Computer-Aided Measurement of Solid Breast Tumor Features on Ultrasound Images

Miguel Alemán-Flores, Patricia Alemán-Flores, Luis Álvarez-León, José Manuel Santana-Montesdeoca, Rafael Fuentes-Pavón, AgustínTrujillo-Pino

(1) Departamento de Informática y Sistemas. Universidad de Las Palmas de Gran Canaria.
(2) Sección de Ecografía. Servicio de Radiodiagnóstico. Hospital Universitario Insular de Gran Canaria.  


In this work, we have applied some Computer Vision techniques to the ultrasonographic analysis of solid breast nodules. Specialists have defined a series of parameters which help determine, with a high accuracy, whether a tumor is malignant or benign. However, it is not always possible for a human observer to obtain reliable measurements of the different parameters. The techniques which have been proposed allow achieving very promising results in the characterization of shapes and textures. We have developed an application to process the images and obtain measurements of the diagnostic criteria.

1. Diagnostic Criteria

For the analysis of ultrasound images, we have used the following criteria, based on Stavros et al. [1]:

Benignity Criteria:

Hiperechogenicity (1)
Ellipsoid shape (2)
Two or three gentle and well circumscribed lobulations(3)
Thin echogenic capsule (4)

Malignancy Criteria:

Hipoechogenicity (5)
Taller-than-wide shape (6)
Spiculation (7)
Angular margins (8)
Microlobulations (9)
Acoustic shadowing (10)
Microcalcifications (11)
Ramifications (12)

2. Feature Extraction

By means of the application of computer vision techniques [2], we have obtained some measurements for the different diagnostic criteria.

3. Results

Examples of the results generated by the application for different criteria in three breast nodules:

4. Conclusion

With these techniques, we have developed a prototype which provides, on the one hand, objective measurements of the global shape of the tumor, such as its dimensions, its disposition and how ellipsoid it is.
On the other hand, we have analyzed certain particular aspects of the contour, such as microlobulations, angular margins or spiculation. Furthermore, we have considered intensity-related features, such as echogenicity, shadowing and microcalcifications.
The main advantage of this application is the fact that it provides objective, robust and reproducible measurements of ultrasound findings which are now assessed in a subjective way. The results which have been obtained can improve the quality of the diagnosis.

[1] Stavros, A.T., Thickman, D., Rapp, C.L., Dennis, M.A., Parker, S.H., Sisney, G.A.: Solid Breast Nodules: Use of Sonography to Distinguish between Benign and Malignant Lesions. Radiology 196 (1995) 123-134.
[2] Alemán-Flores, M., Alemán-Flores, P., Álvarez-León, L., Santana-Montesdeoca, J.M., Fuentes-Pavón, R., Trujillo-Pino, A.: Computational Techniques for the Support of Breast Tumor Diagnosis on Ultrasound Images. Cuadernos del Instituto Universitario de Ciencias y Tecnologías Cibernéticas, ULPGC 27 (2003) 1-12. 

Computer Vision Approches to Medical Image Analysis

Prague 2004