One of the biggest cancer challenges in low- and middle-income countries (LMICs) is the lack of access to accurate and affordable cellular and molecular diagnostics, which are essential for making informed therapeutic decisions, in particular for breast cancer. With the increased use of low cost ultrasound, it has become possible to readily sample suspicious breast lesions with ?ne needles (?ne needle aspirates, FNA). However, the workup of such specimens is often impossible in many LMIC settings. To address these barriers to diagnosis, Aikili?derived from A.I. and Akili (intelligence in Kiswahili)?seeks to enable the same-day diagnosis of breast cancer at the point-of-care using a low-cost, automated system. The Aikili system is a highly advanced stand-alone diagnostic platform capable of automated cancer diagnosis and receptor sub- typing in near real-time (< 1 hour), at a low cost (<$800 for integrated hardware and $5-10 per test). Building upon our initial development and successful clinical validation of human samples, the goal of this Phase I application is to advance the Aikili technology to signi?cantly improve its usability in resource-limited settings. Speci?cally, we propose to i) upgrade Aikili technology by incorporating a custom-designed disposable cartridge for onsite sample processing and deep learning algorithms for automatic analysis (Aim 1), and ii) evaluate the performance of the upgraded system in LMIC work?ows through a validation study in Kenya (n = 30) (Aim 2). We will consider the Phase I project successful when we can show that the ?eld-optimized Aikili system accurately and reliably detects breast cancer and receptor status in human FNAs compared to accepted gold standards. Successful completion of Phase I would lead to a Phase II application for scale-up of manufacturing and a larger, multi-site clinical validation study. This platform may alter therapeutic paradigms for breast cancer patients in globally and enable appropriate use of chemotherapies and anti-estrogens in limited supply.
We will develop a standalone, AI-powered diagnostic system to enable early cancer detection in low resource settings. Our system has the potential to transform cancer diagnostics in low- and middle-income countries where cytopathology is a major bottleneck; it will augment workflows by enabling non-expert healthcare workers to rapidly establish cancer diagnoses and identify molecular subtypes within 1 hour. This will inform the most appropriate therapeutic choices in collaboration with specialist physicians, as well as reduce patient loss to follow-up (LTF) by providing diagnosis in near real-time.