Abstract
Cervical cancer remains a leading cause of mortality in its invasive stages, presenting a significant global public health challenge, particularly in low- and middle-income countries. Despite technological advancements that have improved the quality of cervical images captured during visual inspections, several challenges persist. This article presents key findings from the CITOBOT-COL translational research project, a large-scale initiative focused on designing CITOBOT as a portable cervical cancer screening device. We detail the comprehensive technological development of CITOBOT, guided by a human-centered design approach, smart prototyping, and the integration of AI. Over four design iterations, we developed and refined CITOBOT v4, a portable device. Prototypes were validated through focus groups and testing by experts in cervical cancer prevention, gynecology, nursing, software, artificial intelligence, computer engineering, and public health, utilizing various anatomical models at the Simulated Hospital Laboratory of Pontificia Universidad Javeriana Cali, Colombia. Additionally, we developed AI algorithms using the Inception V3 network, optimized with Transfer Learning and Fine Tuning, for cervical image classification and offline-operating software that guides the physician through the examination and provides a risk assessment for cervical cancer. Feedback was crucial in assessing and refining the device's functionality, focusing on capturing high-quality cervical images. The development of CITOBOT v4 highlights the importance of fostering innovation in resource-limited settings, offering an effective solution to improve cervical cancer screening and potentially save lives in vulnerable communities.