1. Get disease names (with probabilities) given symptoms
2. Suggest symptoms (correct names, related symptoms)
3. Get information on diseases, medicines
This repo acts as core part of SparrowAI. The API is triggered by IBM Watson Assistant via Sparrow middleware.
Quick setup
#Command to directly pull image from dockerhub and run
docker run -d -p 5000:5000 jaylohokare/diseases-predictor:0.5
#Download the docker compose file and run
docker-compose up
The API will be live at 192.168.99.100:5000 (Windows) 127.0.0.1:5000 (Linux/Mac)
To build using dockerfile:
-
Download database umls.db and place it in umls/databases - https://drive.google.com/file/d/1BmnzCv2hjU49vXxBv_kI4FUae-lIAdz0/view?usp=sharing
-
Download 'data' folder and place it at root location - https://drive.google.com/file/d/1nwD3aEMwW5CF8CWOscOT3IMztuNBpaYq/view?usp=sharing
-
Docker build command -
docker build -m 25g -t medicalapi .
To run after building -
docker run -d -p 5000:5000 medicalapi
Other docker useful commands-
#Clear all docker images cache
docker system prune -a
#Get all docker instances
docker container ls
#List all containers (only IDs)
docker ps -aq
#Stop all running containers
docker stop $(docker ps -aq)
#Remove all containers
docker rm $(docker ps -aq)
#Remove all images
docker rmi $(docker images -q)
#To push to docker hub
docker images
docker tag ID jaylohokare/diseases-predictor:0.3
docker push jaylohokare/diseases-predictor
UMLS databases are based on CUIs for medical lexicons. We have built couple of APIs to convert plain text symptoms to CUIs and then use these CUIs to predict diseases.
/api/validateSymptom
Input:
Takes symptom name (Need not be perfect match to content in DB)
Example - {"name" : string_SymptomName}
Output:
Returns CUI if entered name is exact match to DB content, or returns list of symptoms that are closest match and their CUIs
Example - {"CUI1": string_Symptom1, "CUI2": string_Symptom2....}
/api/getDetails
Input:
Takes symptom CUI
Example - {"code" : string_CUI}
Output:
Returns description from UMLS
/api/sySuggest
Input:
Takes symptom CUI
Example - {"symptom" : string_CUI}
Output:
Returns list of relevant symptoms
Example - [{"label": string_SymptomName, "value": string_SymptomCUI}, {"label": string_SymptomName, "value": string_SymptomCUI}, ....]
/api/predict
Input:
Takes list of symptoms CUIs as input
Example - {"symptoms" :[string_SymptomCUI1, string_SymptomCUI2, ...]}
Output:
Returns list of diseases with their CUIs, probabilities, related symptoms
Example - [{"disease": string_DiseaseName, "disease_cui": string_DiseaseCUI, "prob": float_Probability, "sy": arrayOfStrings_relatedSymptoms}, {}...]
/api/predictFromString
Input:
Takes | delimited string of symptom names as input
Example - {"symptoms" :"symptom1|sympto2|"}
Output:
Returns list of diseases with their CUIs, probabilities, related symptoms
Example - [{"disease": string_DiseaseName, "disease_cui": string_DiseaseCUI, "prob": float_Probability, "sy": arrayOfStrings_relatedSymptoms}, {}...]