TCMPG Platform

Platform Instructions:

Page structure

This page mainly contains four parts:
  1. Platform instructions for usage;
  2. Symptom label selection box: you can choose symptoms according to different categories;
  3. Symptom input box: it can display the symptoms you input manually or selected through the symptom label selection box;
  4. Reference prescription display box: Calculate the symptom set you input by using the best trained model, and display the Top 10(default) herbs and the corresponding dose ratios.

How to use

You can use the platform in the following two ways:
  1. You can manually input symptoms set in the symptom input box, where symptoms are separated by a space;
  2. You can select symptoms in the symptom label selection box, and then compose your symptoms set by clicking the Add button, only one symptom label can be added at a time.

Buttons Introduction

  1. Add: Add the symptoms in the symptom label selection box to the symptom input box, and automatically display them in a standard format;
  2. Submit: Transfer the symptom set in the symptom input box to the backend model for calculation. After clicking submit, the waiting information will be displayed in the symptom input box;
  3. Example: View the standard format in the symptom input box and the corresponding results in the reference prescription display box;
  4. Delete: Clear the content of the symptom input box.

Notice

The error message will show in the reference prescription display box if the following situations occur.
  1. Symptoms are not added in the standard format in the symptom input box;
  2. There are some symptoms do not belong to this dataset in the symptom input box.

Select The Symptoms

Input The Symptoms

Display The Reference Prescription

Select the number of herbs:

Introduction of TCMPG platfrom

Introduction

TCMPG platform is a useful tool to assist doctors in diagnosis and treatment and generate effective reference prescriptions. You can firstly inout your symptoms set. Then, the input will be transferred to the back-end model to generate the reference prescription for you.

Structure

The platform mainly includes the following five parts:
  1. Menu: The menu bar at the top of the web page can provide path guidance. Click "Home", "Prescription Generation", "Introduction", and "Dataset" to display different interfaces;
  2. Home: Entering the website will display our homepage content, you can click the "Enter Platform" button to enter the "Prescription Generation" interface;
  3. Prescription Generation: This interface is the core part of the platform. Before using this platform, you can read the platform instructions carefully, so as to better use our platform;
  4. Introduction: Reading this interface, you can have a more comprehensive understanding of the TCMPG platform;
  5. Dataset: This interface mainly introduces the herbs and symptoms obtained through data processing, and how we process the data.

Dataset

Through more than a year of hard work, we continuously processed 49493 prescriptions, including 2284 herbs and 2509 symptoms, through computer and manual operations such as data screening and cleaning, and format standardization. These prescriptions are derived from 1459 ancient medical books, including "Sheng Ji Zong Lu", "Sheng Hui", "Pu Ji Fang" and so on.
Tips: For more information about datasets, click on the datasets section of the menu bar.


Disclaimer

  1. TCMPG platform is provided free-of-charge in the hope that it will be useful for your research.
  2. TCMPG platform will provides users with reference prescriptions. We bear no responsibility for the reference prescriptions recommended by our platform.

Dataset

Herbs

For more than one year, we processed 84463 prescriptions by computer-aided and manual processing. Through a comprehensive analysis of prescriptions, we have found many situations, such as: incorrect and incomplete herbs, unidentified herbs, herbs without doses and so on. We screened out the prescriptions with the above conditions.

Through several screening processes, we finally retained 49493 prescriptions as the main dataset for experimental study. We have separated a total of 2284 different herbs. All the herbs in the dataset can be viewed in the selection box below.

You can browse herbs from the drop-down box. The first drop-down box is the first letter of pinyin, and the corresponding herbs will show in the second drop-down box.

TCM Symptoms and Diseases

Next, we manually label the indication descriptions of 49493 prescriptions. However, the indication descriptions are unstructured, which cannot be directly used to build model and conduct research.

Through manually processing, we marked the unstructured indication descriptions as symptom labels. You can view the processing results through the example below.

In the process of labeling, there are some labels with similar meanings but different expressions. We represent them with a unified label. Finally, we obtained 2509 kinds of symptoms and diseases.

Example:
Indication description: Postpartum pus and blood dysentery persists, intra-abdominal pain, unwillingness to eat, and gradually become weak.
Labels: Postpartum, pus and blood dysentery, abdominal pain, lack of appetite, weakness.

Copyright © 2022 College of Intelligence and Computing, Tianjin University All rights reserved.