Machine Learning Facilitates Hotspot Classification in PSMA-PET/CT with Nuclear Medicine Specialist Accuracy
Gallium-68 prostate-specific membrane antigen positron emission tomography <b>(</b><sup>68</sup>Ga-PSMA-PET) is a highly sensitive method to detect prostate cancer (PC) metastases. Visual discrimination between malignant and physiologic/unspecific tracer accumulation by a nuc...
Main Authors: | , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-08-01
|
Series: | Diagnostics |
Subjects: | |
Online Access: | https://www.mdpi.com/2075-4418/10/9/622 |
id |
doaj-60e67ac95eda4abfb8f4ba8bd4e425a4 |
---|---|
record_format |
Article |
spelling |
doaj-60e67ac95eda4abfb8f4ba8bd4e425a42020-11-25T03:51:23ZengMDPI AGDiagnostics2075-44182020-08-011062262210.3390/diagnostics10090622Machine Learning Facilitates Hotspot Classification in PSMA-PET/CT with Nuclear Medicine Specialist AccuracySobhan Moazemi0Zain Khurshid1Annette Erle2Susanne Lütje3Markus Essler4Thomas Schultz5Ralph A. Bundschuh6Department of Nuclear Medicine, University Hospital Bonn, 53127 Bonn, GermanyDepartment of Nuclear Medicine, Nuclear Medicine, Oncology and Radiotherapy Institute, 21061 Islamabad, PakistanDepartment of Nuclear Medicine, University Hospital Bonn, 53127 Bonn, GermanyDepartment of Nuclear Medicine, University Hospital Bonn, 53127 Bonn, GermanyDepartment of Nuclear Medicine, University Hospital Bonn, 53127 Bonn, GermanyDepartment of Computer Science, University of Bonn, 53115 Bonn, GermanyDepartment of Nuclear Medicine, University Hospital Bonn, 53127 Bonn, GermanyGallium-68 prostate-specific membrane antigen positron emission tomography <b>(</b><sup>68</sup>Ga-PSMA-PET) is a highly sensitive method to detect prostate cancer (PC) metastases. Visual discrimination between malignant and physiologic/unspecific tracer accumulation by a nuclear medicine (NM) specialist is essential for image interpretation. In the future, automated machine learning (ML)-based tools will assist physicians in image analysis. The aim of this work was to develop a tool for analysis of <sup>68</sup>Ga-PSMA-PET images and to compare its efficacy to that of human readers. Five different ML methods were compared and tested on multiple positron emission tomography/computed tomography (PET/CT) data-sets. Forty textural features extracted from both PET- and low-dose CT data were analyzed. In total, 2419 hotspots from 72 patients were included. Comparing results from human readers to those of ML-based analyses, up to 98% area under the curve (AUC), 94% sensitivity (SE), and 89% specificity (SP) were achieved. Interestingly, textural features assessed in native low-dose CT increased the accuracy significantly. Thus, ML based on <sup>68</sup>Ga-PSMA-PET/CT radiomics features can classify hotspots with high precision, comparable to that of experienced NM physicians. Additionally, the superiority of multimodal ML-based analysis considering all PET and low-dose CT features was shown. Morphological features seemed to be of special additional importance even though they were extracted from native low-dose CTs.https://www.mdpi.com/2075-4418/10/9/622prostate cancer (PC)prostate-specific membrane antigen (PSMA)positron emission tomography (PET)computed tomography (CT)machine learning (ML) |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Sobhan Moazemi Zain Khurshid Annette Erle Susanne Lütje Markus Essler Thomas Schultz Ralph A. Bundschuh |
spellingShingle |
Sobhan Moazemi Zain Khurshid Annette Erle Susanne Lütje Markus Essler Thomas Schultz Ralph A. Bundschuh Machine Learning Facilitates Hotspot Classification in PSMA-PET/CT with Nuclear Medicine Specialist Accuracy Diagnostics prostate cancer (PC) prostate-specific membrane antigen (PSMA) positron emission tomography (PET) computed tomography (CT) machine learning (ML) |
author_facet |
Sobhan Moazemi Zain Khurshid Annette Erle Susanne Lütje Markus Essler Thomas Schultz Ralph A. Bundschuh |
author_sort |
Sobhan Moazemi |
title |
Machine Learning Facilitates Hotspot Classification in PSMA-PET/CT with Nuclear Medicine Specialist Accuracy |
title_short |
Machine Learning Facilitates Hotspot Classification in PSMA-PET/CT with Nuclear Medicine Specialist Accuracy |
title_full |
Machine Learning Facilitates Hotspot Classification in PSMA-PET/CT with Nuclear Medicine Specialist Accuracy |
title_fullStr |
Machine Learning Facilitates Hotspot Classification in PSMA-PET/CT with Nuclear Medicine Specialist Accuracy |
title_full_unstemmed |
Machine Learning Facilitates Hotspot Classification in PSMA-PET/CT with Nuclear Medicine Specialist Accuracy |
title_sort |
machine learning facilitates hotspot classification in psma-pet/ct with nuclear medicine specialist accuracy |
publisher |
MDPI AG |
series |
Diagnostics |
issn |
2075-4418 |
publishDate |
2020-08-01 |
description |
Gallium-68 prostate-specific membrane antigen positron emission tomography <b>(</b><sup>68</sup>Ga-PSMA-PET) is a highly sensitive method to detect prostate cancer (PC) metastases. Visual discrimination between malignant and physiologic/unspecific tracer accumulation by a nuclear medicine (NM) specialist is essential for image interpretation. In the future, automated machine learning (ML)-based tools will assist physicians in image analysis. The aim of this work was to develop a tool for analysis of <sup>68</sup>Ga-PSMA-PET images and to compare its efficacy to that of human readers. Five different ML methods were compared and tested on multiple positron emission tomography/computed tomography (PET/CT) data-sets. Forty textural features extracted from both PET- and low-dose CT data were analyzed. In total, 2419 hotspots from 72 patients were included. Comparing results from human readers to those of ML-based analyses, up to 98% area under the curve (AUC), 94% sensitivity (SE), and 89% specificity (SP) were achieved. Interestingly, textural features assessed in native low-dose CT increased the accuracy significantly. Thus, ML based on <sup>68</sup>Ga-PSMA-PET/CT radiomics features can classify hotspots with high precision, comparable to that of experienced NM physicians. Additionally, the superiority of multimodal ML-based analysis considering all PET and low-dose CT features was shown. Morphological features seemed to be of special additional importance even though they were extracted from native low-dose CTs. |
topic |
prostate cancer (PC) prostate-specific membrane antigen (PSMA) positron emission tomography (PET) computed tomography (CT) machine learning (ML) |
url |
https://www.mdpi.com/2075-4418/10/9/622 |
work_keys_str_mv |
AT sobhanmoazemi machinelearningfacilitateshotspotclassificationinpsmapetctwithnuclearmedicinespecialistaccuracy AT zainkhurshid machinelearningfacilitateshotspotclassificationinpsmapetctwithnuclearmedicinespecialistaccuracy AT annetteerle machinelearningfacilitateshotspotclassificationinpsmapetctwithnuclearmedicinespecialistaccuracy AT susannelutje machinelearningfacilitateshotspotclassificationinpsmapetctwithnuclearmedicinespecialistaccuracy AT markusessler machinelearningfacilitateshotspotclassificationinpsmapetctwithnuclearmedicinespecialistaccuracy AT thomasschultz machinelearningfacilitateshotspotclassificationinpsmapetctwithnuclearmedicinespecialistaccuracy AT ralphabundschuh machinelearningfacilitateshotspotclassificationinpsmapetctwithnuclearmedicinespecialistaccuracy |
_version_ |
1724488097552400384 |