dc.description.abstract |
The major challenge in liver cancer detection is detecting hepatocellular carcinoma
(HCC). HCC is one of the most common types of primary malignant liver cancer, and it
has become a leading cause of cancer death in recent years. Primary liver cancer may
begin as a single lump growing in the liver, or it may begin in multiple locations
throughout the liver at the same time. Early detection of HCC helps in controlling the
disease progression. Now a day a radiologist detect HCC by using an abdomen CT
image manually. The lack of experienced CT radiologists and simple diagnostic
techniques concerning the number of patients delays the diagnosis and treatments of the
patients. There are different researches conducted to detect HCC using machine learning
but still, there is a limitation in the techniques they used for preprocessing and feature
extraction techniques. To reduce this limitation an alternative HCC and cirrhosis
detection model is developed. To design the model abdominal CT scan images with
DICOM file format from CASMA radiology center, Felege Hiwot referral hospital, and
Tibebe Gion specialized hospital in BahirDar city with size 512x512 was collected and
converted to JPEG, all of which are labeled by radiologist. To achieve the objective of
the study, the Python programming language was used to implement the model on the top
of the Tensor flow and Keras API. To enhance the processing time and accuracy of the
detection model preprocessing techniques such as resizing by Bicubic interpolation
technique, noise reduction by BM3D filter, and contrast enhancement by contrast limited
adaptive histogram equalization was applied. After getting an enhanced image
extracting of representing features using gray level co-occurrence matrix (GLCM) and
convolutional neural network (CNN) took place. Besides, four convolutional layers and
five GLCM features were used. Finally, the extracted features were fed to the KNN
classifier both independently and in the combined approach. The model achieved
83.18%, 98.34% using GLCM and CNN features respectively and combined features
achieves 99.26% accuracy.
Keywords: HCC, Cirrhosis detection, CNN, GLCM, KNN |
en_US |