Artificial intelligence powered mobile health apps for skin cancer detection: current challenges and a systems thinking approach for improved public health outcomes in low- and middle-income countries

Comprehensive Summary

Mukherjee discusses an emerging development in diagnosis of skin lesions, requiring only a smartphone camera. The apps have the ability to analyze abnormalities to gauge risk and propose recommendations for care. In this article, the benefits and risks of releasing such apps are discussed. Mukherjee uses a causal loop diagram, or CLD, to analyze the following four variables: data acquisition and quality, AI model development and validation, user experience and accessibility, and public health impact. These four pillars, as Mukherjee names them, are used to weigh the overall impact of AI powered apps in skin cancer diagnosis, including both positive and negative effects within each. Mukherjee recommends that clinicians, AI developers, and other concerned experts convene to discuss the development of such technology to mitigate possible consequences of such models becoming widely available.

Outcomes and Implications

These developments could be particularly beneficial to low and middle income countries and households. Requiring only a smartphone camera, rather than additional medical equipment, this method lacks the high costs of most new diagnostic technology, a cost which often prevents communities from utilizing new developments. However, the smartphone app could potentially lead patients to negate the necessity of a physician altogether, relying solely on the AI model for treatment recommendations. Should overdiagnosis occur as result of the apps, dermatology and pathology resources in low and middle income countries are not sufficient to handle the influx of patients, many of which would be false diagnoses. Another concern for implementation is demographic limitations as to how accessible the app will be.