Several large prospective and randomised trials have shown that breast-conserving therapy (BCT) is equivalent to modified radical mastectomy in terms of overall survival and local control, becoming the preferred treatment for women with early stage breast cancer.

Although considerable research has been put into BCT techniques, diverse aesthetic results are common, highlighting the importance of this evaluation in institutions performing breast cancer treatment in order to improve working practices. Approximately one-third of patients still report fair or poor cosmetic results.

This project explores new objective methodologies to predict the overall aesthetic result of BCT. Towards that end, synergies were made between two usually separated areas: concepts from the pattern recognition area were applied to help computationally determine the aesthetic result of BCT. Secondly, automation of the entire process was sought after by using automatic image analysis techniques, promoting the wide acceptance of the method. Finally,  the extension of this technique to the aesthetic assessment of other breast interventions was studied as in this area the evaluation of results still suffers the same limitations.

The first major activity was creating a database of images of patients undergoing breast cancer conservative treatment. A second major task focused on generating features in the digital image. These features must capture all factors impacting the overall aesthetic outcome and must be measurable in the digital photograph of the patient. The possible groups of discriminative features considered were those already identified by experts as relevant in surgical procedure assessment.

It is universally accepted that the aesthetic result is essentially determined by visible skin changes caused by radiotherapy, such as hyperpigmentation and telangiectasias, or those caused by surgery, such as an obvious scar, and by changes in the breast size or shape. For a discriminant representation it is necessary to extract from the digital image quantities that translate this perceptual information of the aesthetic result. Firstly, asymmetry measurements were used, which are undoubtedly the most publicised objective assessment of the aesthetic result. Since there are different measurements of asymmetry, all indices introduced in the literature were measured, which did not limit the performance of the subsequent tasks.

New asymmetry measurements were also introduced that seemed complementary to existing ones. All these measurements require a correction of scale to cm, possible with marks made on the skin of the patient at a known distance. As this need for additional marks limits the automation of the process, the adequacy of dimensionless measurements of asymmetry were studied and introduced in this study for the first time. These relative values are sufficient ​​for the decision process and therefore scale marks are not required.

The objective measurement of color change was also introduced in this area for the first time, which made it possible to compare the treated and the untreated breast, something that is highly valued in subjective ratings. The hyperpigmentation is a typical result of radiotherapy, and depends of the radiation dosage. The contribution of the surgical scar is the most difficult to quantify due to its strong subjectivity. Aspects such as size, position, contrast with the surroundings contribute to the overall perception. In this work, the visibility of the scar was translated as a difference of local color.

The third task was to extract various image features. The automation of this operation - to make the system completely objective and simple to use - requires the prior identification of fiducial points, namely the position of the nipples, breast contour and jugularis notch. Finally, all the results were integrated in a system to evaluate the aesthetic result of the breast cancer treatment, available to all units working in this area. This system is already being used internationally in several prospective studies.

New algorithms were proposed for the aesthetic evaluation of the results of breast cancer conservative treatment. New algorithms were also proposed to automatically detect fiducial points on the image, as well as new measurements to capture the factors relevant to the overall aesthetic result, and new algorithms for evaluating the overall aesthetic result from individual measurements. Furthermore, a new system integrating all these innovations was proposed and developed. As a result, this project contributed to advancing the research in this area.

Project leader: INESC Porto

Contacts: Jaime S. Cardoso

Partners:

Instituto de Engenharia de Sistemas e Computadores do Porto - INESC Porto

Faculdade de Medicina da Universidade do Porto

Financial support:

Financed by the ERDF – European Regional Development Fund through the COMPETE Programme (operational programme for competitiveness) and by National Funds through the FCT – Fundação para a Ciência e a Tecnologia (Portuguese Foundation for Science and Technology).