What is it about?
he article discusses the use of mobile technology, specifically an on-device inference app, for detecting skin cancer through machine learning. Mobile health (mHealth) is seen as a powerful tool for delivering medical applications. Machine learning has been successful in identifying diseases through medical images, but it usually requires a large amount of data for training. Storing and processing such data can be challenging for mobile applications. To address this, some studies propose using cloud-based machine learning, but this comes with issues like latency and privacy concerns. To overcome these challenges, the article introduces an on-device inference app. The app is designed to run machine learning models directly on the mobile device, eliminating the need for cloud-based processing. The authors demonstrate the concept using a dataset of skin cancer images. They pre-train a model using a substantial number of images and then deploy this model on a mobile device. When the app encounters a new skin cancer image, all computations happen locally on the device, reducing delays, saving bandwidth, and enhancing privacy. In summary, the article explores a practical approach to bring machine learning for skin cancer detection directly to mobile devices, addressing issues related to data processing and privacy.
Featured Image
Photo by Matthew Tkocz on Unsplash
Why is it important?
The on-device inference app for skin cancer detection using machine learning on mobile devices is important for several reasons: (1) Reduced Latency: Processing data locally on the mobile device minimizes the time it takes to analyze and provide results. This is crucial in medical applications where quick and timely diagnosis can be critical for patient outcomes. (2) Bandwidth Savings: By performing computations locally on the device, the need for transmitting large amounts of data to a cloud server is minimized. This can result in significant bandwidth savings, making the application more feasible in areas with limited or expensive internet access. (3) Improved Privacy: Local processing of data on the mobile device enhances privacy. Since sensitive medical information is not sent to external servers, there is a reduced risk of unauthorized access or breaches, addressing concerns related to patient data confidentiality. (4) Accessibility: Mobile devices are ubiquitous, even in remote or resource-constrained areas. The ability to perform skin cancer detection on mobile devices makes such diagnostic tools more accessible to a wider population, potentially improving healthcare in diverse settings. (5) Real-time Analysis: On-device processing allows for real-time analysis of medical images. This is valuable for healthcare professionals who may need immediate insights during patient examinations. (6) Resource Efficiency: Cloud-based machine learning often requires substantial computing resources, and it might not be practical for resource-constrained mobile devices. On-device processing optimizes resource utilization and makes the application more feasible on a wider range of devices.
Perspectives
here are some perspectives on the importance of the on-device inference app for skin cancer detection using machine learning on mobile devices: (1) Patient Empowerment: The on-device app empowers patients by providing a tool for preliminary skin cancer detection on their own mobile devices. This can lead to increased awareness, early detection, and proactive healthcare management. (2) Global Healthcare Accessibility: In regions with limited access to advanced medical facilities, the on-device app can serve as a cost-effective and accessible solution for skin cancer screening. This has the potential to make healthcare services more widely available globally. (3) Cost-Effectiveness: Utilizing on-device processing reduces the dependency on expensive cloud-based resources. This cost-effectiveness is essential for implementing healthcare solutions, especially in environments with budget constraints. (4) Integration with Telemedicine: The app aligns well with the growing trend of telemedicine. Patients can use the app for initial assessments, and healthcare professionals can then follow up with more targeted evaluations, potentially streamlining the diagnostic process. (5) Research and Data Collection: The deployment of machine learning models on mobile devices allows for the collection of valuable data directly from users. Researchers can analyze this data to gain insights into skin cancer patterns, contributing to the broader understanding of the disease. (6) Adaptability to Diverse Settings: Mobile devices are versatile and adaptable to different environments, including rural and remote areas. The on-device app's ability to function independently of constant internet connectivity makes it suitable for a wide range of healthcare scenarios. (7) Ethical Considerations: As the app processes data locally, it addresses ethical concerns related to data privacy and security. This is particularly relevant in healthcare, where maintaining patient confidentiality is of utmost importance. (8) Educational Tool for Healthcare Professionals: The on-device app can serve as an educational tool for healthcare professionals, aiding in the continuous learning and development of skills related to skin cancer diagnosis. It complements traditional medical education with technological advancements. (9) Scalability and Integration: The app's on-device approach allows for scalability without overburdening central servers. It can be integrated into existing healthcare systems and scaled up to accommodate a growing user base. (10) Potential for Other Medical Applications: The concept of on-device machine learning has broader implications beyond skin cancer detection. It sets a precedent for developing similar applications for other medical conditions, potentially revolutionizing how various diseases are diagnosed and managed using mobile technology.
Dr. HDR. Frederic ANDRES, IEEE Senior Member, IEEE CertifAIEd Authorized Lead Assessor (Affective Computing)
National Institute of Informatics
Read the Original
This page is a summary of: Machine Learning on Mobile: An On-device Inference App for Skin Cancer Detection, June 2019, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.1109/fmec.2019.8795362.
You can read the full text:
Contributors
The following have contributed to this page
