Triage amalgamated dermoscopic algorithm (TADA) for skin cancer screening

Dermoscopic structures and patterns used in melanoma detection

Melanoma is the leading cause of skin cancer-related deaths. Yet, early detection remains the most cost-effective means of preventing death from melanoma. Early detection can be achieved by a physician and/or the patient (also known as a self-skin exam). Skin exams performed by physicians are further enhanced using dermoscopy. Dermoscopy is a non-invasive technique that allows for the visualization of subsurface structures that are otherwise not visible to the naked eye. Evidence demonstrates that dermoscopy improves the diagnostic accuracy for skin cancer, including melanoma; it decreases the number of unnecessary skin biopsies of benign lesions and improves the benign-to-malignant biopsy ratio. Yet, these improvements are contingent on acquiring dermoscopy training. Dermoscopy is used by clinicians who evaluate skin lesions and perform skin cancer screenings. In general, under dermoscopy nevi tend to appear as organized lesions, with one or two structures and colors, and no melanoma-specific structures. In contrast, melanomas tend to manifest a disorganized pattern, with more than two colors and, usually, at least one melanoma-specific structure. This review is intended to familiarize the reader with the dermoscopic structures and patterns used in melanoma detection.

Utility of Dermoscopy Training in Improving Diagnostic Accuracy of Skin Lesions Among Physician Assistant Students

INTRODUCTION

Patients often first present to their primary care provider for skin lesion concerns, and dermoscopy is a tool that enhances diagnostic acumen of both malignant and benign skin lesions. Physician assistants (PAs) frequently serve as primary care and dermatology providers, but to our knowledge, no current research on dermoscopy expertise with PAs exists. We hypothesize that PA students could be taught dermoscopy based on the triage amalgamated dermoscopic algorithm (TADA) to increase their diagnostic skill, as previously shown with medical students.

METHODS

Dermoscopy was taught to first-year PA students at all 5 PA programs in the state of Minnesota. The training was 50 minutes in length and focused on the fundamentals of the TADA method. Physician assistant students participated in a pretraining and post-training test, consisting of 30 dermoscopic images.

RESULTS

A total of 139/151 (92%) PA students completed both the pretraining and post-training tests. Overall, mean scores for all students increased significantly ( P < .0001) after dermoscopy training was given (18.5 ± 7.1 vs. 23.8 ± 6.7).

CONCLUSION

Our study demonstrates that after TADA training, PA students improved their ability to assess dermoscopy images of both skin cancer and benign lesions accurately, suggesting that PAs can be trained as novice dermoscopists and provide better dermatologic care to patients. We strongly encourage integration of dermoscopy into didactic education across PA programs. Implementing a dermoscopy curriculum in established PA programs will enable future PAs to provide better clinical care when evaluating skin lesions.

Issues in Melanoma Detection: Semisupervised Deep Learning Algorithm Development via a Combination of Human and Artificial Intelligence

BACKGROUND

Automatic skin lesion recognition has shown to be effective in increasing access to reliable dermatology evaluation; however, most existing algorithms rely solely on images. Many diagnostic rules, including the 3-point checklist, are not considered by artificial intelligence algorithms, which comprise human knowledge and reflect the diagnosis process of human experts.

OBJECTIVE

In this paper, we aimed to develop a semisupervised model that can not only integrate the dermoscopic features and scoring rule from the 3-point checklist but also automate the feature-annotation process.

METHODS

We first trained the semisupervised model on a small, annotated data set with disease and dermoscopic feature labels and tried to improve the classification accuracy by integrating the 3-point checklist using ranking loss function. We then used a large, unlabeled data set with only disease label to learn from the trained algorithm to automatically classify skin lesions and features.

RESULTS

After adding the 3-point checklist to our model, its performance for melanoma classification improved from a mean of 0.8867 (SD 0.0191) to 0.8943 (SD 0.0115) under 5-fold cross-validation. The trained semisupervised model can automatically detect 3 dermoscopic features from the 3-point checklist, with best performances of 0.80 (area under the curve [AUC] 0.8380), 0.89 (AUC 0.9036), and 0.76 (AUC 0.8444), in some cases outperforming human annotators.

CONCLUSIONS

Our proposed semisupervised learning framework can help with the automatic diagnosis of skin disease based on its ability to detect dermoscopic features and automate the label-annotation process. The framework can also help combine semantic knowledge with a computer algorithm to arrive at a more accurate and more interpretable diagnostic result, which can be applied to broader use cases.

Dermoscopy of Small Diameter Melanomas with the Diagnostic Feasibility of Selected Algorithms-A Clinical Retrospective Multicenter Study

Simple Summary

The early detection of melanoma determines the recovery of the patient. Dermoscopy, which is one of the diagnostic tools for pigmented lesions, is characterized by high sensitivity and specificity, giving the clinician the possibility to detect the presence of abnormal structures before their clinical presentation. There are a small number of dermoscopic analyses of pigmented lesions of less than 6 mm in diameter in the published literature. The authors attempted to identify characteristic dermoscopic structures typical for melanomas of less than 5 mm in diameter in comparison with a group of melanomas exceeding this dimension at an identical clinical stage. It was found that dermoscopy in the secondary prevention of micromelanomas (appearing mainly as brown lesions) revealed the presence of dotted or polymorphous vessels, with architectural disorder in half of cases. Moreover, spitzoid, multicomponent asymmetric or nonmelanoma-specific patterns prevailed. Knowledge of these dermoscopic features brings the clinician closer to an early diagnosis of melanoma with a diameter of 5 mm or less.

Abstract

Objective: The aim of the study was to verify two hypotheses. The first concerned the possibility of diagnostic dermoscopic differentiation between cutaneous melanomas of the histopathological category in situ (pTis) and thin melanomas (pT1a) in terms of their diameter. The second assessed the diagnostic feasibility of two dermoscopic algorithms aiming to detect ≤ 5.0 mm-sized melanomas histopathologically confirmed as pTis and pT1a. Methods: Dermoscopic images of consecutive cases of histopathologically confirmed melanomas were evaluated by three independent investigators for the presence of the predefined criteria. The melanomas were subdivided according to their diameter into small melanomas, so-called micromelanomas (microM)—sized ≤ 5.0 mm and >5.0 mm, according to published definitions of small melanocytic lesions. The Triage Amalgamated Dermoscopic Algorithm (TADA) and the revisited 7-point checklist of dermoscopy (7-point) algorithm were chosen for the diagnostic feasibility. Odds ratios and corresponding 95% confidence limits (CL) were calculated using the logistic regression adjusted for age for the melanoma-specific dermoscopic structures, the dermoscopic patterns and the diagnostic feasibility of the 7-point checklist and TADA algorithms. The p-values of the results were corrected using the Bonferroni method. Results: In total, 106 patients with 109 melanomas, 50 sized ≤ 5.0 mm and 59 exceeding the diameter of 5.0 mm, were retrospectively analyzed. The prevalent general pattern of microM was the spitzoid one (48% vs. 11.86%, p = 0.0013). Furthermore, 40% of microM vs. 6.78% melanomas sized > 5.0 mm (p = 0.0023) did not present melanoma-specific patterns. The asymmetric multicomponent pattern was present in 64.41% melanomas sized > 5.0 mm and in 26.00% microM (p = 0.0034). The asymmetry of structures or colors was detected in 56% microM vs. 89.83% (p = 0.0020) and 56% microM and 94.92% (p = 0.000034) melanoma sized > 5.0 mm, respectively. The differences in frequency of the detected dermoscopic structures specific to melanomas revealed that microM are almost deprived of negative networks (p = 0.04), shiny white structures (p = 0.0027) and regression features (p = 0.00003). Neither prominent skin markings nor angulated lines were found in the entire study group. Out of the vascular structures, microM presented only dotted (32%) or polymorphous (28%) vessels, although more rarely than melanomas sized > 5.0 mm (66.1% p = 0.017 and 49% p > 0.05, respectively). The diagnostic feasibility revealed a score ≥ 3 of the 7-point algorithm (indicative for malignancy) in 60% microM and 98.31% melanomas sized > 5.0 mm (p = 0.000006). The TADA algorithm revealed melanoma-specific patterns in 64% microM and 96.61% > 5.0 mm-sized melanomas (p = 0.00006) and melanoma-specific structures in 72% and 91.53% (p > 0.05), respectively. Conclusion: In the dermoscopy, 40% of micromelanomas histopathologically staged as pTis and pT1a did not reveal melanoma-specific patterns. Among the general melanocytic patterns, the spitzoid one was the most frequently found in melanomas sized ≤ 5.0 mm. The 7-point checklist and TADA dermoscopic algorithms were helpful in the identification of the majority of melanomas sized ≤ 5.0 mm.

Dermoscopy: A Review of the Structures That Facilitate Melanoma Detection

Melanoma is currently the fifth most common cancer in the United States, resulting in more than 9000 deaths each year. Despite numerous improvements in the management of advanced melanoma, the cornerstone to ensuring a cure remains early detection. Both patient and physician awareness regarding the signs and symptoms of early melanoma remain paramount. As a result, much effort has been and continues to be expended in developing and refining effective diagnostic algorithms to help identify melanomas and differentiate them from nevi, such as the ABCDE rule (A for asymmetry, B for border irregularity, C for color variegation, D for diameter >6 mm, and E for evolution in lesion size, shape, or color). To assist in the detection of more subtle melanomas requires technology to augment a visual examination. Toward this end, a simple instrument called a dermatoscope has transformed not only the appreciation of the morphology of melanoma but also its growth dynamics. The discipline of dermoscopy has improved the detection of melanoma and other skin cancers, has resulted in the detection of thinner melanomas, and has helped improve the ability to differentiate nevi (benign lesions) from melanomas, which, in turn, has resulted in fewer biopsies of benign lesions. Since patients often first present to their primary care physicians for their health-related concerns, it is imperative that primary care physicians be able to recognize the lesions that are suspicious for melanoma. This review is intended to introduce osteopathic physicians to the dermoscopic features associated primarily with melanomas located on nonglabrous skin.

Dermoscopy Training Effect on Diagnostic Accuracy of Skin Lesions in Canadian Family Medicine Physicians Using the Triage Amalgamated Dermoscopic Algorithm

Background

Accurate identification of cutaneous lesions is an essential skill for family medicine physicians (FMPs). Studies show significant improvement in skin cancer detection with dermoscopy use. Frontline FMPs are an ideal target group for dermoscopy training. The 3-step Triage Amalgamated Dermoscopic Algorithm (TADA) facilitates high sensitivity and specificity for pigmented and nonpigmented skin lesions. Step I requires unequivocal identification of dermoscopic features for 1 of 3 benign skin lesions: angioma, dermatofibroma, or seborrheic keratosis. If absent, steps II and III are applied assessing for features of architectural disorder and malignancies with organized, symmetric patterns, respectively.

Objective

To assess FMPs' diagnostic accuracy of benign and malignant skin lesions before and after training in TADA step I.

Methods

In this repeated-measures observational study, 33 dermoscopy-naive FMPs attending an introductory dermoscopy workshop each assessed gross and corresponding dermoscopic photographic images of 50 pigmented and nonpigmented skin lesions (23 benign, 27 malignant) for features of TADA step I lesions before and after training. Analyses compared diagnostic accuracy in relation to training and baseline physician characteristics.

Results

Diagnostic accuracy improved from 76.4% to 90.8% (P < 0.001) and from 85.0% to 90.0% (P = 0.01), respectively, for all lesions and for all TADA I lesions. Female sex was significant as a predictor of individual posttraining performance (all lesions combined, P = 0.02).

Conclusions

Results show significant improvement in diagnostic accuracies for benign and malignant skin lesions with introductory dermoscopy training using TADA step I. This will reduce unnecessary benign lesion excision and enhance referral sensitivity, conserving specialist resources.

Dermoscopy for melanoma detection and triage in primary care: a systematic review

OBJECTIVE

Most skin lesions first present in primary care, where distinguishing rare melanomas from benign lesions can be challenging. Dermoscopy improves diagnostic accuracy among specialists and is promoted for use by primary care physicians (PCPs). However, when used by untrained clinicians, accuracy may be no better than visual inspection. This study aimed to undertake a systematic review of literature reporting use of dermoscopy to triage suspicious skin lesions in primary care settings, and challenges for implementation.

DESIGN

A systematic literature review and narrative synthesis.

DATA SOURCES

We searched MEDLINE, Cochrane Central, EMBASE, Cumulative Index to Nursing and Allied Health Literature, and SCOPUS bibliographic databases from 1 January 1990 to 31 December 2017, without language restrictions.

INCLUSION CRITERIA

Studies including assessment of dermoscopy accuracy, acceptability to patients and PCPs, training requirements, and cost-effectiveness of dermoscopy modes in primary care, including trials, diagnostic accuracy and acceptability studies.

RESULTS

23 studies met the review criteria, representing 49 769 lesions and 3708 PCPs, all from high-income countries. There was a paucity of studies set truly in primary care and the outcomes measured were diverse. The heterogeneity therefore made meta-analysis unfeasible; the data were synthesised through narrative review. Dermoscopy, with appropriate training, was associated with improved diagnostic accuracy for melanoma and benign lesions, and reduced unnecessary excisions and referrals. Teledermoscopy-based referral systems improved triage accuracy. Only three studies examined cost-effectiveness; hence, there was insufficient evidence to draw conclusions. Costs, training and time requirements were considered important implementation barriers. Patient satisfaction was seldom assessed. Computer-aided dermoscopy and other technological advances have not yet been tested in primary care.

CONCLUSIONS

Dermoscopy could help PCPs triage suspicious lesions for biopsy, urgent referral or reassurance. However, it will be important to establish further evidence on minimum training requirements to reach competence, as well as the cost-effectiveness and patient acceptability of implementing dermoscopy in primary care.

TRIAL REGISTRATION NUMBER

CRD42018091395.

Using Dermoscopy to Identify Melanoma and Improve Diagnostic Discrimination

Use of dermoscopy and detection algorithms by primary care physicians can enhance assessment of clinically suspicious lesions compared with that of naked eye examinations.