The rapid increase in telemedicine coupled with recent advances in diagnostic artificial intelligence (AI) create the imperative to consider the opportunities and risks of inserting AI-based support ...into new paradigms of care. Here we build on recent achievements in the accuracy of image-based AI for skin cancer diagnosis to address the effects of varied representations of AI-based support across different levels of clinical expertise and multiple clinical workflows. We find that good quality AI-based support of clinical decision-making improves diagnostic accuracy over that of either AI or physicians alone, and that the least experienced clinicians gain the most from AI-based support. We further find that AI-based multiclass probabilities outperformed content-based image retrieval (CBIR) representations of AI in the mobile technology environment, and AI-based support had utility in simulations of second opinions and of telemedicine triage. In addition to demonstrating the potential benefits associated with good quality AI in the hands of non-expert clinicians, we find that faulty AI can mislead the entire spectrum of clinicians, including experts. Lastly, we show that insights derived from AI class-activation maps can inform improvements in human diagnosis. Together, our approach and findings offer a framework for future studies across the spectrum of image-based diagnostics to improve human-computer collaboration in clinical practice.
Background Evolving dermoscopic terminology motivated us to initiate a new consensus. Objective We sought to establish a dictionary of standardized terms. Methods We reviewed the medical literature, ...conducted a survey, and convened a discussion among experts. Results Two competitive terminologies exist, a more metaphoric terminology that includes numerous terms and a descriptive terminology based on 5 basic terms. In a survey among members of the International Society of Dermoscopy (IDS) 23.5% (n = 201) participants preferentially use descriptive terminology, 20.1% (n = 172) use metaphoric terminology, and 484 (56.5%) use both. More participants who had been initially trained by metaphoric terminology prefer using descriptive terminology than vice versa (9.7% vs 2.6%, P < .001). Most new terms that were published since the last consensus conference in 2003 were unknown to the majority of the participants. There was uniform consensus that both terminologies are suitable, that metaphoric terms need definitions, that synonyms should be avoided, and that the creation of new metaphoric terms should be discouraged. The expert panel proposed a dictionary of standardized terms taking account of metaphoric and descriptive terms. Limitations A consensus seeks a workable compromise but does not guarantee its implementation. Conclusion The new consensus provides a revised framework of standardized terms to enhance the consistent use of dermoscopic terminology.
Whether machine-learning algorithms can diagnose all pigmented skin lesions as accurately as human experts is unclear. The aim of this study was to compare the diagnostic accuracy of state-of-the-art ...machine-learning algorithms with human readers for all clinically relevant types of benign and malignant pigmented skin lesions.
For this open, web-based, international, diagnostic study, human readers were asked to diagnose dermatoscopic images selected randomly in 30-image batches from a test set of 1511 images. The diagnoses from human readers were compared with those of 139 algorithms created by 77 machine-learning labs, who participated in the International Skin Imaging Collaboration 2018 challenge and received a training set of 10 015 images in advance. The ground truth of each lesion fell into one of seven predefined disease categories: intraepithelial carcinoma including actinic keratoses and Bowen's disease; basal cell carcinoma; benign keratinocytic lesions including solar lentigo, seborrheic keratosis and lichen planus-like keratosis; dermatofibroma; melanoma; melanocytic nevus; and vascular lesions. The two main outcomes were the differences in the number of correct specific diagnoses per batch between all human readers and the top three algorithms, and between human experts and the top three algorithms.
Between Aug 4, 2018, and Sept 30, 2018, 511 human readers from 63 countries had at least one attempt in the reader study. 283 (55·4%) of 511 human readers were board-certified dermatologists, 118 (23·1%) were dermatology residents, and 83 (16·2%) were general practitioners. When comparing all human readers with all machine-learning algorithms, the algorithms achieved a mean of 2·01 (95% CI 1·97 to 2·04; p<0·0001) more correct diagnoses (17·91 SD 3·42 vs 19·92 4·27). 27 human experts with more than 10 years of experience achieved a mean of 18·78 (SD 3·15) correct answers, compared with 25·43 (1·95) correct answers for the top three machine algorithms (mean difference 6·65, 95% CI 6·06–7·25; p<0·0001). The difference between human experts and the top three algorithms was significantly lower for images in the test set that were collected from sources not included in the training set (human underperformance of 11·4%, 95% CI 9·9–12·9 vs 3·6%, 0·8–6·3; p<0·0001).
State-of-the-art machine-learning classifiers outperformed human experts in the diagnosis of pigmented skin lesions and should have a more important role in clinical practice. However, a possible limitation of these algorithms is their decreased performance for out-of-distribution images, which should be addressed in future research.
None.
Most of the studies characterizing DNA methylation patterns have been restricted to particular genomic loci in a limited number of human samples and pathological conditions. Herein, we present a ...compromise between an extremely comprehensive study of a human sample population with an intermediate level of resolution of CpGs at the genomic level. We obtained a DNA methylation fingerprint of 1628 human samples in which we interrogated 1505 CpG sites. The DNA methylation patterns revealed show this epigenetic mark to be critical in tissue-type definition and stemness, particularly around transcription start sites that are not within a CpG island. For disease, the generated DNA methylation fingerprints show that, during tumorigenesis, human cancer cells underwent a progressive gain of promoter CpG-island hypermethylation and a loss of CpG methylation in non-CpG-island promoters. Although transformed cells are those in which DNA methylation disruption is more obvious, we observed that other common human diseases, such as neurological and autoimmune disorders, had their own distinct DNA methylation profiles. Most importantly, we provide proof of principle that the DNA methylation fingerprints obtained might be useful for translational purposes by showing that we are able to identify the tumor type origin of cancers of unknown primary origin (CUPs). Thus, the DNA methylation patterns identified across the largest spectrum of samples, tissues, and diseases reported to date constitute a baseline for developing higher-resolution DNA methylation maps and provide important clues concerning the contribution of CpG methylation to tissue identity and its changes in the most prevalent human diseases.
Background Early detection of melanoma is the best way to improve prognosis. Digital follow-up (DFU) programs of populations at high risk could be an efficient strategy for detecting early melanomas ...with low morbidity. Objective We sought to report the added value of the use of the “two-step method” (digital total body photography and digital dermatoscopy). Methods This was an analysis of the surveillance of 618 patients at high risk for melanoma included in our DFU program from 1999 to 2008. Results A total of 11,396 lesions were monitored (mean 18.44/patient) during a median follow-up of 96 months (median 10 visits/patient). A total of 1152 lesions, 1.86 per patient, were excised. Almost 70% (798) were lesions previously registered at least twice, whereas 356 (30%) were detected and removed in the same visit. During follow-up, 98 melanomas (8.5% of excised lesions) were diagnosed in 78 patients (12.6%). In all, 53 melanomas were in situ (53.3%), whereas invasive (45) showed a Breslow index of less than 1 mm (median 0.5 mm) and none were ulcerated. Limitations Because there are no control groups we cannot determine if the combined use of total body photography and digital dermatoscopy is more beneficial than these techniques used separately. Conclusion DFU with total body photography and dermatoscopy in a selected population at high risk demonstrated the early detection of melanomas with a low rate of excisions. Long-term follow-up is required to allow the detection of slow-growing melanomas. Based on our 10-year experience, melanomas can be diagnosed at any time, suggesting that in a population at high risk for melanoma, DFU should be maintained over time.
Background Cyclin-dependent kinase inhibitor 2A ( CDKN2A ) is the major high-risk susceptibility gene for melanoma. Objective We sought to evaluate the effect of CDKN2A mutations in Spanish patients ...with a high risk of developing melanoma and the association with clinical and family history features. Methods A cross-sectional study design was used to analyze the CDKN2A impact in 702 Spanish patients with a high risk of developing melanoma. Results The CDKN2A mutation prevalence was 8.5% in patients with sporadic multiple primary melanoma and 14.1% in familial melanoma. Number of cases in the family, number of primary melanomas, and age of onset were associated with the presence of CDKN2A mutation. Having a CDKN2A mutation in the family increased the prevalence of other cancers (prevalence ratio PR 2.99, P = .012) and prevalence of pancreatic (PR 2.97, P = .006), lung (PR 3.04, P < .001), and breast (PR 2.19, P = .018) cancers but not nephrourologic or colon cancer. Limitations Smoking status was not assessed in the individuals with lung cancer. Conclusions Melanoma-prone families with mutations in CDKN2A have an increased prevalence of a broad spectrum of cancers including lung, pancreatic, and breast cancer. This information should be included in genetic counseling and cancer prevention programs for CDKN2A mutation carriers.
Summary Background Vismodegib, a first-in-class Hedgehog-pathway inhibitor, is approved for use in adults with advanced basal-cell carcinoma. Patients with multiple basal-cell carcinomas, including ...those with basal-cell nevus (Gorlin) syndrome, need extended treatment. We assessed the safety and activity of two long-term intermittent vismodegib dosing regimens in patients with multiple basal-cell carcinomas. Methods In this randomised, regimen-controlled, double-blind, phase 2 trial, we enrolled adult patients with multiple basal-cell carcinomas, including those with basal-cell nevus syndrome, who had one or more histopathologically confirmed and at least six clinically evident basal-cell carcinomas. From a centralised randomisation schedule accessed via an interactive voice or web-based response system, patients were randomly assigned (1:1) to treatment group A (150 mg oral vismodegib per day for 12 weeks, then three rounds of 8 weeks of placebo daily followed by 12 weeks of 150 mg vismodegib daily) or treatment group B (150 mg oral vismodegib per day for 24 weeks, then three rounds of 8 weeks of placebo daily followed by 8 weeks of 150 mg vismodegib daily). Treatment assignment was stratified by diagnosis of basal-cell nevus syndrome, geographical region, and immunosuppression status. The primary endpoint was percentage reduction from baseline in the number of clinically evident basal-cell carcinomas at week 73. The primary analysis was by intention to treat. The safety population included all patients who received at least one dose of study drug. This trial is registered with ClinicalTrials.gov , number NCT01815840 , and the study is ongoing. Findings Between April 30, 2013, and April 9, 2014, 229 patients were randomly assigned treatment, 116 in treatment group A and 113 in treatment group B. The mean number of basal-cell carcinoma lesions at week 73 was reduced from baseline by 62·7% (95% CI 53·0–72·3) in treatment group A and 54·0% (43·6–64·4) in treatment group B. 216 (95%) of 227 patients included in the safety analysis had at least one treatment-emergent adverse event deemed to be related to study treatment (107 94% of 114 in treatment group A and 109 97% of 113 in treatment group B). The most common grade 3 or worse treatment-related adverse events were muscle spasms (four 4% patients in treatment group A vs 12 11% in treatment group B), increased blood creatine phosphokinase (one 1% vs four 4%), and hypophosphataemia (zero vs three 3%). Serious treatment-emergent events were noted in 22 (19%) patients in treatment group A and 19 (17%) patients in treatment group B. Four (2%) patients died from adverse events; one (pulmonary embolism in treatment group A) was possibly related to treatment. Interpretation Both intermittent dosing schedules of vismodegib seemed to show good activity in long-term regimens in patients with multiple basal-cell carcinomas. Further study is warranted. Funding F Hoffmann-La Roche.
So far, no common environmental and/or phenotypic factor has been associated with melanoma and renal cell carcinoma (RCC). The known risk factors for melanoma include sun exposure, pigmentation and ...nevus phenotypes; risk factors associated with RCC include smoking, obesity and hypertension. A recent study of coexisting melanoma and RCC in the same patients supports a genetic predisposition underlying the association between these two cancers. The microphthalmia-associated transcription factor (MITF) has been proposed to act as a melanoma oncogene; it also stimulates the transcription of hypoxia inducible factor (HIF1A), the pathway of which is targeted by kidney cancer susceptibility genes. We therefore proposed that MITF might have a role in conferring a genetic predisposition to co-occurring melanoma and RCC. Here we identify a germline missense substitution in MITF (Mi-E318K) that occurred at a significantly higher frequency in genetically enriched patients affected with melanoma, RCC or both cancers, when compared with controls. Overall, Mi-E318K carriers had a higher than fivefold increased risk of developing melanoma, RCC or both cancers. Codon 318 is located in a small-ubiquitin-like modifier (SUMO) consensus site (ΨKXE) and Mi-E318K severely impaired SUMOylation of MITF. Mi-E318K enhanced MITF protein binding to the HIF1A promoter and increased its transcriptional activity compared to wild-type MITF. Further, we observed a global increase in Mi-E318K-occupied loci. In an RCC cell line, gene expression profiling identified a Mi-E318K signature related to cell growth, proliferation and inflammation. Lastly, the mutant protein enhanced melanocytic and renal cell clonogenicity, migration and invasion, consistent with a gain-of-function role in tumorigenesis. Our data provide insights into the link between SUMOylation, transcription and cancer.
Celotno besedilo
Dostopno za:
DOBA, IJS, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
Background Garment-related terms have been used to describe the pattern of distribution of giant congenital melanocytic nevi (GCMN). Objective We sought to describe patterns of distribution of GCMN ...and propose a classification scheme. Methods Photographic records of patients with GCMN from the Hospital Clinic of Barcelona were analyzed and a classification based on observed GCMN distribution patterns was created. The classification was independently applied by 8 observers to cases found in the literature. The interobserver agreement was assessed. Results Among 22 patients we observed 6 repeatable patterns of distribution of GCMN, which we termed the “6B”: bolero (involving the upper aspect of the back, including the neck), back (on the back, without involvement of the buttocks or shoulders), bathing trunk (involving the genital region and buttocks), breast/belly (isolated to the chest or abdomen without involvement of bolero or bathing trunk distributions), body extremity (isolated to extremity), and body (both bolero and bathing trunk involvement). Our literature search found 113 cases of GCMN, which we were able to classify into 1 of the 6B patterns with an overall kappa of 0.89. Limitations Some patterns occur infrequently with a dearth of images available for analysis. Conclusions The anatomic distribution of GCMN occurs in 6 recognizable and repeatable patterns.
PDF
