TY - GEN
T1 - Non-Small Cell Lung Cancer (NSCLC) Classification Using Convolutional Neural Network (CNN)
AU - Darmawan, Bunga Mastiti
AU - Agustin, Rizki Wulan
AU - Noviandini, Farah
AU - Endarko,
N1 - Publisher Copyright:
© 2022 American Institute of Physics Inc.. All rights reserved.
PY - 2022/11/10
Y1 - 2022/11/10
N2 - Small Cell Lung Cancer (SCLC) and Non-Small Cell Lung Cancer (NSCLC) are types of Lung cancer. SCLC usually occurs in patients who have a history of heavy smoking and spreads more rapidly than NSCLC. However, about 80-85% of all lung cancer cases are NSCLC types that mostly attack men and women. This study aimed to classify NSCLC into squamous cell carcinoma, adenocarcinoma, large cell carcinoma, and normal lung and to compare the architecture VGG19 and ResNet50 for classification NSCLC. 1000 images data for each class was used in this study from CT Scan images. Three processes were used for classification processes, such as preprocessing, classification, and validation. The resizing and grayscale process was conducted for preprocessing step to ensure all input images are uniform. The significant result was achieved in the ResNet50 architecture, with an accuracy of 98.35% on testing data, 99.87% on training data, and 96% on validation data. Meanwhile, the best performance on the validation data in the normal class was that the results of precision, sensitivity, F1-score, and specificity were 100%, 100%, 100%, and 100%, respectively.
AB - Small Cell Lung Cancer (SCLC) and Non-Small Cell Lung Cancer (NSCLC) are types of Lung cancer. SCLC usually occurs in patients who have a history of heavy smoking and spreads more rapidly than NSCLC. However, about 80-85% of all lung cancer cases are NSCLC types that mostly attack men and women. This study aimed to classify NSCLC into squamous cell carcinoma, adenocarcinoma, large cell carcinoma, and normal lung and to compare the architecture VGG19 and ResNet50 for classification NSCLC. 1000 images data for each class was used in this study from CT Scan images. Three processes were used for classification processes, such as preprocessing, classification, and validation. The resizing and grayscale process was conducted for preprocessing step to ensure all input images are uniform. The significant result was achieved in the ResNet50 architecture, with an accuracy of 98.35% on testing data, 99.87% on training data, and 96% on validation data. Meanwhile, the best performance on the validation data in the normal class was that the results of precision, sensitivity, F1-score, and specificity were 100%, 100%, 100%, and 100%, respectively.
UR - http://www.scopus.com/inward/record.url?scp=85142928176&partnerID=8YFLogxK
U2 - 10.1063/5.0103184
DO - 10.1063/5.0103184
M3 - Conference contribution
AN - SCOPUS:85142928176
T3 - AIP Conference Proceedings
BT - 4th International Conference of Science and Education Science, IConSSE 2021
A2 - Nugroho, Didit Budi
A2 - Setiawan, Andreas
A2 - Wibowo, Nur Aji
A2 - Riyanto, Cucun Alep
A2 - Aminu, November Rianto
PB - American Institute of Physics Inc.
T2 - 4th International Conference of Science and Education Science: Integrating Rapid Technology and Whole Person Education in Science and Science Education to Encounter the New Normal Era, IConSSE 2021
Y2 - 7 September 2021 through 8 September 2021
ER -
