-
End output of program
-
Output of program
-
First attempt to get contourns
-
Visual detection of variables (rows go to origin to detected point)
-
Colored contourns
Inspiration
As we are students of software specialization that we wanted a new challenge, we never tried before this kind of projects (computing vision, health themes, ...) and as we see a real world problem where we could try to help we decide it to try it
What it does
Its a CLI's program where have to give a psla and sax dicom files path. Then gets some variables from the files to throw some conclusions (if it's a healthy heart or not) and also it save it as excel file.
How we built it
We built it as a python project with its virtual environment with the requirements needed, also needs a special program called Theseract that it is a optical character recognition open source to get the BPM from the files, the rest of the needed parameters are calculed using opencv mainly.
Challenges we ran into
The most significative challenge that the team ran into consisted in the identification of the contourns of the image. Differenciating between the underside and the outer layer of the blood vessel ended up being an arduous task
Accomplishments that we're proud of
Since distinguishing between the outer and inner layer was one of the most challenging and troublesome problem that the team integrants had to face, it's one of the most relevant acomplishments. It should also be noted that finding an external program that could be included in the project to easen the identification of text was also remarkable
What we learned
During the course of this project, the team members, which were not familiarized with image interpretation and treatment, had to learn about that topic and CLI programs
What's next for DICOM analysis heart hypertrophy
Alongside the existing metrics that might indicate the existence of hypertophy, the usage of both supervised and unsupervised machine learning methods could be considered so as to improve the precision of the obtained results, improve the sistole and diastole detection and add new functionalities as read directories and custom save.
Log in or sign up for Devpost to join the conversation.