In this paper, we propose a novel approach to adapt 2D region growing algorithms to volumetric segmentation of liver and spleen from Computed Tomography (CT) scans. Abdominal organ segmentation is an essential and time-consuming task in clinical radiology. The possibility to implement a semi-automatic segmentation system could speed up the time required to label the images and to improve the delineation results, minimizing both intra- and inter-operator variability. The proposed region growing algorithm exploits an initial seed point to perform the first slice-wise segmentation. Then, starting from this area, all other seeds are automatically discovered taking advantage of two data structures that we called Moving Average Seed Heatmap (MASH) and Area Union Map (AUM). The implemented mechanism avoids the choice of unsuitable seeds and the exclusion of irrelevant organs and tissues from the CT scan. We assessed the validity of the proposed liver and spleen segmentation method on two publicly available datasets: SLIVER07 and Medical Segmentation Decathlon Task 09 (MSD 09), respectively. The proposed method allowed us to obtain promising results for both liver and spleen segmentation, with a Dice Coefficient higher than 93% for the liver segmentation task and a Dice Coefficient greater than 92% for the spleen segmentation task on the designated validation sets.

A Novel Approach Based on Region Growing Algorithm for Liver and Spleen Segmentation from CT Scans / Prencipe, Berardino; Altini, Nicola; Cascarano, Giacomo Donato; Guerriero, Andrea; Brunetti, Antonio. - STAMPA. - 12463:(2020), pp. 398-410. [10.1007/978-3-030-60799-9_35]

A Novel Approach Based on Region Growing Algorithm for Liver and Spleen Segmentation from CT Scans

Prencipe, Berardino;Altini, Nicola;Cascarano, Giacomo Donato;Guerriero, Andrea;Brunetti, Antonio
2020-01-01

Abstract

In this paper, we propose a novel approach to adapt 2D region growing algorithms to volumetric segmentation of liver and spleen from Computed Tomography (CT) scans. Abdominal organ segmentation is an essential and time-consuming task in clinical radiology. The possibility to implement a semi-automatic segmentation system could speed up the time required to label the images and to improve the delineation results, minimizing both intra- and inter-operator variability. The proposed region growing algorithm exploits an initial seed point to perform the first slice-wise segmentation. Then, starting from this area, all other seeds are automatically discovered taking advantage of two data structures that we called Moving Average Seed Heatmap (MASH) and Area Union Map (AUM). The implemented mechanism avoids the choice of unsuitable seeds and the exclusion of irrelevant organs and tissues from the CT scan. We assessed the validity of the proposed liver and spleen segmentation method on two publicly available datasets: SLIVER07 and Medical Segmentation Decathlon Task 09 (MSD 09), respectively. The proposed method allowed us to obtain promising results for both liver and spleen segmentation, with a Dice Coefficient higher than 93% for the liver segmentation task and a Dice Coefficient greater than 92% for the spleen segmentation task on the designated validation sets.
2020
Intelligent Computing Theories and Application : 16th International Conference, ICIC 2020, Bari, Italy, October 2–5, 2020. Proceedings, Part I
978-3-030-60798-2
Springer
A Novel Approach Based on Region Growing Algorithm for Liver and Spleen Segmentation from CT Scans / Prencipe, Berardino; Altini, Nicola; Cascarano, Giacomo Donato; Guerriero, Andrea; Brunetti, Antonio. - STAMPA. - 12463:(2020), pp. 398-410. [10.1007/978-3-030-60799-9_35]
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11589/206766
Citazioni
  • Scopus 3
  • ???jsp.display-item.citation.isi??? ND
social impact