Artificial Intelligence against Breast Cancer (ANNES-BC-Project)

作者：Parmeggiani Domenico^*; Avenia Nicola; Sanguinetti Alessandro; Ruggiero Roberto; Docimo Giovanni; Siciliano Mattia; Ambrosino Pasquale; Madonna Imma; Peltrini Roberto; Parmeggiani Umberto

来源：Annali Italiani di Chirurgia, 2012, 83(1): 1-5.

• Artificial Neural Network
• Breast cancer
• Breast risk lesion

INTRODUCTION: Our preliminary study examined the development of an advanced innovative technology with the objectives of: %26lt;br%26gt;- developing methodologies and algorithms for a Artificial Neural Network (ANN) system, improving mammography and ultra-sonography images interpretation; %26lt;br%26gt;- creating autonomous software as a diagnostic tool for the physicians, allowing the possibility for the advanced application of databases using Artificial Intelligence (Expert System). %26lt;br%26gt;MATERIALS AND METHODS: Since 2004 550 F patients over 40 yrs old were divided in two groups: %26lt;br%26gt;1) 310 pts underwent echo every 6 months and mammography every year by expert radiologists. %26lt;br%26gt;2) 240 pts had the same screening program and were also examined by our diagnosis software, developed with ANN-ES technology by the Engineering Aircraft Research Project team. The information was continually updated and returned to the Expert System, defining the principal rules of automatic diagnosis. %26lt;br%26gt;RESULTS:In the second group we selected: Expert radiologist decision; ANN-ES decision; Expert radiologists with ANN-ES decision. The second group had significantly better diagnosis for cancer and better specificity for breast lesions risk as well as the highest percentage account when the radiologist%26apos;s decision was helped by the ANN software. The ANN-ES group was able to select, by anamnestic, diagnostic and genetic means, 8 patients for prophylactic surgery, finding 4 cancers in a very early stage. %26lt;br%26gt;DISCUSSION AND CONCLUSION: Although it is only a preliminary study,this innovative diagnostic tool seems to provide better positive and negative predictive value in cancer diagnosis as well as in breast risk lesion identification.

• 出版日期2012-2