Exploiting probabilistic uncertainty measures for respiratory motion prediction

Robert Dürichen; Tobias Wissel; Achim Schweikard

Exploiting probabilistic uncertainty measures for respiratory motion prediction

Robert Dürichen, Tobias Wissel, Achim Schweikard

Institute of Robotics and Cognitive Systems

Abstract

To precisely ablate moving tumours in robotic radiotherapy, time latencies have to be
compensated. In the past, several non-probabilistic regression techniques have been
investigated. Recently, relevance vector machines (RVM), a probabilistic regression technique,
have been successfully evaluated on one of the most comprehensive data sets and outperformed
six other prediction algorithms like support vector regression and wavelet based least mean
square (wLMS) [1]. The method has the distinct advantage that each predicted point is assumed
to be drawn from a normal distribution. So far only the mean of the distribution was used. We
want to investigate how the predicted variance can be used as safety feature and to construct
hybrid prediction algorithms.

Original language	English
Pages	59
Number of pages	1
Publication status	Published - 01.06.2014
Event	Proceedings of the 28th International Congress and Exhibition on Computer Assisted Radiology and Surgery - Fukuoka, Japan Duration: 25.06.2014 → 28.06.2014

Conference

Conference	Proceedings of the 28th International Congress and Exhibition on Computer Assisted Radiology and Surgery
Abbreviated title	(CARS'14)
Country/Territory	Japan
City	Fukuoka
Period	25.06.14 → 28.06.14

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being
SDG 9 Industry, Innovation, and Infrastructure

Access to Document

2014 - CARS - Dürichen - Exploiting Probabilistic Uncertainty Measures for Respiratory Motion PredictionFinal published version, 348 KB

Cite this

@conference{26906c012de143e892a8062acae189a0,

title = "Exploiting probabilistic uncertainty measures for respiratory motion prediction",

abstract = "To precisely ablate moving tumours in robotic radiotherapy, time latencies have to be compensated. In the past, several non-probabilistic regression techniques have been investigated. Recently, relevance vector machines (RVM), a probabilistic regression technique, have been successfully evaluated on one of the most comprehensive data sets and outperformed six other prediction algorithms like support vector regression and wavelet based least mean square (wLMS) [1]. The method has the distinct advantage that each predicted point is assumed to be drawn from a normal distribution. So far only the mean of the distribution was used. We want to investigate how the predicted variance can be used as safety feature and to construct hybrid prediction algorithms.",

author = "Robert D{\"u}richen and Tobias Wissel and Achim Schweikard",

note = "Motion Compensation in Radiosurgery; Proceedings of the 28th International Congress and Exhibition on Computer Assisted Radiology and Surgery ; Conference date: 25-06-2014 Through 28-06-2014",

year = "2014",

month = jun,

day = "1",

language = "English",

pages = "59",

}

TY - CONF

T1 - Exploiting probabilistic uncertainty measures for respiratory motion prediction

AU - Dürichen, Robert

AU - Wissel, Tobias

AU - Schweikard, Achim

N1 - Motion Compensation in Radiosurgery

PY - 2014/6/1

Y1 - 2014/6/1

N2 - To precisely ablate moving tumours in robotic radiotherapy, time latencies have to be compensated. In the past, several non-probabilistic regression techniques have been investigated. Recently, relevance vector machines (RVM), a probabilistic regression technique, have been successfully evaluated on one of the most comprehensive data sets and outperformed six other prediction algorithms like support vector regression and wavelet based least mean square (wLMS) [1]. The method has the distinct advantage that each predicted point is assumed to be drawn from a normal distribution. So far only the mean of the distribution was used. We want to investigate how the predicted variance can be used as safety feature and to construct hybrid prediction algorithms.

AB - To precisely ablate moving tumours in robotic radiotherapy, time latencies have to be compensated. In the past, several non-probabilistic regression techniques have been investigated. Recently, relevance vector machines (RVM), a probabilistic regression technique, have been successfully evaluated on one of the most comprehensive data sets and outperformed six other prediction algorithms like support vector regression and wavelet based least mean square (wLMS) [1]. The method has the distinct advantage that each predicted point is assumed to be drawn from a normal distribution. So far only the mean of the distribution was used. We want to investigate how the predicted variance can be used as safety feature and to construct hybrid prediction algorithms.

UR - https://www.rob.uni-luebeck.de/index.php?id=276&author=0:2646&L=0

M3 - Conference Papers

SP - 59

T2 - Proceedings of the 28th International Congress and Exhibition on Computer Assisted Radiology and Surgery

Y2 - 25 June 2014 through 28 June 2014

ER -

Exploiting probabilistic uncertainty measures for respiratory motion prediction

Abstract

Conference

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this