The evolution of medical diagnosis
Data science has become the most favourite career choice today. It attracts individuals from diverse backgrounds and has found applications in every aspect of our day to day life. The way we use internet today, has flooded the world with data and this helps in designing solutions for various businesses problems.
Healthcare stands as one of the best examples of its applications. Most of the diagnosis to identify a pathological condition is computer assisted and is carried out with the help of models generated using huge samples of data. Many sophisticated classification models take into account the age and several vital signs of the patient. This goes a long way in minimising the effects of subjective analysis by individual doctors.
From diversity perspective, this needs doctors and engineers to work hand in hand. To get good quality data with very less noise, it is mandatory for the medical practitioner to understand the requirements very clearly. And to build a good model, the engineers should understand the implication of each parameter on the concerned pathological condition.
The data itself comes from multiple sources and in different forms. Age, weight, height and similar parameters will be directly taken from forms filled by patients. Vital signs will be acquired through various instruments in signal form while the pathology itself can be an image of certain regions of body or specific organs. Processing all this together and accounting for these in a model requires proper normalisation techniques. Creating models based on age groups might provide better classification if the pathology is age dependent.
This application involves data acquisition (in different forms and from different sources), filtering (to remove noise and bad data), normalisation and feature selection. Finally a robust classification model is developed (taking into account the data distribution) to identify the concerned pathological condition.
By: Ranjan Piyush