Artificial intelligence (AI) and especially deep learning (DL) are applied in many fields and are the leading technologies towards robot autonomy. Proper DL algorithms and sufficient amount of medical data may be applied in medical robotics to increase autonomy, safety, accuracy and precision of a medical procedure. We investigate in this paper the implementation and the integration of DL techniques in a new robotic system for prostate biopsy -PROST. The DL algorithm, named PROST-Net, is a convolutional neural network (CNN) employed for the segmentation of the prostate in different types of medical images: pre-operative magnetic resonance (MRI) and intra-operative ultrasound (US). The US images come from different acquisition planes (axial and sagittal) with different alignments of the sensors (convex and linear) making the design of the CNN challenging. Tests on patient data produced an accuracy of 86% in US images and 77% in MRI and were estimated by using Dice similarity Coefficient (DC).The biopsy robot will use the output of PROST-Net for the initial fusion of pre-operative and intra-operative images to define the biopsy targets and the planning of the procedure. Real-time processing of the data with PROST-Net will empower dynamic update of the initial fusion by following the current position

Autonomy in robotic prostate biopsy through AI-assisted fusion

Palladino, Luigi
;
Maris, Bogdan;Fiorini, Paolo
2021-01-01

Abstract

Artificial intelligence (AI) and especially deep learning (DL) are applied in many fields and are the leading technologies towards robot autonomy. Proper DL algorithms and sufficient amount of medical data may be applied in medical robotics to increase autonomy, safety, accuracy and precision of a medical procedure. We investigate in this paper the implementation and the integration of DL techniques in a new robotic system for prostate biopsy -PROST. The DL algorithm, named PROST-Net, is a convolutional neural network (CNN) employed for the segmentation of the prostate in different types of medical images: pre-operative magnetic resonance (MRI) and intra-operative ultrasound (US). The US images come from different acquisition planes (axial and sagittal) with different alignments of the sensors (convex and linear) making the design of the CNN challenging. Tests on patient data produced an accuracy of 86% in US images and 77% in MRI and were estimated by using Dice similarity Coefficient (DC).The biopsy robot will use the output of PROST-Net for the initial fusion of pre-operative and intra-operative images to define the biopsy targets and the planning of the procedure. Real-time processing of the data with PROST-Net will empower dynamic update of the initial fusion by following the current position
978-1-6654-3684-7
Image segmentation , Ultrasonic imaging , Magnetic resonance imaging , Biopsy , Transformers , Real-time systems , User experience
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/11562/1065181
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? 1
social impact