Authors

AUTHOR 1: ESRAA HASSAN,

AUTHOR 2: SHAZA BAHNACY

AUTHOR 3: HABIBA TAREK

AUTHOR 4: MAI HAZEM

SUPERVISED BY: DR. WALAA HASSAN EL-ASHMAWIE AND ENG. LOBNA SHAHEEN.

Publishing Date

26TH OCTOBER 2020.

Abstract

At least 1.5 million people each year are being killed or furtherly sickened due to reading the medical prescription incorrectly in accordance with the National Academy of Sciences. This proposed system presents a solution to this issue through a software application that recognizes handwritten medicine names in English language and returns a readable digital text. This will help in diminishing the instances of distortion of medication names and will assist drug specialists with limiting their doubts when selling to the patient , the medicine. The proposed system shall be able to scan the prescription, recognize it then return in the form of a digital text and checks the availability of the medicine as it will be connected to the pharmacy’s system.


1.1 Purpose of this document

The main purpose of this document is to outline the description of the medical prescription recognition system. The system aims read medical prescriptions and return readable digital text for the prescribed medicines. Furthermore, the document will present the details of the system’s functions and features. This document also shows an explicit user interface and how the system will deal with the user interactions. This Software Requirements Specification Document (SRS) is needed for the graduation project at Misr International University (MIU).

1.2 Scope of this document

This document provides detailed information about the system as the system overview, system scope , business context and the similar systems to our proposed systems, also this document provides detailed description of the system and it’s users as user problem statement, user objectives, user characteristics , system context, also it provides detailed description about the functional requirements of the system and non-functional requirements as the APIs used and design constraints also the document shows the user interface design of the system in addition to the project plan that shows the start and end dates of a several elements of the project.

1.3 System Overview

The first stage, the pre-processing phase consists of:

• Image Normalization [1]: the size of the image is normalized by cropping white spaces and converting the image into black and white.

• Morphological operation [2]: the morphological operation technique applied on the image to make all the images of the same size based on a comparison of the corresponding pixel in the input image with its neighbors.

• Image cropping: the crop operation is applied to crop the image into 3- parts.

– First part: from the beginning of the prescription till (R/) symbol which includes the name of the doctor.

The second stage, the processing phase consists of:

• Feature extraction [3]: in which the most important features in an image is extracted, it consists of:

– Convolution layer [4]: it includes the input image, a feature detector and a feature map, then the filter is taken and applied pixel block by pixel block through the multiplication matrix to the preprocessed middle image so the feature map is filled or completed. Many feature maps are created to get our first Convolutional layer.

– ReLU Layer [5]: the Rectified Linear Unit (ReLU layer)is another step to the

Convolution layer as an activation function is applied to the feature map to increase the non-linearity in the network.

– MaxPooling Layer [6]: in order to achieve spatial variance, we use the max pooling technique to gradually reduce the input representation size as it makes it easier to detect and identify objects wherever they are located inside the image.

– Flattening: he pooled feature map is flattened into a sequential long vector to allow the information to enter the input layer in the ANN [7] to be furtherly processed.

• Classification in which data is trained and tested to be output:

– Fully connected layer [8]: it uses the softmax activation function to get probabilities of the input image being classified into a particular medicine class.

3. The third stage, the post processing phase consists of:

• Optical Character Recognition (OCR) [9] technique is applied on the resulted medicines if accuracy 50% or less to process character by character and comparing the OCR result with a data set contains all the medicine names to recognize which medicine in the dataset nearest to the result.

1.4 System Scope

Medical prescriptions should be scanned by a camera phone with high camera resolution, photos should be taken from an appropriate and zoomed in angle, to ensure that the scanned image is clear and it’s content is visible.