Learn how easy it is to sync an existing GitHub or Google Code repo to a SourceForge project! See Demo


PYKE and Certainty factors

  • Inayath

    Dear Bruce,

    Thanks for inviting me to this forum. Certainty factors play an important role in medical diagnosis. Although we doctors do not explicitly do calculations to come to a diagnosis we however subconsciously do them. I consider the real time doctors diagnosis system akin to artificial neural network system wherein different weights and values are alloted to each fact, the backpropagating learning algorithm based on his experience alter the weights of the nodes/facts. I will updating this page http://naimath.sourceforge.net/?q=content/how-do-we-diagnose very soon will the whole theory and discuss in depth all the steps.

    Here's a rough example of certainty factors for chest pain -

    Following questions are asked and their answers asserted as facts (in brackets):

    1. What is the duration of chest pain? (int)
    2. How is the onset of chest pain? ( Sudden or gradual)
    3. Does the pain aggravate on exertion? (Yes - exertion pain, No)
    4. What is the character of the chest pain? (Heaviness/Band like, Burning, Sharp)
    5. Does the pain radiate to left arm/neck? (Yes - radiates, No)
    6. Is the pain associated with meal intake? (Yes - meals, No)
    7. What is the severity of chest pain? (1,2,3,4)
    8. Is the pain relieved by sublingual nitrates? (Yes, No)

    These are just a few questions, based on which we can come to a diagnosis -

    sudden onset      0.6           0.2               0.3
    gradual onset      0.3           0.5               0.3
    on exertion yes    0.8           -0.8              0.3
    on exertion no     -0.4          0.4               0.4
    c - tightness        0.8           0.4               0.2
    c- burning           -0.4          0.6               -0.4
    c- sharp               0.3           0.1               0.5
    radiation left arm 0.8           -0.8              0.3
    ass with meals     0.1           0.7               -0.2
    severity 3,4         0.5           0.4                0.2

    Based on this findings we come to a diagnosis by adding up all the values. However, how these values are choosen is debatable.

    Once the disease is triggered it can further trigger other diseases like for eg. angina + syncope + dyspnea = Aortic Stenosis, angina > 20 min = infarction, etc

    But for every disease there is some confirmation method for eg. in the above table if all the findings from history make angina a likely diagnosis we confirm it by getting an EKG whose certainity factor is much high (+100 or -100).

    I am still not sure about how much we should rely on CF, and i need to consult a few expert doctors and see if i can come up with something more concrete.

    Many thanks for your interest in MES, i will be posting the entire plan by this weekend.

    Thank you.