Hemolysis is a challenging problem and still represents a frequent source of errors in blood test laboratory practice. Due to the broad and heterogeneous bias induced in the measurement of several parameters by hemolysis, inaccurate results may be reported, and the patient may be required to repeat sample collection, delaying diagnosis. Existing automated laboratory devices including hemolysis detection are not suitable for lower volume and smaller sample collection sites. In many situations, hemolysis is still detected by visual inspection of the sample after centrifugation, during the blood test pre-analytical stage. Visual inspection is highly dependent on a qualified workforce, subjective to interpretation discrepancies, and thus difficult to standardize. The paper aims to describe the design and performance of a portable device for measuring hemolyzed samples based on computer vision and neural network. The results indicate that the device provides hemolysis indexes with sufficient accuracy to guide laboratory decision in the blood test pre-analytical stage.