Performance Evaluation of the Pallor in the Diagnosis of Anemia

Abstract

With minimal resources and trained labour, digital image processing techniques can be used for diagnostic support system in illness identification at an early stage. These methods can also aid doctors during clinical evaluations by removing the need for invasive pathological investigations. By just viewing the colour content of the image of the pallor site, various blood components such as hemoglobin and billirubin can be analysed and classed in terms of the colour properties of the image. This research effort is an attempt to propose an image based method of quantifying hemoglobin(Hb) by examining the brightness, red component and texture of the patient's palpebral conjunctiva image after image super resolution. According to WHO recommendations, a computational model based on network of neurons is employed to relate the hemoglobin level to be measured with the level measured by the conventional invasive technique. Furthermore, based on the models’ testing results, patients with Hb concentrations below 11g/dL were identified as anemic. The predicted categorization findings were then compared to the actual Hb value acquired from an invasive test using a confusion matrix. The proposed algorithm is shown to have a sensitivity of 91 percent and a specificity of 99 percent in diagnosing anemia. The proposed approach is useful not only for detecting severe but moderate anemia too and will be valuable in assisting medical practitioners in making accurate anemia diagnosis in clinics with limited available resources for healthcare using non invasive procedure.