Page 88 - sjsi
P. 88
Research Article: Alsibai & Heydari 88
repeated in the test and train directories. Figure 3 depicts a sample of the ultrasound
Therefore, one of the directories was neglected. images present in this dataset.
Figure 3: Samples of infected and healthy ovaries in Dataset A
Data Pre-Processing preprocessed so that they are uniform in this
There are several steps that may be taken to aspect. The preprocessing pipeline include:
preprocess medical ultrasound images before 1. Resizing all the images into the same size of
they are input into a deep learning model for (224, 224), which is the image size that
training. It's important to note that the specific DenseNet201 was trained on.
preprocessing steps will depend on the 2. Transforming the images into tensors. In
characteristics of the images and the specific Pytorch which is the deep learning of choice for
requirements of the deep learning model being this project, a tensor is a multi-dimensional array
used. Since the model chosen for this project is that is similar to a Numpy array, but can be
the DenseNet201 architecture which is a pre- operated on by the GPU, which makes it more
efficient for certain types of computations.
trained model on the ImageNet dataset, the Tensors can be used to store a wide variety of
ovarian ultrasound images that will be the input data, including images, videos, and audio, and
to this model need to be pre-processed the are an important building block in PyTorch. The
same way as the images this model was resulting tensor has the same number of
originally trained on. These images vary in size dimensions as the original image, with the size
and dimensions therefore they need to be of each dimension being the size of the
SJSI – 2023: VOLUME 1-1
