Parasite Classification in Biomedical Imaging Utilizing Vision Mamba

Parasites can lead to a variety of illnesses and may progress into chronic conditions with long-term health implications. Early diagnosis remains the cornerstone of effective disease prevention. One of the diagnostic method is performed by examining microscopic images of various samples collected from individuals. Artificial intelligence plays a crucial role in preventing misdiagnosis and misclassification in this context. We employed Vision Mamba, a state-of-the-art backbone, on an eight-class dataset and conducted a comparative analysis against other current state-of-the-art architectures and models. Three variants of the Vision Mamba model were evaluated, and the best performance was achieved with the Vim-Base variant. The weighted precision, recall, and accuracy scores all reached $\mathbf{9 9. 8 5 \%}$, reflecting an exceptional model performance. These results demonstrate that thanks to the strong architecture and hardware-aware design of the Vision Mamba model, it remains highly competitive and outperforms other models on the same dataset.