top of page
  • Writer's pictureDr Edin Hamzić

From Genotype Data to Artificial Image Objects: An Interesting Approach for Schizophrenia Diagnosis

Updated: Jan 18

Why Do I Write These Types of Blog Posts?

  • One of the interests I will write about is the intersection of genomics and artificial intelligence in general and, more specifically, machine learning as a subdomain of artificial intelligence.

  • So, occasionally, I will post these hopefully insightful posts about interesting scientific papers that focus on this area.

Classifying Patients With Schizophrenia Artificial Image Objects Generated From Genotype Data

  • 🛎️ I already wrote personally the most excin exciting segment of the paper in the above subtitle: the authors applied a really interesting approach where they converted genotype data into artificial image objects to apply a convolutional neural network (CNN) to classify patients with schizophrenia.

  • The whole approach had a pretty reasonable accuracy of 80.6%. Honestly, I don't know if this is the first time this approach has been used, as the authors do not claim the novelty of the approach.

  • ℹ️ However, it would be interesting to see a comparison study where one would compare both genotype data as tabular data used as an input for the neural networks and artificial image objects for convolutional neural networks and assess their accuracy in classifying individuals with specific disorders.

  • Machine learning specialists might have a straightforward answer to this question without even doing comparative analysis since they have a more profound knowledge of how different neural network architectures actually learn and would perform on the type of data, such as genotype data.

  • ℹ️ In any case, it is an interesting feature engineering approach worth exploring further.


bottom of page