Entity

Time filter

Source Type

Lauzacco, Italy

Pella A.,Polytechnic of Milan | Cambria R.,Italian National Cancer Institute | Riboldi M.,Polytechnic of Milan | Riboldi M.,Biongineering Unit | And 13 more authors.
Medical Physics | Year: 2011

Purpose: The goal of this study is to investigate the advantages of large scale optimization methods vs conventional classification techniques in predicting acute toxicity for urinary bladder and rectum due to prostate irradiation. Methods: Clinical and dosimetric data of 321 patients undergoing prostate conformal radiotherapy were recorded. Gastro-intestinal and genito-urinary acute toxicities were scored according to the Radiation Therapy Oncology GroupEuropean Organization for Research and Treatment of Cancer (RTOGEORTC) scale. Patients were classified in two categories to separate mild (Grade2) from severe toxicity levels (Grade2). Machine learning methods at different complexity were implemented to predict toxicity as a function of multiple variables. The first approach consisted of a large scale optimization method, based on genetic algorithms (GAs) and artificial neural networks (ANN). The second approach was a binary classifier based on support vector machines (SVM). Results: The ANN and SVM-based solutions showed comparable prediction accuracy, exhibiting an area under the receiver operating characteristic (ROC) curve of 0.7. Different sensitivity and specificity features were measured for the two approaches. The ANN algorithm showed enhanced sensitivity if combined with appropriate classification criteria. Conclusions: The results demonstrate that high sensitivity in toxicity prediction can be achieved with optimized ANNs, that are put forward to represent a valuable support in medical decisions. Future studies will be focused on enlarging the available patient database to increase the reliability of toxicity prediction algorithms and to define optimal classification criteria. © 2011 American Association of Physicists in Medicine. Source


Torshabi A.E.,Biongineering Unit | Pella A.,Polytechnic of Milan | Riboldi M.,Polytechnic of Milan | Baroni G.,Biongineering Unit | Baroni G.,Polytechnic of Milan
Technology in Cancer Research and Treatment | Year: 2010

The use of external surrogates to predict tumor motion in real-time for extra-cranial sites requires the use of accurate correlation models. This is extremely challenging when motion prediction is to be performed over several breathing cycles, as occurs for realtime tumor tracking with Cyberknife® Synchrony®. In this work we compare three different approaches to infer tumor motion based on external surrogates, since no comparative study is available to assess the accuracy of correlation models in tumor tracking over a long time period. We selected 20 cases in a database of 130 patients treated with real-time tumor tracking by means of the Synchrony® module. The implemented correlation models comprise linear/quadratic correlation, artificial neural networks and fuzzy logic. The accuracy of each correlation model is evaluated on the basis of ground truth tumor position information acquired during treatment, as detected by means of stereoscopic X-ray imaging. Results show that the implemented models achieve an error reduction with respect to Synchrony®, measured at the 95% confidence level, up to 10.8% for the fuzzy logic approach. This latter is able to partly reduce the incidence of tumor tracking errors above 6 mm, resulting in improved accuracy for larger discrepancies. In conclusion, complex models are suggested to predict tumor motion over long time periods. This leads to an effective improvement with respect to Cyberknife® Synchrony®. Future studies will investigate the sensitivity of the implemented models to the input database, in order to define optimal strategies. Source

Discover hidden collaborations