Smartphone applications for melanoma detection by community, patient and generalist clinician users: a review

Summary

Smartphone health applications (‘apps’) are widely available but experts remain cautious about their utility and safety. We reviewed currently available apps for the detection of melanoma (July 2014), aimed at general community, patient and generalist clinician users. A proforma was used to extract and assess each app that met the inclusion criteria, and we undertook content analysis to evaluate their content and the evidence applied in their development. Thirty‐nine apps were identified with the majority available only for Apple users. Over half (n =22) provided information or education about melanoma, ultraviolet radiation exposure prevention advice, and skin self‐examination strategies, mainly using the ABCDE (A, Asymmetry; B, Border; C, Colour; D, Diameter; E, Evolving) method. Half (n =19) helped users take and store images of their skin lesions either for review by a dermatologist or for self‐monitoring to identify change, an important predictor of melanoma; a similar number (n =18) used reminders to help users monitor their skin lesions. A few (n =9) offered expert review of images. Four apps provided a risk assessment to patients about the probability that a lesion was malignant or benign, and one app calculated users’ future risk of melanoma. None of the apps appeared to have been validated for diagnostic accuracy or utility using established research methods. Smartphone apps for detecting melanoma by nonspecialist users have a range of functions including information, education, classification, risk assessment and monitoring change. Despite their potential usefulness, and while clinicians may choose to use apps that provide information to educate their patients, apps for melanoma detection require further validation of their utility and safety.

Smartphones are rapidly evolving from being solely devices for communication and entertainment to include specialized applications (‘apps’) that are intimately involved in many aspects of daily life. A vast range of health apps is now available to assist users (> 13 000 in a 2012 report), for example to monitor their pulse and blood pressure, or to track their food intake and exercise undertaken to manage weight loss. Furthermore, two out of three U.S. clinicians already use smartphone health apps in their practice to manage a range of conditions.1 Some apps have been evaluated, such as those to assist in managing diabetes2 and pain,3 and to aid monitoring of anticoagulation therapy4 and epilepsy,5 but such evaluation is not common among apps aimed at general community users, probably due to the rapid evolution and commercial drivers of this field. While smartphones have been hailed as ‘new clinical tools in oncology’,1 many experts remain cautious about the utility of the thousands of apps currently available, either free or at a small charge, for the prevention, detection and management of cancer.6  7 It has also been suggested that apps for detecting cancers tend to lack scientific and specialty input,8 and the use of technology to deliver cancer follow‐up has only begun to be studied for safety and utility: a recently published systematic review found only two randomized studies that had used smartphone technology.9

A number of dermatology‐specific apps have been developed which aim to help previously unaffected individuals or those previously diagnosed with a skin cancer to decide if they should seek medical review for a skin lesion,10  11 and to assist nonspecialist clinicians such as general practitioners (GPs) to make decisions about whether to reassure the patient that their lesion is benign or to refer for specialist assessment.10 However, a recent comparison of the accuracy of four smartphone apps in assessing melanoma risk demonstrated wide variation in performance and utility.7 Furthermore, smartphone apps for the identification or management of cancer including melanoma have not been subject to any sort of validation or regulatory controls in the U.S.A., the U.K. or elsewhere. Therefore, while they have the potential to improve patient and nonspecialist clinician assessment and patient–clinician communication about potential skin cancer there is the risk that these apps could actually harm users. The advice could be inaccurate or misleading, apps could be used as a substitute for a clinical consultation, and they could even delay melanoma diagnosis.12 This risk is particularly concerning for melanoma compared with other cancers as the majority are detected by the patient rather than their clinician, and the time taken to present to a clinician from first noticing a skin change or symptom is longer than for all other cancers except those of the head and neck.13

In this paper we report a review of currently available smartphone apps for the detection of melanoma aimed at general community, patient or generalist clinician users, evaluating their content and the evidence applied in their development.

Methods

In July 2014 we searched the online stores of the two most popular smartphone providers (Apple14 and Android,15 which together provide more than 90% of cancer‐related apps6), for health apps that suggested any kind of support for previously unaffected individuals, or those previously diagnosed with a skin cancer, to detect melanoma. Our search terms included ‘skin cancer’, ‘mole’ and ‘melanoma’. We assessed the descriptions of each app identified by the searches, and included in this review all apps that stated an aim to help to detect melanoma. We excluded all apps that aimed to provide entertainment, cosmetics advice or general medical information alone. We also excluded those designed only for use by skin cancer specialists (dermatologists and plastic surgeons) including apps which use a dermoscopy attachment to the phone, but included any aimed specifically at generalist clinicians. We also excluded apps that were not available in English.

We developed a proforma specifically for this review, and one researcher (A.P.K.) extracted data about all the apps that met the inclusion criteria. The proforma was completed based only on the online descriptions; no apps were downloaded for data extraction purposes. All included apps were evaluated for their operating systems (Apple, Android, both), whether they were derived from research and/or validated, and the year of the latest update. The included apps were also evaluated for (i) general information about melanoma and/or skin cancer, ultraviolet radiation (UVR) and sun exposure preventive advice, and skin self‐examination strategies; (ii) risk factor assessment; (iii) cataloguing and/or classifying using ‘image analysis’, defined as a comparison between the lesion and a stored image or an algorithm to assess the image, or the forwarding of images to a dermatologist; and (iv) monitoring changes in moles over time.

Content analysis and critical appraisal were then performed independently by two researchers (A.P.K. and F.M.W.) to analyse the findings and refine the categories. The emerging categories were discussed and critiqued by all the experienced researchers, whose differing perspectives and expertise facilitated a robust and critical interpretation of assigned themes, and discussion and resolution of any differences by consensus.

Results

We identified 39 unique smartphone apps that met our inclusion criteria (Tables 1 and 2). Most are available for Apple Inc. devices only (26, 66·7%), with eight available for Android devices only (20·5%), and just five available for both platforms (12·8%).

Properties of available smartphone apps

Table 1 demonstrates the 39 smartphone apps currently available (July 2014) for previously unaffected individuals or those previously diagnosed with a skin cancer in the online stores of the two main application providers. Most of the apps (n =28) were available to download for free, with the remaining apps having an added cost to download (range £0·69–£4·99).

Table 2 shows the properties and functions of these apps. The commonest function was to provide general information about melanoma and/or skin cancer (n =22). Nineteen apps catalogued and classified lesions by capturing images, and applying integral algorithms, and 18 apps aimed to monitor skin lesions.

Information about melanoma and/or skin cancer was provided in a number of ways including access to online libraries. The ABCDE method (A, Asymmetry; B, Border; C, Colour; D, Diameter; E, Evolving) was most commonly used to provide education about skin self‐examination and self‐assessment of moles (n =12).16 Most of the apps provided more than one type of function. For example, the Mollie's Fund app provided information on skin self‐examination, the ABCDE method, and how to protect from UVR exposure; and the UMSkinCheck app provided information about the ABCDE method, the commonest types of skin lesions, and how to protect against UVR exposure. Four apps (Dermatology Planet, iDoc24, Skin Cancer, Skin of Mine) also offered guidance on how to find a dermatologist in the U.S.A. One app (UMSkinCheck) stated it was specifically designed for non‐Hispanic white U.S. citizens.

Cataloguing and classifying lesions was mainly undertaken by comparing the users’ own photographic images obtained by the smartphone against a set of exemplar images, such as in Embarrassing Bodies – My MoleChecker. Some apps contained integral algorithms for photographic analysis such as pattern recognition and real‐time computer image analysis technology in Doctor Mole and SpotMole, enabling the user to take a photographic image of the mole for assessment of each of the ABCDE criteria, and reporting changes in the features of the mole compared with previous photographic images. One app, Melanoma Visual Risk Calculator, used a Visual Analogue Scale (VAS) to compare the user's skin lesion against the ABCDE criteria. It was noteworthy that seven apps (Mole Detective, Mole Monitor, Skin of Mine, Skin Prevention – Photo Body Map, Skin Scanner, Skin Vision, SkinXM) provided limited or no information on how the photographic images were processed and analysed to provide advice on likelihood of melanoma. Apps from Apple Inc. are available for iPhones, iPads and iPod touch with software 3.1.3 and later, and apps from Android for all Android phones with software 1.6 and up. A minority of apps (n =9), such as Dermlink.md and Mole Check App, offered prompt expert review of images taken by the user. These were often accompanied by advice on making an appointment with a dermatologist, and functioned only in the U.S.

Using smartphone apps to monitor skin lesions was undertaken via the storage of photographic images of their skin lesions. Users were encouraged to track the evolution of individual lesions by comparing serial images over time. While some such as Doctor Mole and Skin Prevention used simple comparisons with stored images, others used more sophisticated algorithms, such as Skin Analytics, which claimed to use ‘real time vision technology’. Some, such as the Doctor Mole app, provided feedback to patients based on the ABCDE method; others, such as the Embarrassing Bodies – My MoleChecker and the Mole Detective apps allowed users to set reminders to take another photograph for comparison after a selected time. These apps also suggested that users could share the images during consultations with their dermatologist or family physician.

Only four apps provided tools to assess a skin lesion for likelihood of melanoma. The Doctor Mole and Skin Doctor apps used a computerized algorithm based on the ABCDE method, and the Melanoma Visual Risk Calculator provided a risk approximation based on patients’ responses to a VAS. iSkin claimed to gather information about a mole and then calculate risk of melanoma, but provided no further information about the method involved.

A single app offered assessment of a person's baseline risk factor: UMSkinCheck used eight risk factors (region, sex, race, age, complexion, tanning, small moles and freckling) to calculate 5‐year absolute risk for melanoma. However, there was no information about the model used to underpin this calculation.

Development and performance of apps

Even though several of the apps incorporated the ABCDE method, only one app acknowledged any research and developmental processes; the Melanoma Visual Risk Calculator was based on the diagnostic value of the ABCDE criteria for patient users.17 Clinician involvement in the development of apps was clearly stated in only four apps. None of the apps were reported to have been validated.

Version updates

Among the apps, the range of latest updates was 2009–14, with 17 updated in 2013–14, nine in 2012, seven in 2011, two in 2010 and one in 2009. Three apps (Doctor Mole – Skin Cancer app, Mollie's Fund, nēvus) had their Apple and Android versions updated in different years.

Discussion

This review has identified smartphone apps for detecting melanoma by nonspecialist users including previously unaffected individuals, patients previously diagnosed with skin cancer, and generalist clinicians. We have evaluated the content of the apps with respect to general information about melanoma and/or skin cancer, UVR and sun exposure preventive advice, skin self‐examination strategies, assessment of current and future melanoma risk, personal or expert image analysis to classify lesions, and lesion monitoring over time. We have identified 39 smartphone apps with the majority available only for Apple users. One‐third of these apps provided information only. A similar number enabled patients to capture and store images of their skin lesions either for review by a dermatologist or for self‐monitoring to identify change, an important predictor of melanoma. Four apps provided a risk assessment to patients about the probability that a lesion was malignant or benign, and only one app incorporated a validated risk model giving individuals information about their future risk of melanoma. None of the apps appeared to have been validated for diagnostic accuracy or utility using established research methods. There was limited information about whether the apps were developed with clinician involvement, and some apps had not been updated for more than 3 years.

While a recent systematic review of smartphone apps for the prevention, detection and management of cancer in general concluded that there has been ‘a lot of action, but not in the right direction’, as they tend not to incorporate technological innovation,6 we have identified a range of uses of smartphone technology for melanoma detection. For example, image analysis allows for self‐monitoring, and the use of reminders and alerts promote the added use of monitoring skin changes over time. Furthermore, the technology can use a ‘teledermatology’ model, and send the image for expert review and advice. Nevertheless, one‐third of the apps that we identified only promoted awareness of cancer symptoms or solar protection, with no potential for interaction or specialist overview, despite the strong evidence suggesting that the provision of information alone is not an effective intervention to improve patient self‐management of chronic disease in primary care.6  18 Adding self‐monitoring techniques could be more effective than information‐provision alone.19 Moreover, as communication with clinicians may improve adherence to e‐healthcare interventions,20 apps that facilitate communication with skin specialists may have utility, particularly for individuals previously diagnosed with a skin cancer.

Almost a quarter of the apps used a teledermatology model to send the image for expert review. Evaluations of remote assessment of clinical photographs and/or dermoscopic images have given good diagnostic values (up to 98% sensitivity), suggesting the likely reliability of this method in smartphone apps assuming adequate image quality.21  22  23  24 Most evaluations describe dermatologists or plastic surgeons as experts, but a fully qualified dermatologist may not need to support skin apps, and there could be roles for generalist clinicians and allied health professionals, particularly if using a dermoscope attached to a phone.25 Fewer apps used automated image analysis for skin self‐examination. While this is clearly feasible,26 there is less clear evidence around its diagnostic accuracy or utility. Indeed, a recent report highlighted the diagnostic inaccuracies of four smartphone apps for melanoma detection, with wide variations in sensitivity and specificity to detect malignant skin lesions.7

Twelve apps made use of the ABCDE method27 to inform and educate users. The Doctor Mole and Skin Doctor apps also used a computerized algorithm based on the ABCDE method to help people assess the risk of melanoma directly for a specific lesion. Most apps that we identified were developed in the U.S.A. where the ABCDE method is widely used, even though there is evidence that the revised ‘Glasgow’ 7‐point checklist has higher sensitivity for diagnosing cutaneous melanoma when compared with the ABCDE method.28 Future apps could consider incorporating and validating the 7‐point checklist for melanoma detection, particularly for use in the U.K. population where it was developed and recently validated in a primary care population.29 Two apps were not included in the review as they were marketed for doctor's use only. iDoc24 Pro was the ‘professional’ version of iDoc24, aiming to ‘help clinicians assess whether to refer a patient to a dermatologist’, and Melanoma Risk Assessment Tool (Apple; last updated 2011) used the patient's personal and medical history and the result of a back and shoulders examination to calculate an estimate of a patient's absolute melanoma risk.

We found that some apps have not been updated for more than 3 years. Even when updated recently, there was no discussion around whether the updates were based on new evidence, and we were not able to distinguish whether commercial or other apps were updated most regularly. We also found surprisingly scanty scientific evidence about the development or evaluation of the smartphone apps included in our review; only one app provided a single reference to a peer‐reviewed publication. Reviews of smartphone apps for dermatology in general,7  11 and for colorectal cancer,8 have reported similar findings. There seems to be a profound mismatch between the promise of the app developers on the one hand and evidence of clinical validity and utility on the other hand, as demonstrated in this study. This is of concern for two reasons: firstly, it is important to establish the safety, including accuracy and utility, of these apps, particularly as the market is unregulated; secondly, clinicians may be benefiting financially from both the sale of the apps and the teledermatology role, and this should be disclosed.

Strengths and limitations

Previous reviews of smartphone apps focused mainly on those developed by Apple Inc.3  30 This review extends knowledge to apps for melanoma detection developed by Google. This is the first review that evaluates the content of smartphone apps available for melanoma detection for previously unaffected individuals or those previously diagnosed with a skin cancer, and generalist clinicians. It is also the first review to outline the content of available apps, whether they are derived from research evidence, and whether they have been validated.

We recognize that there are limitations to this review. It was restricted to the descriptions available from the online stores; more information may have been available after purchasing or using the apps, but this was not undertaken in order to ensure equity across the review. Moreover, we felt that users should be able to make an informed choice about the apps by reading their description in the online stores. We reviewed apps only available in English; therefore we may have failed to identify any apps published in other languages. In reality this is unlikely as the U.S.A. and U.K. were the main countries publishing these apps, with a few also coming from Western Europe and South America. Finally, we only reviewed apps applicable to Apple and Android phones as we knew that more than 90% of apps are available via these two platforms. It is likely that small platforms would use the same apps – as some were available via both of the major platforms – and we are therefore confident that few are missing.

Clinical and policy implications

A recent systematic review mapping mobile health research over the last decade identified 117 publications, but concluded that few included large or rigorous evaluation of the novel technologies.31 Another recent review has also identified patients as having a stronger desire to be informed about underpinning scientific evidence than commercial stakeholders.32 Our review has shown little evidence of clinical‐ or research‐based input; therefore, more research is needed to evaluate their content, validity and utility to support detection of melanoma in previously unaffected individuals and those previously diagnosed with a skin cancer. A particular priority should be to evaluate how skin self‐monitoring could help people appraise their skin changes and make appropriate help‐seeking decisions. Without a more explicit research base, clinicians should be cautious about supporting the use of such apps to detect melanoma earlier, although they may choose to use apps for education and/or information alone. It is noteworthy that although the U.S. Food and Drug Administration (FDA) does not currently regulate the mobile health industry, there are plans to do so in a similar way to regulation for medical devices;33 the European Medicines Agency has signalled similar intentions.34 Further evidence on clinical validity and utility are likely to be required for these regulators.

Nevertheless, these apps continue to have great potential.35 Ideally, evidence is needed on the diagnostic accuracy of these tools in the hands of intended users compared with a reference standard diagnosis. If shown to be accurate, trials of their clinical utility would be needed to assess their cost‐effectiveness in promoting earlier diagnosis and reducing consultations for benign lesions. Evidence‐based smartphone apps could then contribute towards minimizing the diagnostic delay experienced by people with melanoma, which may be significantly associated with mortality.36

Acknowledgments

We would like to thank Daniel Obute who conducted the preliminary searches.

References

Author notes

Funding sources: This report is independent research arising from a Clinician Scientist award supported by the National Institute for Health Research (RG 68235).

Conflicts of interest None declared.

The views expressed in this publication are those of the authors and not necessarily those of the NHS, the National Institute for Health Research, or the Department of Health.