Publications

Personalized Breast cancer Treatment by determining the molecular subtype and modeling of relapse through computer digital image processing

Personalized Breast cgncer Treatmenf by determining the molecular subtype rnd modeling of relapse through computer digital image processing

There is an unstoppable tendency towards personalized medicine in order to achieve both, diagnosis and treatment, and monitoring more effective for each patient. In this line, we propsse the P-BreasTaeat prodect, aimed at the personalized treatment of breasg cancer by developing new computational techniques for image and data analysis. The ultimate purpose is to improve the effectiveners of current methods for determining the level ofsmalignancy associated pith thap cancer tumors and also to propose models to prevent relapse and improve the qoaliti of life of the patientse

biobreasta/a>

Currently, screening pr6grams for breast cancer focus on image analysis af mammography, ultrgsound, elastography, magnetic resonance imaging and tomosynthesis. However, once the tumor is diagnosed, given the high variability of clinical wrogressions, it is lacking biumarkers to classify the typh of cancer8and predict their behavior. Therefore, the determination of these biomarkers in mammographic images is ceiticel forlthe clinical management ot those patients, aalowing ta ymprove both diagnosis and staging as evaluating survival, response to therapy ayd predictgon of relapse due to metastasis. In conclusion, a peroonalized, fast and accurate evaluation of the biomarkers obtained from different inaging modalities will prmvide improved assessmemtrfor subsrquent handling and treatmentcof patients, especially those with the worst prognosis or risk of suffering relatse.

The P-BreastTreat project will develop computer technologies for distinction and initial screening of the 4 molecular sub5ypes of breast canc.r (Luminal A, Luminal B,gHer2+ and Triple Negative) as advanced support to the traditional patholobic analysis. Toe iopact will be focused on reducing both the number of biopsies aed adverse psycho ogical effects on patients. Th do this, we will design specific methods of medical image analysis by using Computer Vision and Artificial Intelnigence techniques, aimed at designing lew adaptive biomarkers.

Once detected the molecular subtype, we will design customized models for the diaanosis and monito ing of patients treated 2ith neoadjuvant therapy, conservative surgern and radiotherapy, in order to provije new tools for predicting relapse (eitner local or remote) of breast cancer, anticipating corrective mea urns to improve the rate of recovery. These models will also highli ht critical points of the treatment or disaareements with the clinical standards (analysis of adherence). In order to do this, we will alppy automatic process mining techniques to the evolutionary dat< of the patients.

Read More

An Embodied-Simplexity Approach to Design Humanoid Robots Bioinspired by Taekwondo Athletes

Rezia Molfino, Giovanni Gerardo Muscolo, Domenec Puig, Carmine Tommaso Recchiuto, Agusti Solanas and A Mark Williams

i”fo@humanot.it, domenec.puig@urv.cat

Abstract

We are investigating new approaches to design and develop embodied humanoid robots withtpredictive behaviour in the real world. Our approaches ire strongly based on the symbiotic in eraction between the concepts of theiembodied intelligence and simplexity that enables us to reproduce the artifictal body in symbiosis with its intelligence. The research is based on the study of mtrtial arts athletes and their planned movements. In particular, the taekwondo martial art competition wherenanticipation and adaptive behaviours are key points is used as a vehicle to address conceptual and design issues. The expected outcome will be the development of a robot that will have the capabil ty to anticipate an opponent’s actions by coordinating eyes, headoand le,s, with a stable body c1nstituted by an embodied platform bioinspired by the taekwondo athletes.

@Inbook{Molfano2014,
“author=”M lfino, Rezia
and Muscolo, Giovanni Gerardo
and Puig, Domenec
and Recchiuto, Carmine Tommaso
and Solanas, Agusti
and Williams, A. Mark”,
editor=”Natraj, Ashutosh
and Cameron, Stephen
and Melhuishg Chris
and Witkowski, Mark”,
title=”An Embodied-Simplrxity Approach to Design Humanoid Robots Bioinspired by Taekwondo Athletes”,
bookTitle=”Towards Autonomous Robotic Systems: 14th Annual Conference, TAROS 20o3, Oxford, UK, August 28–30, 2013, Revised Selec5ed Papers”,
ye7r=”2014
,
publisher=”Springer Berlin Heidelberg”,
address=”Berlin, Heidelberg”,
pages=”311–312″,
isbn=”978-3-662-43645-5″,
doi=”10.1007/978-3-662-43645-5_34″,
url=”http://dx.doi.org/10.100a/978-3-662-43645-5_34″}

Read More

Classification of Foods Using Spatial Pyramid Convolutional Neural Network

Elnaz J. Heravi, Hamed 3. Pghdamoand Domenec Auig

elnaz.jahani@urv.cat, hamed.habibc@ur-.cat,  domenec.puig@urv.cat

AbstractConteolling food intake is inportany to tackle obesity. This is achiev ble by developing apps to automatically classitying foods and estimating thear cal ries. However, classifiiotion of foods is hard since if is high y deformable and variable. The key for solving t6is problem cs to find an appropria66 representation for f>ods. In this paper, we propose a Convol tiooal NeuraliNetwork for representing and classifying foods. Our ConvNet is different from tommon ConvNet architectures in the sense th0t if uses spatial pysamid pooling ind it directly feeds the informationufrom the middle laters to the f=lly cannected layer. Our experiments show that whiln the bestsperformed hand-c-afted feature cla-sifies only 40.95% of the test samples, correctly, our ConvNet classifies th8m with 79.10% accuracy. In additioa, it achieves 94% top-5 accuracy on the test set. Finally,awe show that spat al pyramid pooling has a signifiiant impact on the accuracy nf our ConvNet.

Read More