characteristic (AUROC). total of 569 images were taken from 495 patients (n=280, 57% women, n=215, 43% men mean age 55, SD 17 years) for this study. On these images, CNN-S achieved an AUROC of 0.861 (95% CI 0.830-0.889 i P /i & .001), and CNN-S2 achieved an AUROC of 0.831 (95% CI 0.798-0.861 i P /i =.009), with both outperforming CNN-NS, which achieved an AUROC of 0.759 (95% CI 0.722-0.794 i P /i & .001 i P /i =.009). When the CNNs were matched to the mean sensitivity and specificity of the teledermatologists, the model’s resultant sensitivities and specificities were surpassed by the teledermatologists. However, when compared to CNN-S, the differences were not statistically significant ( i P /i =.10 i P /i =.05). Performance across all CNN models and teledermatologists was influenced by the image quality. NNs trained on standardized images had improved performance and therefore greater generalizability in skin cancer classification when applied to an unseen data set. This is an important consideration for future algorithm development, regulation, and approval. Further, when tested on these unseen test images, the teledermatologists i clinically /i outperformed all the CNN models however, the difference was deemed to be statistically insignificant when compared to CNN-S.
Publisher: Elsevier BV
Date: 05-2017
Publisher: AMPCo
Date: 22-07-2019
DOI: 10.5694/MJA2.50289
Abstract: To assess changes in the choice of skin biopsy technique for assessing invasive melanoma in Victoria, and to examine the impact of partial biopsy technique on the accuracy of tumour microstaging. Retrospective cross-sectional review of Victorian Cancer Registry data on invasive melanoma histologically diagnosed in Victoria during 2005, 2010, and 2015. 400 patients randomly selected from each of the three years, stratified by final tumour thickness: 200 patients with thin melanoma ( 4.0 mm). Partial and excisional biopsies, as proportions of all skin biopsies rates of tumour base transection and T-upstaging, and mean tumour thickness underestimation following partial biopsy. 833 excisional and 337 partial diagnostic biopsies were undertaken. The proportion of partial biopsies increased from 20% of patients in 2005 to 36% in 2015 (P < 0.001) the proportion of shave biopsies increased from 9% in 2005 to 20% in 2015 (P < 0.001), with increasing rates among dermatologists and general practitioners. Ninety-four of 175 shave biopsies (54%) transected the tumour base wide local excision subsequently identified residual melanoma in 65 of these cases (69%). Twenty-one tumours diagnosed by shave biopsy (12%) were T-upstaged. With base-transected shave biopsies, tumour thickness was underestimated by a mean 2.36 mm for thick, 0.48 mm for intermediate, and 0.07 mm for thin melanomas. Partial biopsy, particularly shave biopsy, was increasingly used for diagnosing invasive melanoma between 2005 and 2015. Shave biopsy has a high rate of base transection, reducing the accuracy of tumour staging, which is crucial for planning appropriate therapy, including definitive surgery and adjuvant therapy.
Publisher: Elsevier BV
Date: 04-2020
Publisher: Wiley
Date: 03-2011
DOI: 10.1111/J.1440-0960.2010.00727.X
Abstract: As melanoma incidence in Australia continues to rise, targeting high-risk in iduals for early detection is of paramount importance. We aimed to design a population-specific risk assessment tool to improve on the use of intuition alone for assignment of surveillance strategies for high-risk in iduals and help communicate risk more accurately and effectively to patients. Methods used in the development of breast cancer risk models were adopted. Data from a large meta-analysis was used to determine risk estimates. Attributable risk was calculated for each risk factor using data from the Victorian Melanoma Service. Local prevalence data from state cancer registries was incorporated to estimate 5-year risk of melanoma. Independent risk factors identified were common naevi, atypical naevi, hair colour, freckles, family history of melanoma and personal history of non-melanoma skin cancer. Personal history of melanoma was the strongest risk factor for developing another (relative risk 7.28, 7.24). Absolute risk for in iduals varies greatly with age, risk factor profiles and proximity to the equator. We have developed a melanoma risk assessment tool based on the best available information (alculator). The tool is easily modified as new information becomes available.
Publisher: Wiley
Date: 15-11-2023
DOI: 10.1111/AJD.13946
Abstract: Artificial Intelligence (AI) is the ability for computers to simulate human intelligence. In dermatology, there is substantial interest in using AI to identify skin lesions from images. Due to increasing research and interest in the use of AI, the Australasian College of Dermatologists has developed a position statement to inform its members of appropriate use of AI. This article presents the ACD Position Statement on the use of AI in dermatology, and provides explanatory information that was used to inform the development of this statement.
Publisher: AMPCo
Date: 07-2015
DOI: 10.5694/MJA15.00148
Publisher: Wiley
Date: 23-12-2020
DOI: 10.1111/AJD.13227
Abstract: Partial biopsies are sometimes used for melanoma diagnosis with anticipated time and cost savings compared to excisional biopsy. However, their impact on subsequent melanoma management is unknown. Determine factors related to choice of partial over excisional biopsy to diagnose invasive melanoma and examine the effect of partial biopsies on definitive melanoma management. Retrospective repeated cross-sectional population-based study through the Victorian Cancer Registry of diagnosed melanomas in 2005, 2010 and 2015. A random s le of 400 patients per year, stratified by tumour thickness, was selected. A total of 1200 patients had 833 excisional and 337 partial biopsies. Omission of suspected diagnosis on pathology requests affected 46% (532/1151) of all diagnostic biopsies. Diagnostic suspicion did not influence preference for partial over excisional biopsy [Odds Ratio (OR) 1.2, 95%CI 0.8-1.7 P = 0.40]. The partial:excisional biopsy usage ratio was higher in patients aged > 50 years than patients aged <50 years [relative risk ratios (RRR) 1.5 95%CI 1.0 to 2.2 P = 0.03]. In 34% and 17% of tumours diagnosed with punch and shave, respectively, three procedures were required for definitive excision instead of two, compared with 5% of excisional biopsies When partial biopsy was used, patients were at greater risk of requiring three-staged excisions when controlled for age, anatomical site, melanoma subtype and thickness (RRR 6.7 95%CI 4.4-10.1 P < 0.001). Diagnostic suspicion does not appear to be a major factor influencing choice of biopsy technique. Using partial biopsy to diagnose melanoma often leads to an extra procedure for definitive treatment compared with excisional biopsy.
Publisher: American Society of Clinical Oncology (ASCO)
Date: 03-2019
DOI: 10.1200/JOP.18.00570
Publisher: Wiley
Date: 17-01-2016
DOI: 10.1111/PCMR.12450
Abstract: BRAF mutations at codons L597 and K601 occur uncommonly in melanoma. Clinical and pathological associations of these mutations were investigated in a cohort of 1119 patients with known BRAF mutation status. A BRAF mutation was identified in 435 patients Mutations at L597 and the K601E mutation were seen in 3.4 and 3.2% of these, respectively. K601E melanomas tended to occur in male patients, a median age of 58 yr, were generally found on the trunk (64%) and uncommonly associated with chronically sun-damaged (CSD) skin. BRAF L597 melanomas occurred in older patients (median 66 yr), but were associated with CSD skin (extremities or head and neck location - 73.3%, P = 0.001). Twenty-three percent of patients with V600E- and 43% of patients with K601E-mutant melanomas presented with nodal disease at diagnosis compared to just 14% of patients with BRAF wild-type tumors (P = 0.001 and 0.006, respectively). Overall, these mutations represent a significant minority of BRAF mutations, but have distinct clinicopathological phenotypes and clinical behaviors.
Publisher: Wiley
Date: 04-08-2014
DOI: 10.1111/PCMR.12295
Abstract: Activating mutations in the GTPase RAC1 are a recurrent event in cutaneous melanoma. We investigated the clinical and pathological associations of RAC1(P29S) in a cohort of 814 primary cutaneous melanomas with known BRAF and NRAS mutation status. The RAC1(P29S) mutation had a prevalence of 3.3% and was associated with increased thickness (OR=1.6 P = 0.001), increased mitotic rate (OR=1.3 P = 0.03), ulceration (OR=2.4 P = 0.04), nodular subtype (OR=3.4 P = 0.004), and nodal disease at diagnosis (OR=3.3 P = 0.006). BRAF mutant tumors were also associated with nodal metastases (OR=1.9 P = 0.004), despite being thinner at diagnosis than BRAF WT (median 1.2 mm versus 1.6 mm, P < 0.001). Immunohistochemical analysis of 51 melanomas revealed that 47% were immunoreactive for RAC1. Melanomas were more likely to show RAC1 immunoreactivity if they were BRAF mutant (OR=5.2 P = 0.01). RAC1 may therefore be important in regulating the early migration of BRAF mutant tumors. RAC1 mutations are infrequent in primary melanomas but may accelerate disease progression.
Publisher: AMPCo
Date: 10-2017
DOI: 10.5694/MJA17.00123
Abstract: A Cancer Council Australia multidisciplinary working group is currently revising and updating the 2008 evidence-based clinical practice guidelines for the management of cutaneous melanoma. While there have been many recent improvements in treatment options for metastatic melanoma, early diagnosis remains critical to reducing mortality from the disease. Improved awareness of the atypical presentations of this common malignancy is required to achieve this. A chapter of the new guidelines was therefore developed to aid recognition of atypical melanomas. Main recommendations: Because thick, life-threatening melanomas may lack the more classical ABCD (asymmetry, border irregularity, colour variegation, diameter > 6 mm) features of melanoma, a thorough history of the lesion with regard to change in morphology and growth over time is essential. Any lesion that is changing in morphology or growing over a period of more than one month should be excised or referred for prompt expert opinion. Changes in management as a result of the guidelines: These guidelines provide greater emphasis on improved recognition of the atypical presentations of melanoma, in particular nodular, desmoplastic and acral lentiginous subtypes, with particular awareness of hypomelanotic and amelanotic lesions.
Publisher: Springer Science and Business Media LLC
Date: 05-2018
Publisher: AMPCo
Date: 12-02-2020
DOI: 10.5694/MJA2.50502
Publisher: JMIR Publications Inc.
Date: 12-09-2022
DOI: 10.2196/35150
Abstract: Convolutional neural networks (CNNs) are a type of artificial intelligence that shows promise as a diagnostic aid for skin cancer. However, the majority are trained using retrospective image data sets with varying image capture standardization. The aim of our study was to use CNN models with the same architecture—trained on image sets acquired with either the same image capture device and technique (standardized) or with varied devices and capture techniques (nonstandardized)—and test variability in performance when classifying skin cancer images in different populations. In all, 3 CNNs with the same architecture were trained. CNN nonstandardized (CNN-NS) was trained on 25,331 images taken from the International Skin Imaging Collaboration (ISIC) using different image capture devices. CNN standardized (CNN-S) was trained on 177,475 MoleMap images taken with the same capture device, and CNN standardized number 2 (CNN-S2) was trained on a subset of 25,331 standardized MoleMap images (matched for number and classes of training images to CNN-NS). These 3 models were then tested on 3 external test sets: 569 Danish images, the publicly available ISIC 2020 data set consisting of 33,126 images, and The University of Queensland (UQ) data set of 422 images. Primary outcome measures were sensitivity, specificity, and area under the receiver operating characteristic curve (AUROC). Teledermatology assessments available for the Danish data set were used to determine model performance compared to teledermatologists. When tested on the 569 Danish images, CNN-S achieved an AUROC of 0.861 (95% CI 0.830-0.889) and CNN-S2 achieved an AUROC of 0.831 (95% CI 0.798-0.861 standardized models), with both outperforming CNN-NS (nonstandardized model P=.001 and P=.009, respectively), which achieved an AUROC of 0.759 (95% CI 0.722-0.794). When tested on 2 additional data sets (ISIC 2020 and UQ), CNN-S (P .001 and P .001, respectively) and CNN-S2 (P=.08 and P=.35, respectively) still outperformed CNN-NS. When the CNNs were matched to the mean sensitivity and specificity of the teledermatologists on the Danish data set, the models’ resultant sensitivities and specificities were surpassed by the teledermatologists. However, when compared to CNN-S, the differences were not statistically significant (sensitivity: P=.10 specificity: P=.053). Performance across all CNN models as well as teledermatologists was influenced by image quality. CNNs trained on standardized images had improved performance and, therefore, greater generalizability in skin cancer classification when applied to unseen data sets. This finding is an important consideration for future algorithm development, regulation, and approval.
Publisher: JMIR Publications Inc.
Date: 10-12-2021
DOI: 10.2196/35391
Abstract: Convolutional neural networks (CNNs) are a type of artificial intelligence that show promise as a diagnostic aid for skin cancer. However, the majority are trained using retrospective image data sets of varying quality and image capture standardization. The aim of our study is to use CNN models with the same architecture, but different training image sets, and test variability in performance when classifying skin cancer images in different populations, acquired with different devices. Additionally, we wanted to assess the performance of the models against Danish teledermatologists when tested on images acquired from Denmark. Three CNNs with the same architecture were trained. CNN-NS was trained on 25,331 nonstandardized images taken from the International Skin Imaging Collaboration using different image capture devices. CNN-S was trained on 235,268 standardized images, and CNN-S2 was trained on 25,331 standardized images (matched for number and classes of training images to CNN-NS). Both standardized data sets (CNN-S and CNN-S2) were provided by Molemap using the same image capture device. A total of 495 Danish patients with 569 images of skin lesions predominantly involving Fitzpatrick skin types II and III were used to test the performance of the models. Four teledermatologists independently diagnosed and assessed the images taken of the lesions. Primary outcome measures were sensitivity, specificity, and area under the curve of the receiver operating characteristic (AUROC). A total of 569 images were taken from 495 patients (n=280, 57% women, n=215, 43% men mean age 55, SD 17 years) for this study. On these images, CNN-S achieved an AUROC of 0.861 (95% CI 0.830-0.889 P .001), and CNN-S2 achieved an AUROC of 0.831 (95% CI 0.798-0.861 P=.009), with both outperforming CNN-NS, which achieved an AUROC of 0.759 (95% CI 0.722-0.794 P .001 P=.009). When the CNNs were matched to the mean sensitivity and specificity of the teledermatologists, the model’s resultant sensitivities and specificities were surpassed by the teledermatologists. However, when compared to CNN-S, the differences were not statistically significant (P=.10 P=.05). Performance across all CNN models and teledermatologists was influenced by the image quality. CNNs trained on standardized images had improved performance and therefore greater generalizability in skin cancer classification when applied to an unseen data set. This is an important consideration for future algorithm development, regulation, and approval. Further, when tested on these unseen test images, the teledermatologists clinically outperformed all the CNN models however, the difference was deemed to be statistically insignificant when compared to CNN-S. VM received speakers fees from Merck, Eli Lily, Novartis and Bristol Myers Squibb. VM is the principal investigator for a clinical trial funded by the Victorian Department of Health and Human Services with 1:1 contribution from MoleMap.
Publisher: Wiley
Date: 28-11-2014
DOI: 10.1111/AJD.12121
Abstract: Although there has been improvement in clinical diagnosis of pigmented superficial spreading melanomas (SSM), less common melanoma subtypes have different clinical features and may be more difficult to diagnose. The objective was to assess diagnostic accuracy for different melanoma subtypes. A retrospective review was made of a random selection of SSM, nodular melanomas (NM), desmoplastic melanomas (DM) and acral lentiginous melanomas (ALM) biopsied between February 2001 and May 2012 and referred to the Victorian Melanoma Service. Clinical differential diagnoses listed on pre-operative biopsy pathology request forms were recorded. Sensitivity for the diagnosis of melanoma was used as a marker of diagnostic accuracy. In total 111 SSM, 121 NM, 43 DM and 30 ALM were biopsied by 222 clinicians. Whereas diagnostic sensitivity for SSM and ALM were similar (77%, 95% CI 69-85% and 73%, 95% CI 58-89%, respectively) diagnostic sensitivity was lower for NM (41%, 95% CI 33-50%) and DM (21%, 95% CI 9-33%). Both NM and DM were diagnosed at greater tumour thickness (median 3.0 mm and 4.0 mm) than SSM and ALM (both median 1.0 mm). Amelanosis was associated with lower diagnostic sensitivity for SSM (0 vs 82%, P < 0.01), NM (19 vs 51%, P < 0.01) andDM (10 vs 32%, P = 0.07). Dermatologists were more accurate than non-dermatologists for NM (diagnostic sensitivity 57 vs 32%, P < 0.01) and ALM (diagnostic sensitivity 94 vs 43%, P = 0.02). Misdiagnosis of melanoma varies according to subtype and is particularly problematic for NM, DM and hypopigmented melanomas. Greater awareness of the different criteria required to diagnose these melanomas is needed.
Publisher: Wiley
Date: 30-11-2017
DOI: 10.1111/PCMR.12544
Abstract: Metastasis represents the end product of an elaborate biological process, which is determined by a complex interplay between metastatic tumour cells, host factors and homoeostatic mechanisms. Cutaneous melanoma can metastasize haematogenously or lymphogenously. The three predominant models that endeavour to explain the patterns of melanoma progression are the stepwise spread model, the simultaneous spread model and the model of differential spread. The time course to the development of metastases differs between the different metastatic routes. There are several clinical and histopathological risk factors for the different metastatic pathways. In particular, patient sex and the anatomical location of the primary tumour influence patterns of disease progression. There is limited existing evidence regarding the relationship between tumour mutation status, other diagnostic and prognostic biomarkers and the metastatic pathways of primary cutaneous melanoma. This knowledge gap needs to be addressed to better identify patients at high risk of disease recurrence and personalize surveillance strategies.
Publisher: Elsevier BV
Date: 03-2017
DOI: 10.1016/J.JAAD.2016.08.009
Abstract: Scalp melanomas have more aggressive clinicopathological features than other melanomas and mortality rates more than twice that of melanoma located elsewhere. We sought to describe the survival of patients with scalp melanoma versus other cutaneous head and neck melanoma (CHNM), and explore a possible independent negative impact of scalp location on CHNM survival. A retrospective cohort study was performed of all invasive primary CHNM cases seen at a tertiary referral center over a 20-year period. Melanoma-specific survival (MSS) was compared between scalp melanoma and other invasive CHNM. Multivariable Cox proportional hazards regression was performed to determine associations with survival. On univariate analysis, patients with scalp melanoma had worse MSS than other CHNM (hazard ratio 2.22, 95% confidence interval 1.59-3.11). Scalp location was not associated with MSS in CHNM on multivariable analysis (hazard ratio 1.11, 95% confidence interval 0.77-1.61) for all tumors together, but remained independently associated with MSS for the 0.76- to 1.50-mm thickness stratum (hazard ratio 5.51, 95% confidence interval 1.55-19.59). Disease recurrence was not assessed because of unavailable data. The poorer survival of scalp melanoma is largely explained by greater Breslow thickness and a higher proportion of male patients.
Publisher: Wiley
Date: 18-08-2022
DOI: 10.1111/AJD.13690
Abstract: Artificial intelligence (AI) technology is becoming increasingly accurate and prevalent for the diagnosis of skin cancers. Commercially available AI diagnostic software is entering markets across the world posing new legal and ethical challenges for both clinicians and software companies. Australia has the highest rates of skin cancer in the world and is poised to be a significant benefactor and pioneer of the technology. This review describes the legal and ethical considerations raised by the emergence of artificial intelligence in skin cancer diagnosis and proposes recommendations for best practice.
Publisher: JMIR Publications Inc.
Date: 28-01-2022
Abstract: number of studies have shown promising performance of artificial intelligence (AI) algorithms for diagnosis of lesions in skin cancer. To date, none of these have assessed algorithm performance in the real-world setting. he aim of this project is to evaluate practical issues of implementing a convolutional neural network developed by MoleMap Ltd and Monash University eResearch in the clinical setting. articipants were recruited from the Alfred Hospital and Skin Health Institute, Melbourne, Australia, from November 1, 2019, to May 30, 2021. Any skin lesions of concern and at least two additional lesions were imaged using a proprietary dermoscopic camera. Images were uploaded directly to the study database by the research nurse via a custom interface installed on a clinic laptop. Doctors recorded their diagnosis and management plan for each lesion in real time. A pre-post study design was used. In the preintervention period, participating doctors were blinded to AI lesion assessment. An interim safety analysis for AI accuracy was then performed. In the postintervention period, the AI algorithm classified lesions as benign, malignant, or uncertain after the doctors’ initial assessment had been made. Doctors then had the opportunity to record an updated diagnosis and management plan. After discussing the AI diagnosis with the patient, a final management plan was agreed upon. articipants at both sites were high risk (for ex le, having a history of melanoma or being transplant recipients). 743 lesions were imaged in 214 participants. In total, 28 dermatology trainees and 17 consultant dermatologists provided diagnoses and management decisions, and 3 experienced teledermatologists provided remote assessments. A dedicated research nurse was essential to oversee study processes, maintain study documents, and assist with clinical workflow. In cases where AI algorithm and consultant dermatologist diagnoses were discordant, participant anxiety was an important factor in the final agreed management plan to biopsy or not. lthough AI algorithms are likely to be of most use in the primary care setting, higher event rates in specialist settings are important for the initial assessment of algorithm safety and accuracy. This study highlighted the importance of considering workflow issues and doctor-patient-AI interactions prior to larger-scale trials in community-based practices. linicalTrials.gov NCT04040114 t2/show/NCT04040114
Publisher: Wiley
Date: 30-03-2020
Publisher: Wiley
Date: 28-02-2020
DOI: 10.1002/IJC.32930
Publisher: Elsevier BV
Date: 10-2014
DOI: 10.1016/J.JAAD.2014.04.068
Abstract: To calculate melanoma rate of growth (ROG), previous studies have relied on subjective patient recall to estimate time delay to diagnosis. To objectively calculate ROG by measuring the rate of increase in melanoma thickness between 2 sequential biopsy specimens over time. This was a retrospective review of 51 melanomas in which pathologic misdiagnosis of a partial biopsy specimen caused a delay before referral and excisional biopsy between January 1998 and January 2013. ROG was calculated as rate of increase in tumor thickness between biopsy specimens. The median delay between the 2 biopsy specimens was 27 months (range, 3-89 months). Biopsy specimens of melanomas that were obtained initially in their in situ phase were thinner at excision compared to those that were first obtained as invasive tumors (median, 0.7 vs. 3.2 mm P<.01) and had a lower ROG (median, 0.04 vs. 0.11 mm/month P=.05). Faster growth was associated with increased tumor thickness, higher mitotic rate, symptoms, elevation, and amelanosis. Partial biopsy specimens may not be representative of deepest tumor thickness. We have demonstrated an objective measure of melanoma growth rate using sequential biopsy specimens. The correlation between faster growth and aggressive tumor features supports what others have found and validates the historical measure of growth rate as a reliable clinical marker.
Publisher: Elsevier BV
Date: 08-2018
Publisher: Elsevier BV
Date: 12-2019
Publisher: American Association for Cancer Research (AACR)
Date: 09-2013
DOI: 10.1158/1078-0432.CCR-13-0398
Abstract: Purpose: The mutation load in melanoma is generally high compared with other tumor types due to extensive UV damage. Translation of exome sequencing data into clinically relevant information is therefore challenging. This study sought to characterize mutations identified in primary cutaneous melanomas and correlate these with clinicopathologic features. Experimental Design: DNA was extracted from 34 fresh-frozen primary cutaneous melanomas and matched peripheral blood. Tumor histopathology was reviewed by two dermatopathologists. Exome sequencing was conducted and mutation rates were correlated with age, sex, tumor site, and histopathologic variables. Differences in mutations between categories of solar elastosis, pigmentation, and BRAF/NRAS mutational status were investigated. Results: The average mutation rate was 12 per megabase, similar to published results in metastases. The average mutation rate in severely sun damaged (SSD) skin was 21 per Mb compared with 3.8 per Mb in non-SSD skin (P = 0.001). BRAF/NRAS wild-type (WT) tumors had a higher average mutation rate compared with BRAF/NRAS–mutant tumors (27 vs. 5.6 mutations per Mb P = 0.0001). Tandem CC& TT/GG& AA mutations comprised 70% of all dinucleotide substitutions and were more common in tumors arising in SSD skin (P = 0.0008) and in BRAF/NRAS WT tumors (P = 0.0007). Targetable and potentially targetable mutations in WT tumors, including NF1, KIT, and NOTCH1, were spread over various signaling pathways. Conclusion: Melanomas arising in SSD skin have higher mutation loads and contain a spectrum of molecular subtypes compared with BRAF- and NRAS-mutant tumors indicating multigene screening approaches and combination therapies may be required for management of these patients. Clin Cancer Res 19(17) 4589–98. ©2013 AACR.
Publisher: The Royal Australian College of General Practitioners
Date: 06-2020
Publisher: Impact Journals, LLC
Date: 06-12-2015
Abstract: Melanoma is often caused by mutations due to exposure to ultraviolet radiation. This study reports a recurrent somatic C > T change causing a P131L mutation in the RQCD1 (Required for Cell Differentiation1 Homolog) gene identified through whole exome sequencing of 20 metastatic melanomas. Screening in 715 additional primary melanomas revealed a prevalence of ~4%. This represents the first reported recurrent mutation in a member of the CCR4-NOT complex in cancer. Compared to tumors without the mutation, the P131L mutant positive tumors were associated with increased thickness (p = 0.02), head and neck (p = 0.009) and upper limb (p = 0.03) location, lentigo maligna melanoma subtype (p = 0.02) and BRAF V600K (p = 0.04) but not V600E or NRAS codon 61 mutations. There was no association with nodal disease (p = 0.3). Mutually exclusive mutations of other members of the CCR4-NOT complex were found in ~20% of the TCGA melanoma dataset suggesting the complex may play an important role in melanoma biology. Mutant RQCD1 was predicted to bind strongly to HLA-A0201 and HLA-Cw3 MHC1 complexes. From thirteen patients with mutant RQCD1, an anti-tumor CD8⁺ T cell response was observed from a single patient's peripheral blood mononuclear cell population stimulated with mutated peptide compared to wildtype indicating a neoantigen may be formed.
Publisher: Elsevier BV
Date: 05-2019
Publisher: Elsevier BV
Date: 04-2013
DOI: 10.1016/J.JAAD.2012.09.047
Abstract: There is a growing body of evidence that nodular melanoma (NM), because of its association with increased growth rate and thickness at diagnosis, accounts for a substantial proportion of melanoma deaths. We sought to assess the contribution of NM to melanoma deaths in comparison with other tumor subtypes. Four cohorts were established comprising 5775 cases of invasive primary cutaneous melanoma reported to the Victorian Cancer Registry during 1989, 1994, 1999, and 2004. Original pathology reports were reviewed. Age-standardized melanoma incidence rates were compared from 1989 to 2004 with annual percentage change using Poisson regression. The incidence of thick tumors (>4 mm) increased by 3.8% (95% confidence interval 1.4 to 6.2) and 2.5% (95% confidence interval -0.5 to 5.5) per year for male and female patients, respectively. The median thickness of NM at diagnosis was 2.6 mm compared with 0.6 mm for superficial spreading melanoma. A third of patients who died from melanoma during the follow-up period had thick tumors (>4 mm), most of which were nodular subtype (61%). NM accounted for 14% of invasive melanomas, but was responsible for 43% of melanoma deaths in a total of 57,461 person-years of follow-up. By comparison, superficial spreading melanoma contributed 56% of invasive melanoma but only 30% of deaths. Pathology review was limited to reports only. Mortality information relied mostly on death certificate information. The incidence of thick melanomas continues to increase. Nodular melanoma is clinically distinct and the predominant contributor to melanoma-related deaths, representing a public health challenge in reducing skin cancer mortality.
Publisher: Wiley
Date: 14-01-2017
DOI: 10.1111/AJD.12437
Abstract: Scalp melanoma has a worse prognosis than melanoma elsewhere, though the reasons for this are poorly understood. Current literature describing the clinicopathological associations of scalp melanoma is sparse. This study aims to compare clinical and histological features of scalp melanoma with other cutaneous head and neck melanomas (CHNM). A cross-sectional study was performed of all primary CHNM cases seen by the Victorian Melanoma Service between 1994 and 2014, using prospectively recorded clinical data. Invasive and in situ melanomas were compared separately. Invasive scalp melanoma was associated with male sex (OR, 2.7 95% CI, 1.9-3.9), increasing age (OR, 1.02 per year increase in age 95% CI, 1.01-1.03), being first noticed by a person other than self, spouse/relative or doctor (OR, 2.9 95% CI, 1.5-5.7), amelanosis (OR, 1.6 95% CI, 1.1-2.3), and increased growth rate (OR, 1.14 per 1 mm/month growth rate increase 95% CI, 1.04-1.26). Compared with other CHNM, scalp melanoma had greater median Breslow thickness (2.8 vs 1.2 mm) and was independently associated with satellite metastases (OR, 4.7 95% CI, 1.9-11.5) and nodular subtype (OR, 1.8 95% CI, 1.1-3.1). In situ scalp melanoma was associated with male sex, increasing age and solar keratoses. Scalp melanoma tends to occur in older men, is often rapidly growing and amelanotic, and is associated with high risk histological features. As it is likely to be overlooked, increased recognition of the atypical presentations of scalp melanoma is required.
Publisher: Wiley
Date: 29-04-2022
DOI: 10.1111/AJD.13848
Abstract: Clinical quality registries aim to identify significant variations in care and provide anonymised feedback to institutions to improve patient outcomes. Thirty‐six Australian organisations with an interest in melanoma, raised funds through three consecutive Melanoma Marches, organised by Melanoma Institute Australia, to create a national Melanoma Clinical Outcomes Registry (MelCOR). This study aimed to formally develop valid clinical quality indicators for the diagnosis and early management of cutaneous melanoma as an important step in creating the registry. Potential clinical quality indicators were identified by examining the literature, including Australian and international melanoma guidelines, and by consulting with key melanoma and registry opinion leaders. A modified two‐round Delphi survey method was used, with participants invited from relevant health professions routinely managing melanoma as well as relevant consumer organisations. Nineteen participants completed at least one round of the Delphi process. 12 of 13 proposed clinical quality indictors met the validity criteria. The clinical quality indicators included acceptable biopsy method, appropriate excision margins, standardised pathology reporting, indications for sentinel lymph node biopsy, and involvement of multidisciplinary care and referrals. This study provides a multi‐stakeholder consensus for important clinical quality indicators that define optimal practice that will now be used in the Australian Melanoma Clinical Outcomes Registry (MelCOR).
Publisher: Oxford University Press (OUP)
Date: 20-12-2018
DOI: 10.1111/CED.13343
Start Date: 2011
End Date: 2014
Funder: National Health and Medical Research Council
View Funded ActivityStart Date: 2017
End Date: 2019
Funder: National Health and Medical Research Council
View Funded Activity