We suggest a low cost, non invasive healthcare system that measures haemoglobin levels in patients and can be used as a preliminary diagnostic test for anaemia. A combination of image processing, machine learning and deep learning techniques are employed to develop predictive models to measure haemoglobin levels. This is achieved through the color analysis of the fingernail beds, palpebral conjunctiva and tongue of the patients. This predictive model is then encapsulated in a healthcare application. This application expedites data collection and facilitates active learning of the model. It also incorporates personalized calibration of the model for each patient, assisting in the continual monitoring of the haemoglobin levels of the patient. Upon validating this framework using data, it can serve as a highly accurate preliminary diagnostic test for anaemia.