Andrea Guerriero

Dr. Andrea Guerriero

Department of Electrical and Information Engineering (DEI), Polytechnic University of Bari, Via Edoardo Orabona, 4, 70125 Bari, Italy

Research Focus & Metrics

Academic Career Length

18Years (2009-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
9 June 2021 in Informatics
Verified
Journal Article
Open access
PDF file
Nicola AltiniNicola Altini
Nicola Altini
GGGiuseppe De Giosa
Giuseppe De Giosa
NFNicola Fragasso
Nicola Fragasso
CCClaudia Coscia
Claudia Coscia
ESElena Sibilano
Elena Sibilano
Vitoantonio BevilacquaVitoantonio Bevilacqua
Vitoantonio Bevilacqua
The accurate segmentation and identification of vertebrae presents the foundations for spine analysis including fractures, malfunctions and other visual insights. The large-scale vertebrae segmentation challenge (VerSe), organized as a competition at the Medical Image Computing and Computer Assisted Intervention (MICCAI), is aimed at vertebrae segmentation and labeling. In this paper, we propose a framework that addresses the tasks of vertebrae segmentation and identification by exploiting both deep learning and classical machine learning methodologies. The proposed solution comprises two phases: a binary fully automated segmentation of the whole spine, which exploits a 3D convolutional neural network, and a semi-automated procedure that allows locating vertebrae centroids using traditional machine learning algorithms. Unlike other approaches, the proposed method comes with the added advantage of no requirement for single vertebrae-level annotations to be trained. A dataset of 214 CT scans has been extracted from VerSe’20 challenge data, for training, validating and testing the proposed approach. In addition, to evaluate the robustness of the segmentation and labeling algorithms, 12 CT scans from subjects affected by severe, moderate and mild scoliosis have been collected from a local medical clinic. On the designated test set from Verse’20 data, the binary spine segmentation stage allowed to obtain a binary Dice coefficient of 89.17%, whilst the vertebrae identification one reached an average multi-class Dice coefficient of 90.09%. In order to ensure the reproducibility of the algorithms hereby developed, the code has been made publicly available.

23 March 2022 in Applied Sciences
Verified
Journal Article
Open access
PDF file
BPBerardino Prencipe
Berardino Prencipe
Nicola AltiniNicola Altini
Nicola Altini
GCGiacomo Donato Cascarano
Giacomo Donato Cascarano
Antonio BrunettiAntonio Brunetti
Antonio Brunetti
AGAndrea Guerriero
Andrea Guerriero
Vitoantonio BevilacquaVitoantonio Bevilacqua
Vitoantonio Bevilacqua
Liver segmentation is a crucial step in surgical planning from computed tomography scans. The possibility to obtain a precise delineation of the liver boundaries with the exploitation of automatic techniques can help the radiologists, reducing the annotation time and providing more objective and repeatable results. Subsequent phases typically involve liver vessels’ segmentation and liver segments’ classification. It is especially important to recognize different segments, since each has its own vascularization, and so, hepatic segmentectomies can be performed during surgery, avoiding the unnecessary removal of healthy liver parenchyma. In this work, we focused on the liver segments’ classification task. We exploited a 2.5D Convolutional Neural Network (CNN), namely V-Net, trained with the multi-class focal Dice loss. The idea of focal loss was originally thought as the cross-entropy loss function, aiming at focusing on “hard” samples, avoiding the gradient being overwhelmed by a large number of falsenegatives. In this paper, we introduce two novel focal Dice formulations, one based on the concept of individual voxel’s probability and another related to the Dice formulation for sets. By applying multi-class focal Dice loss to the aforementioned task, we were able to obtain respectable results, with an average Dice coefficient among classes of 82.91%. Moreover, the knowledge of anatomic segments’ configurations allowed the application of a set of rules during the post-processing phase, slightly improving the final segmentation results, obtaining an average Dice coefficient of 83.38%. The average accuracy was close to 99%. The best model turned out to be the one with the focal Dice formulation based on sets. We conducted the Wilcoxon signed-rank test to check if these results were statistically significant, confirming their relevance.

20 July 2025 in Applied Sciences
Verified
Journal Article
Open access
PDF file
ACAnnachiara Cariola
Annachiara Cariola
ESElena Sibilano
Elena Sibilano
Antonio BrunettiAntonio Brunetti
Antonio Brunetti
Domenico BuongiornoDomenico Buongiorno
Domenico Buongiorno
AGAndrea Guerriero
Andrea Guerriero
Vitoantonio BevilacquaVitoantonio Bevilacqua
Vitoantonio Bevilacqua
Accurate segmentation of key tumor subregions in adult gliomas from Magnetic Resonance Imaging (MRI) is of critical importance for brain tumor diagnosis, treatment planning, and prognosis. However, this task remains poorly investigated and highly challenging due to the considerable variability in shape and appearance of these areas across patients. This study proposes a novel Deep Learning architecture leveraging modality-specific encoding and attention-based refinement for the segmentation of glioma subregions, including peritumoral edema (ED), necrotic core (NCR), and enhancing tissue (ET). The model is trained and validated on the Brain Tumor Segmentation (BraTS) 2023 challenge dataset and benchmarked against a state-of-the-art transformer-based approach. Our architecture achieves promising results, with Dice scores of 0.78, 0.86, and 0.88 for NCR, ED, and ET, respectively, outperforming SegFormer3D while maintaining comparable model complexity. To ensure a comprehensive evaluation, performance was also assessed on standard composite tumor regions, i.e., tumor core (TC) and whole tumor (WT). The statistically significant improvements obtained on all regions highlight the effectiveness of integrating complementary modality-specific information and applying channel-wise feature recalibration in the proposed model.

1 May 2009 in 2009 IEEE International Conference on Virtual Environments, Human-Computer Interfaces and Measurements Systems
Verified
Conference Paper
Closed access
AGAndrea Guerriero
Andrea Guerriero
CPCiriaco C. Pasquale
Ciriaco C. Pasquale
FRFrancesco Ragni
Francesco Ragni
Environments that integrate network computer into global computing resources have been studied since the development of Internet. In order to provide a platform independent approach to high-performance network computing, several Java-based implementations have been developed. Java overcomes the issue of maintaining different binary codes, multiple execution environments, and complex architectures. It offers the basic infrastructure to integrate computers resources connected to Internet into a distributed computational resource and allows parallel applications execution on different O.S. environment. In this paper we discuss the design and implementation of a network computing prototype system based on Java technology. The proposed multi level architecture uses the Agent paradigm to manage resources and to develop a user friendly interface to share resources in the grid.

17 June 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
AGAndrea Guerriero
Andrea Guerriero
Regional land-use inventories are often used as reference information inagricultural remote sensing, although they may be temporally displaced withrespect to the satellite observations under analysis. This is particularlycritical for hyperspectral crop and land-use mapping, where spectral signaturescan change with phenology, crop rotation, management practices, and land-usetransitions. In this work, we use PRISMA hyperspectral observations to audit thespectral reliability of SIT-derived land-use labels in the Puglia region ofSouthern Italy. The SIT inventory dates back to 2016, whereas the consideredPRISMA acquisitions span the 2019--2025 period. The inventory labels aretherefore treated as weak operational priors rather than contemporaneouspixel-level ground truth.We formulate a field-level reliability-auditing framework that evaluates thespectral support of original SIT level-4 labels in the PRISMA data. For eachsufficiently represented SIT class, spectral prototypes are estimated from thehyperspectral observations, and pixel-level distances from the assigned classprototype are aggregated at the field level. The resulting scores distinguishfield-scene units that are spectrally coherent with their assigned SIT classfrom intermediate spectral support or spectrally atypical cases that should be prioritized for inspection. Rare or insufficiently represented classes are retained in the inventory but excluded from prototype-based reliability estimation.The proposed score measures spectral support for an assigned operational SITlabel, not semantic correctness. Low support identifies spectrally atypicalfield-scene units and provides a diagnostic layer for inventory interpretation,but does not by itself indicate an inventory error.A targeted photo-interpretation check shows that most inspected field-sceneunits are consistent or partially consistent with the original SIT labels.However, spectrally atypical flags do not correspond directly to semantic labelerrors. Instead, they identify field-scene units that are spectrally atypicalwith respect to their assigned class. Such atypicality may reflect true labelinconsistencies, but also temporal mismatch, phenological stage, exposed soil,mixed parcels, or agricultural management practices such as protective covers invineyards.The proposed approach does not claim to detect confirmed land-use change in theabsence of contemporaneous ground truth. Rather, it provides an operationaltriage layer for identifying where legacy inventory labels are well supported byrecent hyperspectral observations and where they require further interpretation.By shifting the focus from classification accuracy to inventory reliabilityauditing, this study demonstrates how PRISMA data can support the criticalassessment of regional land-use information under temporal mismatch.

2 September 2026 in Remote Sensing Applications: Society and Environment
Verified
Journal Article
Closed access
AGAndrea Guerriero
Andrea Guerriero