Digital photography in skin cancer screening by mobile units in remote areas of Brazil

Skin cancer is the most common malignancy in many parts of the world, including Brazil [1],[2]. More than 2 million new non-melanoma skin cancer (NMSC) cases are diagnosed in the United States each year [3], and approximately 76,250 new cases of malignant melanoma were expected to be diagnosed in 2012 [4]. However, the true incidence of NMSC remains unknown because these lesions are not commonly reported to cancer registries. It has been estimated that 25% of all new cancer cases diagnosed in Brazil in 2012 will be skin cancers, with approximately 134,170 new cases of NMSC and 6,230 malignant melanomas expected to be identified. These numbers represent an approximately 16% increase in new NMSC cases and a 4% increase in melanomas compared with 4 years ago [1].

Despite a low mortality rate, NMSC can still cause severe sequelae when diagnosed at advanced stages [5] because these lesions occur predominantly on sun-exposed areas such as the face, which can become disfigured. Moreover, significant morbidity costs may occur [6]. Malignant melanoma, the third most common type of skin cancer, has more aggressive behavior and a worse prognosis, causing a significant decrease in life expectancy and lost productivity [6].

Currently, the best method for the early detection of skin cancer is to identify changes in skin lesions, including the appearance of new growths [4]. However, this strategy is not easily implemented in developing countries due to a shortage of trained professionals and their availability in remote areas. For example, in the State of São Paulo, which has the largest cancer registry in Brazil, 9% of basal cell cancer (BCC) cases, 21% of squamous cell cancer (SCC) cases and 49% of malignant melanomas are diagnosed at stages II, III or IV [7].

Teledermatology provides a new tool for monitoring skin cancer in countries such as Brazil, which has a large area and an unequal population distribution. Teledermatology essentially involves sending digital images to specialized cancer centers for evaluation by trained experts [8]-[10] and provides a platform for professional training programs and physicians to discuss complex cases without the necessity of transporting patients, which may lead to substantial savings in time and cost [11]. This approach also effectively reduces the waiting times for surgical treatment [12] and facilitates the spread of dermatological knowledge into poor regions of the world [13].

Because of the high incidence of skin cancer in Brazil, the difficulty accessing doctors in poor areas and the lack of studies on teledermatology in developing countries, we decided to utilize the existing telecommunication technology and evaluate the accuracy of digital imaging in diagnosing skin cancer in remote areas of Brazil.

The most remote areas of Brazil are located North of the Tropic of Capricorn and near the Equator—areas where solar radiation is very intense [17]. Since the 1980s, these regions have received large numbers of migrants from the Southern and Midwestern regions of Brazil, where people are predominantly of European descent. Thus, the majority of these inhabitants have light skin. Not surprisingly, this migration has increased the incidence of skin cancer in this area [1]. Despite this increase, the Brazilian government is reluctant to invest in permanent programs for skin cancer prevention because of the existing controversy over whether or not screening increases the survival of melanoma patients [18].

Developing countries exhibit an unequal distribution of doctors throughout their territories. In Brazil, the concentration of physicians is higher in metropolitan areas, and there is a huge gap in the availability of these professionals in remote areas [19]. Brazil averages one doctor for every 551 inhabitants [20], with this number varying from one doctor for every 232 inhabitants in the city of São Paulo to one doctor for every 10,000 inhabitants in some remote areas of the Amazon States and Rondônia [20],[21]. This situation is further exacerbated when medical experts, such as dermatologists, are sought. In this case, there may be one dermatologist for every 90,000 inhabitants in certain regions of Brazil [21],[22].

Given this reality, one alternative is teledermatology, which is being promoted worldwide as an effective tool for diagnosing skin cancer in remote areas [13]. This method reduces the waiting times for these patients and results in good customer satisfaction scores [23],[24], satisfactory clinical results [25] and good cost-effectiveness [26]. Despite the fact that teledermatology has been recognized as an important skin cancer screening tool in even large populations [27], few studies have investigated the use of teledermatology for diagnosing skin cancer [28],[29]. This technique could be particularly useful in developing countries [13], where skin cancer predominantly presents in the clinic at advanced stages. To our knowledge, this is the first study to use mobile units for teledermatology.

The present study was a store-and-forward teledermatology study used for skin cancer screening. The study was performed using an MPU in which a physician examined a large number of patients who were referred because of a suspected skin cancer lesion. These patients were subjected to a biopsy and/or lesion removal for histopathological confirmation. Thus, the present study represents a typical screening setting, where a positive test (classification as suspicious by digital image) was verified by the gold standard (skin biopsy). This study showed a high rate of agreement between the MPU physician’s diagnosis by visual inspection and the diagnosis given by the two oncologists using the digital images as a diagnostic tool. This result is quite impressive given the fact that oncologists #1 and #2 performed their analysis in a tertiary cancer hospital (BCH) up to 1,200 miles away from the MPU. These concordance values are even more impressive if one also considers that some of this variability may be due to differences in data interpretation between the physicians and the different experiences of the physicians involved and not due to the technology itself [13]. It seems that the physician’s experience in skin cancer screening may be more important than the technology, a finding that is not surprising. Previous reports have found that inter-observer agreement varies significantly depending on case selection, the use of classification criteria [30] and the level of expertise of the physicians [13],[28]. Thus, the high level of concordance between the oncologists in this study may also be related to the fact that they have more than 10 years of experience in skin cancer screening in Brazil.

When the diagnoses of the physicians were translated to performance indicators of digital imaging, the crude sensitivity for oncologist #1 was 89.3% and the crude sensitivity for oncologist #2 was even higher at 96.2% (Table 3). The PPV was equally high for both physicians, but the specificity and NPV were not particularly impressive. It must be emphasized that these calculations are based on incomplete evaluations [31], as biopsy verification was only performed for test-positive cases, i.e., the digital images classified as malignant. This situation occurred because it would be unethical to perform a biopsy on normal skin tissue. For any type of screening, the optimal test is the one with the highest PPV [31]. In this respect, the teledermatology setting tested here seems to be a highly suitable screening tool.

From the clinical point of view, the type of lesions missed by this screening approach is important; for example, missing a malignant melanoma is more serious than missing a BCC or SCC, which exhibit a substantially more protracted clinical course. According to our histological records, there were 5 histologically-confirmed malignant melanomas in the present series. Of those malignancies, observer 1 diagnosed 1/5 correctly, while observer 2 correctly diagnosed 3/5 cases. This result would justify using dermatoscopy to screen all pigmented lesions. However, this practice is not feasible in large countries with limited resources in rural areas, such as Brazil. A dermatoscope is an expensive piece of equipment, and the proper use of this technique necessitates the practical training of the physicians, precluding the use of this instrument in population-based screening for skin cancer.

Digital image diagnosis approach seems to be a viable option for teledermatology settings that utilize MPUs. One limitation of this study is the fact that we were not able to capture images from a camera attached to a dermatoscope, which is known to further increase diagnostic precision [32]-[34]. We only used a simple digital camera, which is easier for any health professional to handle than a more complex device such as a dermatoscope. Another limitation was the lack of clinical histories for the examined patients (work in agriculture, family history of skin cancer, etc.) and a description of the characteristics of the lesion (raised or flat, lesion size, time of development, etc.) to assist the observing physicians in making the final diagnoses.

Despite the limitations of this study, the high number of evaluated lesions and the high concordance between the observers clearly indicate that teledermatology could be used as an effective tool to screen for malignant skin lesions, especially in the remote areas of developing countries.

This study showed high agreement between direct visual inspection and digital imaging in identifying suspicious skin lesions. Moreover, digital imaging played a previously unrecognized role in predicting skin cancer. Hence, this study suggests that it is feasible to use digital imaging as a tool to screen for skin cancer in a population-based setting and that this approach would be particularly useful in remote areas.