Assessment of the change in accuracy of an artificial intelligence algorithm for the detection of skin cancer in camera images following diversification and training

Abstract Background The US FDA recently stated in its Proposed Regulatory Framework for software as a medical device (SaMD) that “One of the greatest benefits of AI/ML in software resides in its ability to learn from real-world use and experience, and its capability to improve its performance.” This study follows two previous publications which addressed the accuracy of a machine learning algorithm for the detection of malignant melanoma. The aim of this study was to quantify the change in the accuracy following modifications to the algorithm (DERM) for the detection of non-melanoma skin cancers and potential precursors of skin cancer. A secondary aim was to assess any improvement in accuracy associated with continued training of the algorithm.Methods A total of 16,550 images of skin lesions with histopathology based assessment were available for assessment. The primary indicator of diagnostic accuracy was the area under the ROC curve with 95% confidence intervals. Sensitivity and specificity at the most efficient cut-point was also estimated together with the numbers of false negative and false positive results.Results The inclusion of squamous cell cancer, basal cell cancer and intra-epidermal carcinoma in addition to melanoma results in an improvement in the scope of the algorithm. For the most recent version of the algorithm all skin cancers show an area under the ROC curve greater than 95%. For melanoma sensitivity=91% and specificity = 89%; for all non-melanoma skin cancers sensitivity=97% and specificity=94%. Continued training of the algorithm results in a statistically significant (p<0.01) improvement in accuracy which diminishes as the ROC area approaches 100%. Conclusions The results indicate that as the algorithm is used in clinical practice it will become more accurate with continued training but the rate of improvement will diminish as the ROC area approaches 100%. A smartphone or other camera fitted with a dermoscopic lens and with internet access to the algorithm can provide an accurate additional assessment of a suspected skin cancer lesion or precursor for primary care physicians and dermatologists.

Surgical Management of Melanoma and Other Skin Cancers

10.2310/surg.2038 ◽

2015 ◽

Author(s):

Jennifer A. Wargo ◽

Kenneth Tenabe

Keyword(s):

Lymph Node ◽

Surgical Treatment ◽

Skin Cancer ◽

Cell Carcinoma ◽

Stage Iv ◽

Stage Iii ◽

Skin Cancers ◽

Stage Iv Melanoma ◽

The prevalence of malignant skin cancers has increased significantly over the past several years. Approximately 1.2 million cases of non-melanoma skin cancer are diagnosed per year. More alarming, up to 80,000 cases of melanoma are diagnosed per year, an incidence that has been steadily increasing, with a lifetime risk of 1 in 50 for the development of melanoma. The disturbing increase in the incidence of both non-melanoma skin cancer and melanoma can largely be attributed to the social attitude toward sun exposure. The clinical assessment and management of skin lesions can be challenging. This review describes the assessment process, including thorough history and examination; the need for possible biopsy; and excision criteria. Specific types of skin cancer are distinguished and include basal cell carcinoma; squamous cell carcinoma; and melanoma; and for each type the incidence; epidemiology; histologic subtypes; diagnosis; and both surgical and non-surgical treatments are provided. Stages I-IV of melanoma are detailed, with prognostic factors described. Surgical treatment for stages I and II include description of the margins of excision and sentinel lymph node biopsy. The surgical treatment of Stage III melanoma further includes therapeutic lymph node dissection and isolated limb perfusion. Adjuvant therapies are also presented and include radiotherapy and chemotherapy. The additional treatment of metastasectomy for Stage IV melanoma is described. For both Stage III and IV melanoma, the study of vaccines to host immune cells is reported. For Stage IV melanoma, the text also describes immunotherapy treatment. Operative procedures specific to superficial and deep groin dissections are outlined. This review contains 9 figures, 3 tables, and 96 references.

Treatment of Multiple Actinic Keratosis and Field of Cancerization with Topical Piroxicam 0.8% and Sunscreen 50+ in Organ Transplant Recipients: A Series of 10 Cases

Case Reports in Dermatology ◽

10.1159/000481770 ◽

2017 ◽

Vol 9 (3) ◽

pp. 211-216 ◽

Cited By ~ 13

Author(s):

Virginia Garofalo ◽

Alessandra Ventura ◽

Sara Mazzilli ◽

Laura Diluvio ◽

Luca Bianchi ◽

...

Keyword(s):

High Risk ◽

Skin Cancer ◽

Cell Carcinoma ◽

Medical Device ◽

Actinic Keratosis ◽

Organ Transplant ◽

Cyclooxygenase Inhibitors ◽

Organ Transplant Recipient ◽

Skin Cancers ◽

Organ transplant recipient (OTR) subjects are at high risk of skin cancer such as squamous cell carcinoma and basal cell carcinoma. Actinic keratosis (AK) is considered the precursor of these non-melanoma skin cancers. Sun protection is mandatory in subjects with AK and this preventive strategy is very important in OTR. Treatment of the field of cancerization is also crucial to reduce the risk of recurrence of skin lesions in AK and non-melanoma skin cancer patients. Activation of cyclooxygenase 1 and 2 enzymes plays an important role in the pathogenesis of skin cancers. Topical application of cyclooxygenase inhibitors such as diclofenac and, more recently, piroxicam has shown to reduce AK lesions in immunocompetent subjects. A medical device containing piroxicam and SPF 50+ sunscreen filters (P+SS) has been demonstrated to be effective in reducing AK lesions and improving the field of cancerization. We report the effect of P+SS, applied for 16 weeks, in a case series of 10 OTR subjects with multiple AK lesions. P+SS treatment was associated with a relevant AK lesion reduction (>75%) in 7 patients (with a complete clearance in 3 subjects) with an improvement in the field of cancerization. This medical device could be considered a promising long-term curative and preventive treatment in OTR patients at high risk of non-melanoma skin cancers.

Non-Melanoma Skin Cancer in People Living With HIV: From Epidemiology to Clinical Management

Frontiers in Oncology ◽

10.3389/fonc.2021.689789 ◽

2021 ◽

Vol 11 ◽

Author(s):

Emmanuele Venanzi Rullo ◽

Maria Grazia Maimone ◽

Francesco Fiorica ◽

Manuela Ceccarelli ◽

Claudio Guarneri ◽

...

Keyword(s):

Skin Cancer ◽

Early Stage ◽

Treatment Options ◽

Case Series ◽

Surgical Excision ◽

People Living With Hiv ◽

Skin Cancers ◽

Ablative Techniques ◽

Living With Hiv ◽

Skin cancers represent the most common human tumors with a worldwide increasing incidence. They can be divided into melanoma and non-melanoma skin cancers (NMSCs). NMSCs include mainly squamous cell (SCC) and basal cell carcinoma (BCC) with the latest representing the 80% of the diagnosed NMSCs. The pathogenesis of NMSCs is clearly multifactorial. A growing body of literature underlies a crucial correlation between skin cancer, chronic inflammation and immunodeficiency. Intensity and duration of immunodeficiency plays an important role. In immunocompromised patients the incidence of more malignant forms or the development of multiple tumors seems to be higher than among immunocompetent patients. With regards to people living with HIV (PLWH), since the advent of combined antiretroviral therapy (cART), the incidence of non-AIDS-defining cancers (NADCs), such as NMSCs, have been increasing and now these neoplasms represent a leading cause of illness in this particular population. PLWH with NMSCs tend to be younger, to have a higher risk of local recurrence and to have an overall poorer outcome. NMSCs show an indolent clinical course if diagnosed and treated in an early stage. BCC rarely metastasizes, while SCC presents a 4% annual incidence of metastasis. Nevertheless, metastatic forms lead to poor patient outcome. NMSCs are often treated with full thickness treatments (surgical excision, Mohs micro-graphic surgery and radiotherapy) or superficial ablative techniques (such as cryotherapy, electrodesiccation and curettage). Advances in genetic landscape understanding of NMSCs have favored the establishment of novel therapeutic strategies. Concerning the therapeutic evaluation of PLWH, it’s mandatory to evaluate the risk of interactions between cART and other treatments, particularly antiblastic chemotherapy, targeted therapy and immunotherapy. Development of further treatment options for NMSCs in PLWH seems needed. We reviewed the literature after searching for clinical trials, case series, clinical cases and available databases in Embase and Pubmed. We review the incidence of NMSCs among PLWH, focusing our attention on any differences in clinicopathological features of BCC and SCC between PLWH and HIV negative persons, as well as on any differences in efficacy and safety of treatments and response to immunomodulators and finally on any differences in rates of metastatic disease and outcomes.

Skin cancer

10.1093/med/9780199568741.003.0259 ◽

2018 ◽

Author(s):

Rubeta Matin ◽

Jane McGregor ◽

Catherine Harwood

Keyword(s):

Skin Cancer ◽

Cell Carcinoma ◽

Cutaneous Metastasis ◽

Primary Malignancy ◽

Skin Cancers ◽

Cutaneous Lymphomas ◽

The Uk ◽

Late Stages ◽

Skin cancer is very common in the UK, and its incidence is rising rapidly. There are two broad classes of primary skin cancer: non-melanoma and melanoma. Non-melanoma skin cancer is the commonest form (100 000 cases diagnosed annually in the UK), accounting for nine out of ten skin cancers and includes basal cell carcinoma (BCC) and squamous cell carcinoma (SCC). Cutaneous melanoma is less common (10 000 cases diagnosed in the UK annually) but confers a significantly worse prognosis and accounts for 75% of skin cancer related deaths. There are also a number of other, rarer, non-melanoma skin cancers (e.g. appendageal carcinomas, Merkel cell carcinoma, sarcomas, vascular malignancies, and cutaneous lymphomas); however, these account for less than 1% of all skin cancers in the UK and so will not be specifically discussed in this chapter. Cutaneous metastases can occur secondary to any internal cancer or, indeed, to skin cancer (e.g. melanoma). In most cases, cutaneous metastasis occurs after the diagnosis of a primary cancer and usually in late stages of the disease but, in some cases, it may be the first presentation, in which case it should prompt a thorough investigation for the primary malignancy.

Mohs Surgery versus Standard Local Excision for Basal Cell Carcinoma, Squamous Cell Carcinoma, and Melanoma Skin Cancer

Facial Plastic Surgery ◽

10.1055/s-0040-1709142 ◽

2020 ◽

Vol 36 (02) ◽

pp. 133-140

Author(s):

Timothy M. Johnson ◽

Noah R. Smith

Keyword(s):

Risk Factors ◽

Squamous Cell Carcinoma ◽

Skin Cancer ◽

Cell Carcinoma ◽

Local Excision ◽

Mohs Surgery ◽

Skin Cancers ◽

The Face ◽

Carcinoma Squamous Cell ◽

AbstractBasal cell carcinoma, squamous cell carcinoma, and melanoma represent the three most common skin cancers that occur on the face. The most common surgical treatments for facial skin cancers are Mohs surgery and standard local excision. The effective utilization of either of these techniques is based on tumor and patient risk stratification incorporating known risk factors for occult invasion and local recurrence, combined with patient comorbidities, expectations, and desires. Best available evidence highlights multiple and consistent risk factors for each specific skin cancer type, and dictate local control rates reported in the literature. Recognizing gaps in the literature, we compare and review surgical treatment guidelines and data for standard local excision versus Mohs surgery for cutaneous nonmelanoma and melanoma skin cancer. This article serves as a resource for optimal therapeutic decision making for surgical management of skin cancer on the face.

The incidence and clinical analysis of non-melanoma skin cancer

Scientific Reports ◽

10.1038/s41598-021-83502-8 ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Magdalena Ciążyńska ◽

Grażyna Kamińska-Winciorek ◽

Dariusz Lange ◽

Bogumił Lewandowski ◽

Adam Reich ◽

...

Keyword(s):

Skin Cancer ◽

Cell Carcinoma ◽

Treatment Strategies ◽

Cancer Registries ◽

Clinical Analysis ◽

Tumour Site ◽

Skin Cancers ◽

The Face ◽

AbstractNon-melanoma skin cancers (NMSCs) are the most common malignancies diagnosed in Caucasian populations. Basal cell carcinoma (BCC) is the most frequent skin cancer, followed by squamous cell carcinoma (SCC). Unfortunately, most European cancer registries do not record individual types of NMSC. To evaluate the incidence of primary BCCs and SCCs regarding age, sex, tumour site and tumour subtype to determine trends in epidemiology of both cancers. Retrospective analysis of BCCs and SCCs diagnosed and treated across seven sites in Poland from 1999 to 2019. We recorded 13,913 NMSCs occurring in 10,083 patients. BCC represented 85.2% of all cases. SCC patients were older than BCC patients (77.1 ± 11.3 years vs. 70.1 ± 12.3 years, p < 0.01). The nodular subtype was the most common subtype of BCC, followed by the superficial and infiltrative subtypes. The superficial BCC subtype was more common on photoprotected areas (p < 0.01), whereas the nodular BCC subtype occurred on the face (p < 0.01). The high-risk SCC subtypes were more common on face compared to low-risk SCC subtypes (p < 0.01). BCC and SCC are common malignancies developing at various ages and anatomical sites. These data underline the need for better registration policies regarding NMSC in order to improve prevention and treatment strategies for these tumours.

Preoperative AI-Driven Fluorescence Diagnosis of Non-Melanoma Skin Cancer

Diagnostics ◽

10.3390/diagnostics12010072 ◽

2021 ◽

Vol 12 (1) ◽

pp. 72

Author(s):

Victoriya Andreeva ◽

Evgeniia Aksamentova ◽

Andrey Muhachev ◽

Alexey Solovey ◽

Igor Litvinov ◽

...

Keyword(s):

Skin Cancer ◽

Normal Skin ◽

Invasive Procedure ◽

Biopsy Material ◽

Fluorescence Diagnosis ◽

Skin Cancers ◽

Abnormal Cells ◽

Diagnosis Method ◽

The diagnosis and treatment of non-melanoma skin cancer remain urgent problems. Histological examination of biopsy material—the gold standard of diagnosis—is an invasive procedure that requires a certain amount of time to perform. The ability to detect abnormal cells using fluorescence spectroscopy (FS) has been shown in many studies. This technique is rapidly expanding due to its safety, relative cost-effectiveness, and efficiency. However, skin lesion FS-based diagnosis is challenging due to a number of single overlapping spectra emitted by fluorescent molecules, making it difficult to distinguish changes in the overall spectrum and the molecular basis for it. We applied deep learning (DL) algorithms to quantitatively assess the ability of FS to differentiate between pathologies and normal skin. A total of 137 patients with various forms of primary and recurrent basal cell carcinoma (BCC) were observed by a multispectral laser-based device with a built-in neural network (NN) “DSL-1”. We measured the fluorescence spectra of suspected non-melanoma skin cancers and compared them with “normal” skin spectra. These spectra were input into DL algorithms to determine whether the skin is normal, pigmented normal, benign, or BCC. The preoperative differential AI-driven fluorescence diagnosis method correctly predicted the BCC lesions. We obtained an average sensitivity of 62% and average specificity of 83% in our experiments. Thus, the presented “DSL-1” diagnostic device can be a viable tool for the real-time diagnosis and guidance of non-melanoma skin cancer resection.

Aplikasi Mobile Deteksi Dini Kanker Kulit Berdasarkan Image Processing

Jurnal Litbang Edusaintech ◽

10.51402/jle.v2i2.44 ◽

2021 ◽

Vol 2 (2) ◽

pp. 100-106

Author(s):

Fina Royana ◽

Puput Yuniar Maulida ◽

Rully Nurul Hasanah ◽

Sondari Setia Rahayu ◽

Rasim Rasim

Keyword(s):

Machine Learning ◽

Skin Cancer ◽

Cancer Detection ◽

Future Health ◽

Skin Cancers ◽

Threatening Condition ◽

Life Threatening ◽

Management Techniques ◽

Skin Cancer Detection ◽

Currently, between 2 and 3 million non-melanoma skin cancers and 132,000 melanoma skin cancers occur globally each year (WHO, 2017). Skin cancer is one type of cancer that can cause death for many people. Because of this, an application is needed to easily detect skin cancer early that the cancer can be handled with more quickly. Besides, consultations with dermatologists have better prognosis (Avilés-Izquierdo et. al., 2016). Due to that, we built an early skin cancer detection application with dermatologist consultation. Our application helps to diagnose skin cancer before it grows into a life-threatening condition and is crucial to preserving lifestyle, future health, and aesthetics. Besides, thanks to online doctor consultations we have, however, getting diagnosed, prescribed and treated for your issues without spending time travelling to and from the doctors and waiting in queues can be just as effective. We used three management techniques such as machine learning to create data pipelines, build a model, and convert the model to TensorFlow lite with post-training quantization. Android to deploy the TensorFlow lite model and create the application. The application has a real-time connection using firebase. Moreover, cloud to create a simple database for doctor and diagnosis services on firebase.

RNA-seq reveals more consistent reference genes for gene expression studies in human non-melanoma skin cancers

PeerJ ◽

10.7717/peerj.3631 ◽

2017 ◽

Vol 5 ◽

pp. e3631 ◽

Cited By ~ 19

Author(s):

Van L.T. Hoang ◽

Lisa N. Tom ◽

Xiu-Cheng Quek ◽

Jean-Marie Tan ◽

Elizabeth J. Payne ◽

...

Keyword(s):

Skin Cancer ◽

Reference Genes ◽

Ribosomal Proteins ◽

Data Interpretation ◽

Rna Seq ◽

Skin Cancers ◽

Genes Encoding ◽

Gene Expression Studies ◽

Identification of appropriate reference genes (RGs) is critical to accurate data interpretation in quantitative real-time PCR (qPCR) experiments. In this study, we have utilised next generation RNA sequencing (RNA-seq) to analyse the transcriptome of a panel of non-melanoma skin cancer lesions, identifying genes that are consistently expressed across all samples. Genes encoding ribosomal proteins were amongst the most stable in this dataset. Validation of this RNA-seq data was examined using qPCR to confirm the suitability of a set of highly stable genes for use as qPCR RGs. These genes will provide a valuable resource for the normalisation of qPCR data for the analysis of non-melanoma skin cancer.

Increased Incidence Of Melanoma and Non-Melanoma Skin Cancers In Patients With Hairy Cell Leukemia: A Single Institution Experience With 267 Patients From Memorial Sloan-Kettering Cancer Center

Blood ◽

10.1182/blood.v122.21.5274.5274 ◽

2013 ◽

Vol 122 (21) ◽

pp. 5274-5274

Author(s):

Justin M Watts ◽

Ashwin Kishtagari ◽

Sean Devlin ◽

Eytan M Stein ◽

Jae H Park ◽

...

Keyword(s):

Skin Cancer ◽

General Population ◽

Cancer Center ◽

Hispanic Males ◽

Skin Cancers ◽

Increased Risk ◽

Seer Data ◽

Almost All ◽

Abstract Chronic lymphocytic leukemia and other B-cell malignancies have been associated with melanoma and non-melanoma skin cancers (NMSC). However, an analysis of Surveillance, Epidemiology and End Results (SEER) data from 1973-2007 found that hairy cell leukemia (HCL), while associated with an increased second cancer risk overall, was not associated with melanoma. In addition, the incidence of NMSC in HCL patients has not been described to our knowledge. Per recent SEER data, the median age at melanoma diagnosis in the general population was 61 years with an age-adjusted incidence rate of 0.02%/year. Methods We identified 372 patients seen at Memorial Sloan-Kettering Cancer Center (MSKCC) over the past 30 years (1983-2013) with a morphologic diagnosis of HCL. Of these, we found 267 patients with ≥2 months of follow-up. We examined the medical records of these 267 patients for demographic data, treatment with purine analogs (PA), and co-occurring skin cancers, including melanoma, squamous cell carcinoma (SCC), and basal cell carcinoma (BCC). Skin cancers were considered to be “co-occurring” if they were diagnosed up to 1 year before or any time after the diagnosis of HCL. Results In this 267 patient cohort, the median age at HCL diagnosis was 52.1 years (range 19.6-86.1), and the vast majority of patients were white, non-Hispanic males [Table 1]. 225 patients (84%) were treated with a PA, either cladribine or pentostatin. Of 267 patients, 34 (12.7%) developed skin cancer: 11 (4.1%) melanoma and 25 (9.4%) NMSC [Table 2]. Twelve patients had SCC and 22 BCC. Eleven of 34 patients (32%) had >1 type of skin cancer: 9 BCC and SCC, 1 BCC and melanoma, and 1 SCC and melanoma.For the 34 patients with skin cancer, median follow-up from HCL diagnosis was 10 years (0.7-33.6), median age at HCL diagnosis was 57.5 years, and almost all patients were white, non-Hispanic males. Twenty-nine of the 34 patients (85%) received a PA. Nine patients (27%) either did not receive or were diagnosed with skin cancer before PA therapy. Eighteen patients (53%) were diagnosed with skin cancer between 1 year before and 5 years after HCL; 16 patients (47%) were diagnosed >5 years after HCL. Conclusions In 267 HCL patients with very long follow-up, we found a high incidence of all skin cancers (12.7%), melanoma (4.1%), and NMSC (9.4%). Furthermore, the risk of melanoma appears to be considerably higher in the HCL cohort than the general population (0.02%/year). Although these groups were not age, sex, or race-matched, both HCL and melanoma typically occur in white individuals, and one might expect the risk of melanoma to be lower in HCL patients if there was no association given that HCL usually presents at a younger age. Although a previous analysis of SEER data did not show an association between HCL and melanoma, many of these data were collected before PA therapy was introduced. Moreover, almost all melanoma patients in our cohort were previously treated with a PA, possibly explaining the increased risk. The pathogenesis of this apparent association is elusive, but immunosuppression induced by PA therapy in addition to inherent immunosuppression from HCL itself may be responsible. There also appears to be an increased risk of NMSC in our cohort; however, the precise incidence of NMSC in the general population is not available for comparison to our knowledge. Our findings reinforce that HCL patients should be screened aggressively for skin cancer, particularly given the risk of synchronous melanoma. Disclosures: No relevant conflicts of interest to declare.