Original scientific paper
https://doi.org/10.55378/rv.49.1.2
The Efficiency of Artificial Intelligence in Distinguishing Malignant from Benign Ultrasound Features of Breast Tumors
Filip Marković
orcid.org/0009-0006-7547-3883
; Medicinski fakultet Osijek
*
Damir Štimac
; Nacionalna memorijalna bolnica „Dr. Juraj Njavro“ Vukovar
* Corresponding author.
Abstract
In the female population, breast cancer is the most common malignant disease, with a mortality rate of 20%, making it a significant social and health problem, especially in developed countries. Early detection of breast cancer plays a crucial role in improving treatment outcomes. Ultrasound diagnostics represent a significant tool for identifying and differentiating suspicious lesions, but accurate evaluation can be challenging due to the subjectivity of interpretation. In this context, artificial intelligence (AI) and deep learning offer the potential for standardization and enhancement of diagnostic accuracy. This retrospective study was conducted at the Radiology Department of the National Memorial Hospital “Dr. Juraj Njavro” in Vukovar, using 257 ultrasound images from patients who underwent ultrasound examinations and pathohistological analysis via needle biopsy from November 2021 to February 2024. The collected images were adjusted for analysis using the DenseNet neural network. The aim of this research was to examine the effectiveness of the DenseNet neural network in analyzing ultrasound images of breast tumors, focusing on distinguishing malignant from benign formations and the potential to reduce the number of unnecessary invasive diagnostic procedures. The diagnostic efficacy of the model was assessed by comparing its results with pathohistological findings, utilizing standard statistical tests. The model demonstrated an accuracy of 73%, an AUC (area under the curve) value of 0.78, a specificity of 68.8%, a sensitivity of 80%, and an F1-score of 70%.
Keywords
artificial Intelligence; breast tumor; breast ultrasound
Hrčak ID:
330407
URI
Publication date:
11.4.2025.
Visits: 95 *