| Abstrak/Abstract |
Due to the excellent spatial resolution that cardiac magnetic resonance imag-
ing (CMR) offers, it is possible to better extract crucial functional and morphological
aspects for the staging of cardiovascular illness. CMR, with its great spatial resolution,
allows for the improved extraction of crucial functional and morphological elements for
the staging of cardiovascular disease. Cardiologists employ CMR to assess temporal geo-
metric changes and manually estimate heart function by outlining forms. Yet, this work
demands a great deal of accuracy and takes a long time. For cardiac analysis, deep
learning techniques have been widely used. The stages proposed in this study include (i)
converting 3-dimensional images into 2-dimensional ones, (ii) obtaining contour values
from ground truth images, (iii) cropping the image localization area to obtain the region
of interest (RoI), (iv) dataset separation to train, validate, and test data (70%, 20%,
10%), and (v) classify using the ResNet50 V2 model. The ACDC MICCAI 2017 dataset
is used in this study to classify the cardiac into five main categories abnormal and one
healthy the ResNet50 V2 architecture. By having a measurement accuracy of 0.98, perfor-
mance outcomes are indicated. Experimental results show the robustness of the proposed
architecture. |
