Design of a quantitative readout in a point of care device for cisplatin detection

This letter presents a low-cost, lightweight, and automatic readout system for accurate quantification of cisplatin (Cis-Pt) level in a colorimetric POC device. The system relies on an RGB camera to record the color changes in the test tube and an inference function trained via machine learning techniques to estimate Cis-Pt concentration. A suitable training strategy for ensemble learning is adopted to deal with a scenario involving a small training set. In addition, the training strategy is designed to support a gray-box model, thus enabling interpretability. Experimental results show that the readout system based on the proposed strategy is able to outperform the state-of-the-art readout system for the quantification of Cis-Pt level in a colorimetric POC device.