probabilistic-symptom-graph 1.0.3

Using a bipartite graph of medical conditions and symptoms with relative weighted edges, gives the estimated probability of having each illness given a list of symptoms

Probabilistic Symptom Graph

pip install probabilistic-symptom-graph

from src.ProbabilisticSymptomGraph import ProbabilisticSymptomConditionGraph

import networkx as nx
import numpy as np

medical_condition_gexf = "./data/medical-condition-symptom-graph.gexf"
graph = nx.read_gexf(medical_condition_gexf)
sim_matrix = np.load('./data/md-symptom-sim-mat.npy')

symptom_names = []
for node_id, tpe in graph.nodes(data="type"):
    if tpe == "Symptom":
        symptom_names.append(node_id)
symptom_names = sorted(symptom_names)

condition_names = []
for node_id, tpe in graph.nodes(data="type"):
    if tpe == "MedicalCondition":
        condition_names.append(node_id)
condition_names = sorted(condition_names)

probabilistic_graph = ProbabilisticSymptomConditionGraph(condition_names, symptom_names, graph, sim_matrix)
print(" | ".join(probabilistic_graph.get_all_symptoms()[:10]))
probabilistic_graph.get_condition_probs(["acne"])[:5]

Graph Image

Red nodes represent medical conditions and green nodes represent symptoms. This graph data was gathered by mining Wikipedia "Medical Condition (New)" infoboxes and extracting medical condition <=> Symptom pairs. Post processing was done using ChatGPT.

Data Mining

See ./data/WikipediaSymptomExtractor.ipynb