Automatic detection of calcifications in breast cancer diagnosis based on machine learning classifiers
Early detection of breast cancer through mammography is vital, with calcifications in mammograms serving as key indicators. Distinguishing between benign and malignant calcifications is essential for accurate diagnosis and treatment. This study aims to develop a Computer-Aided Detection (CAD) system...
| Main Author: | |
|---|---|
| Format: | Monograph |
| Language: | English |
| Published: |
Universiti Sains Malaysia
2024
|
| Subjects: | |
| Online Access: | http://eprints.usm.my/61346/ http://eprints.usm.my/61346/1/Fatina%20Ham%20Yahya%20Ham-E.pdf |
| _version_ | 1848884685962215424 |
|---|---|
| author | Ham, Fatina Ham Yahya |
| author_facet | Ham, Fatina Ham Yahya |
| author_sort | Ham, Fatina Ham Yahya |
| building | USM Institutional Repository |
| collection | Online Access |
| description | Early detection of breast cancer through mammography is vital, with calcifications in mammograms serving as key indicators. Distinguishing between benign and malignant calcifications is essential for accurate diagnosis and treatment. This study aims to develop a Computer-Aided Detection (CAD) system to identify and classify breast calcifications. Data from confirmed breast cancer cases were collected from the Laboratory Information System (LIS) at the Women Imaging Suite (WISH) of Hospital Universiti Sains Malaysia (HUSM) and cross-verified with the Picture Archiving and Communication System (PACS) to select mammograms showing calcifications that met the inclusion criteria from September 2020 to December 2023. The performance of Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Random Forest (RF) models was evaluated using metrics such as accuracy, F1 score, recall, precision, specificity, sensitivity, and area under the curve (AUC). The SVM model showed balanced performance with 65.22% accuracy and an F1 score of 0.6, indicating a trade-off between precision (54.55%) and recall (66.67%). The KNN model had the lowest performance with 47.83% accuracy and an F1 score of 0.4, highlighting classification challenges. The RF model performed moderately with 60.87% accuracy and an F1 score of 0.47, showing high specificity (71.43%) but lower sensitivity (44.44%). Achieving 95% accuracy remains difficult due to reliance on high pixel value detection, limited complexity of machine learning models, and data constraints. Enhancing feature extraction, data augmentation, and model optimization could improve accuracy. Combining machine learning with deep learning or using ensemble methods offers promise for better classification, ultimately improving patient management. |
| first_indexed | 2025-11-15T19:10:38Z |
| format | Monograph |
| id | usm-61346 |
| institution | Universiti Sains Malaysia |
| institution_category | Local University |
| language | English |
| last_indexed | 2025-11-15T19:10:38Z |
| publishDate | 2024 |
| publisher | Universiti Sains Malaysia |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | usm-613462024-11-14T02:15:14Z http://eprints.usm.my/61346/ Automatic detection of calcifications in breast cancer diagnosis based on machine learning classifiers Ham, Fatina Ham Yahya R Medicine RC254-282 Neoplasms. Tumors. Oncology (including Cancer) Early detection of breast cancer through mammography is vital, with calcifications in mammograms serving as key indicators. Distinguishing between benign and malignant calcifications is essential for accurate diagnosis and treatment. This study aims to develop a Computer-Aided Detection (CAD) system to identify and classify breast calcifications. Data from confirmed breast cancer cases were collected from the Laboratory Information System (LIS) at the Women Imaging Suite (WISH) of Hospital Universiti Sains Malaysia (HUSM) and cross-verified with the Picture Archiving and Communication System (PACS) to select mammograms showing calcifications that met the inclusion criteria from September 2020 to December 2023. The performance of Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Random Forest (RF) models was evaluated using metrics such as accuracy, F1 score, recall, precision, specificity, sensitivity, and area under the curve (AUC). The SVM model showed balanced performance with 65.22% accuracy and an F1 score of 0.6, indicating a trade-off between precision (54.55%) and recall (66.67%). The KNN model had the lowest performance with 47.83% accuracy and an F1 score of 0.4, highlighting classification challenges. The RF model performed moderately with 60.87% accuracy and an F1 score of 0.47, showing high specificity (71.43%) but lower sensitivity (44.44%). Achieving 95% accuracy remains difficult due to reliance on high pixel value detection, limited complexity of machine learning models, and data constraints. Enhancing feature extraction, data augmentation, and model optimization could improve accuracy. Combining machine learning with deep learning or using ensemble methods offers promise for better classification, ultimately improving patient management. Universiti Sains Malaysia 2024-06 Monograph NonPeerReviewed application/pdf en http://eprints.usm.my/61346/1/Fatina%20Ham%20Yahya%20Ham-E.pdf Ham, Fatina Ham Yahya (2024) Automatic detection of calcifications in breast cancer diagnosis based on machine learning classifiers. Project Report. Universiti Sains Malaysia. (Submitted) |
| spellingShingle | R Medicine RC254-282 Neoplasms. Tumors. Oncology (including Cancer) Ham, Fatina Ham Yahya Automatic detection of calcifications in breast cancer diagnosis based on machine learning classifiers |
| title | Automatic detection of calcifications in breast cancer diagnosis based on machine learning classifiers |
| title_full | Automatic detection of calcifications in breast cancer diagnosis based on machine learning classifiers |
| title_fullStr | Automatic detection of calcifications in breast cancer diagnosis based on machine learning classifiers |
| title_full_unstemmed | Automatic detection of calcifications in breast cancer diagnosis based on machine learning classifiers |
| title_short | Automatic detection of calcifications in breast cancer diagnosis based on machine learning classifiers |
| title_sort | automatic detection of calcifications in breast cancer diagnosis based on machine learning classifiers |
| topic | R Medicine RC254-282 Neoplasms. Tumors. Oncology (including Cancer) |
| url | http://eprints.usm.my/61346/ http://eprints.usm.my/61346/1/Fatina%20Ham%20Yahya%20Ham-E.pdf |