Machine learning systems (ML) are reshaping healthcare due to their efficiency and accuracy, but their diagnostic and prognostic powers are undermined by persistent racial and gender bias. For example, some breast cancer tools have been shown to disadvantage black women. I situate issues of bias and discrimination in the above tools as a value and virtues problem, rather than a technical one. Afterwards, I claim that current existing principles in bioethics, like autonomy, beneficence, justice, and non-maleficence, which currently guide how these systems are designed and deployed, cannot adequately shape the moral virtues of designers and the designing environment to mitigate issues above, since they are rooted in Western rights- and utility-based traditions. Finally, the paper then offers a novel intervention by showing why my prized Afro-relational moral principles, care, respect for human integrity, and dignity, are better suited to address bias in ML systems used in healthcare for diagnosis and prognosis. Beyond their African origins, these principles offer globally relevant ethical resources for designing more value-laden, virtues-centred and inclusive technologies. I argue that reorienting bioethics around these values marks a necessary shift in both theory and practice.