Medical Handwritten Prescription Recognition and Information Retrieval using Neural Network

The automatic interpretation of handwritten documents is one of the popular research areas over the last few decades because of the huge scope of its practical applications such as automatic reading of address, bank cheque processing, and hand written text recognition filled on forms. Moreover, information retrieval in offline doctor’s prescription images was not being focused prior the Covid - 19 pandemic. However, over the last two years such prescription images are extensively being exchanged by the patients’ among offline medical consultants for useful advice. Therefore, in this paper the potential concern has been expressed on using computer technology to assess handwriting. Character recognition from the connected alphabets of a word (cursive writing) is a real time challenge. For this, the usage of Extended MNIST has been explored and the results support the efficiency of proposed model to identify the poor legibility of handwriting and transform it into readable correct text recognition. Further, using proposed model of the system the handwritten medical prescription can be converted digitally using electronic writing pad. Such facility will enables patient to take away prescription in the form of digital media which can further be recognized by running it on our model. The application areas wherein the proposed character recognition system can be utilized are recognizing medicine names from doctor’s prescription, historical document recognition, automatic reading of bank’s cheque, automatic postal code identification, converting handwriting in real time, extracting data from filled-in forms etc.