AICOM-MP: an AI-based Monkeypox Detector for Resource-Constrained Environments

Abstract: Under the Autonomous Mobile Clinics (AMCs) initiative, we are developing, open sourcing, and standardizing health AI technologies to enable healthcare access in least developed countries (LDCs). We deem AMCs as the next generation of health care delivery platforms, whereas health AI engines are applications on these platforms, similar to how various applications expand the usage scenarios of smart phones. Facing the recent global monkeypox outbreak, in this article, we introduce AICOM-MP, an AI-based monkeypox detector specially aiming for handling images taken from resource-constrained devices. Compared to existing AI-based monkeypox detectors, AICOM-MP has achieved state-of-the-art (SOTA) performance. We have hosted AICOM-MP as a web service to allow universal access to monkeypox screening technology. We have also open sourced both the source code and the dataset of AICOM-MP to allow health AI professionals to integrate AICOM-MP into their services. Also, through the AICOM-MP project, we have generalized a methodology of developing health AI technologies for AMCs to allow universal access even in resource-constrained environments.

• The paper presents AICOM-MP as a novel AI-based monkeypox detection system tailored for low-resource settings, achieving state-of-the-art accuracy.
• Methodology combines multi-layer attention with EfficientNetB7 and resolution restoration to process low-resolution images effectively on mobile devices.
• Implications include enhanced healthcare delivery in Least Developed Countries and a replicable framework for future AI-driven diagnostic tools.

An Expert Overview of "AICOM-MP: an AI-based Monkeypox Detector for Resource-Constrained Environments"

The paper "AICOM-MP: an AI-based Monkeypox Detector for Resource-Constrained Environments" presents an innovative approach toward improving healthcare accessibility in resource-constrained regions, specifically targeting the ongoing global monkeypox outbreak. This research proposes a novel AI-based detection system, AICOM-MP, designed to function effectively on low-end devices, primarily aiming to be integrated into Autonomous Mobile Clinics (AMCs).

Objectives and Methodology

The primary objective of this study is to enhance healthcare delivery, particularly in Least Developed Countries (LDCs) where healthcare resources and infrastructures are limited. The paper delineates the concept of AMCs as mobile healthcare units equipped with diagnostic tools and AI capabilities for disease screening. AICOM-MP, as a part of these AMCs, is tailored to process images captured via resource-constrained devices, offering an accessible barrier-free solution to monkeypox screening. The development and deployment of these AI engines are compared to smartphone applications, dictating that they adapt to varying usage scenarios dictated by the operational environment.

From a methodological perspective, AICOM-MP is evaluated against existing AI models, showing a notable improvement in detection accuracy. The development strategy for AICOM-MP is twofold: dataset formulation and architectural innovation. The dataset construction emphasizes coverage and diversity, ensuring a robust training set that counters potential biases regarding age, race, and gender. Architecturally, AICOM-MP employs a multi-layer attention model that mimics medical professionals' diagnostic processes, effectively parsing and focusing on medically pertinent image regions. Additionally, the framework incorporates state-of-the-art deep learning models, including EfficientNetB7, optimized for computational efficiency suitable for mobile CPUs.

Evaluation and Results

Markedly, AICOM-MP achieved superior detection results, evidencing state-of-the-art (SOTA) performance across multiple datasets and real-world scenarios. Data processing and model training enhancements, such as the inclusion of resolution restoration units, enable the system to excel in environments with limited data fidelity or adversely affected by low-resolution imaging. The empirical evaluation demonstrated that AICOM-MP surpassed existing AI models, achieving up to 100% accuracy on specific datasets and yielding high generalization capability, particularly on datasets mirroring real-world complexity.

Theoretical and Practical Implications

The implications of this work are manifold. Theoretically, it offers a replicable model for the development of AI-based healthcare solutions adaptable to diverse diseases beyond monkeypox. The research contributes to the broader discourse on accessible healthcare AI deployment, underscoring methodologies that enhance model robustness and bias mitigation during dataset formation. Practically, AICOM-MP and similar models hold the potential to revolutionize medical diagnostics in LDCs, circumventing the traditional healthcare infrastructure limitations by leveraging mobile technology and autonomous units.

Future Directions

The paper indicates prospective advancements towards facilitating offline functionalities for its AI models, especially to cater to LDC populations with limited internet access. The migration of AICOM-MP onto widely-used, low-end mobile platforms is a crucial next step. As the AMCs initiative evolves, there is potential for broadening the spectrum of diseases addressed by health AI engines, informed by data-driven modifications and grounded in the established methodological framework. Such progress could lead to significant strides in achieving universal health coverage as stipulated by the UN’s SDG3 goal.

In conclusion, the development of the AICOM-MP system represents a significant advance in leveraging AI for equitable healthcare. By focusing on accessibility and resource-conscious design, this paper paves the way for future innovations in mobile and telemedicine infrastructure, specifically within the confines of economically challenged constraints.