Non-Invasive Anemia Detection using Machine Learning: A Comparative Analysis | ||||
| Advanced Sciences and Technology Journal | ||||
| Articles in Press, Accepted Manuscript, Available Online from 22 August 2025 | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/astj.2025.390972.1064 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Rasha S. Aboul-Yazeed  ; Mazen Mohamed; Reen Salama; Mahmoud Ahmed
| ||||
| Software Engineering and Information Technology Department, Faculty of Engineering and Technology, Egyptian Chinese University, Cairo, Egypt | ||||
| Abstract | ||||
| Anemia is a common disease affecting over 33% of the world’s population, where 42% of children under the age of six and 40% of pregnant women worldwide are anemic if neglected it can have serious impacts like slow child development, delayed mental and psychological growth, impaired performance at work, and increased vulnerability to infections. Various reasons including poor dietary plans, chronic diseases, or genetic disorders can cause anemia. Traditional detection methods, although reliable, are costly and time-consuming. The proposed paper presents a comparative analysis of three different machine learning models for the non-invasive detection of anemia. Support vector machines (SVMs), k-nearest neighbor (k-NN), and random forest (RF) are applied to three distinct regions of interest (ROI); the conjunctiva of the eye, palpable palm, and fingernails; to determine the model that could achieve the highest anemia detection accuracy per ROI after examining several preprocessing operations such as data cleaning, feature selection and augmentation. Moreover, two different training-testing data split ratios are explored, that are 70-30 and 80-20. Analysis results show that the palpable palm is the optimal region for non-invasive screening of anemia, and k-NN yields the highest accuracy across all regions of 99.88% using an 80-20 data split ratio, exceeding the state-of-the-art. The proposed non-invasive technique for anemia screening and early diagnosis could provide a reliable alternative for quick, low-cost anemia screening, particularly in the countryside and rural areas. | ||||
| Keywords | ||||
| Anemia; Non-Invasive Detection; Eye Conjunctiva; Hand Palm; Fingernails | ||||
|
Statistics Article View: 3 |
||||
