Artificial intelligence-based versus manual assessment of prostate cancer in the prostate gland

a method comparison study

Mike Allan Mortensen*, Pablo Borrelli, Mads Hvid Poulsen, Oke Gerke, Olof Enqvist, Johannes Ulén, Elin Trägårdh, Caius Constantinescu, Lars Edenbrandt, Lars Lund, Poul Flemming Høilund-Carlsen

*Kontaktforfatter for dette arbejde

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

Resumé

Aim: To test the feasibility of a fully automated artificial intelligence-based method providing PET measures of prostate cancer (PCa). Methods: A convolutional neural network (CNN) was trained for automated measurements in 18F-choline (FCH) PET/CT scans obtained prior to radical prostatectomy (RP) in 45 patients with newly diagnosed PCa. Automated values were obtained for prostate volume, maximal standardized uptake value (SUV max), mean standardized uptake value of voxels considered abnormal (SUV mean) and volume of abnormal voxels (Vol abn). The product SUV mean × Vol abn was calculated to reflect total lesion uptake (TLU). Corresponding manual measurements were performed. CNN-estimated data were compared with the weighted surgically removed tissue specimens and manually derived data and related to clinical parameters assuming that 1 g ≈ 1 ml of tissue. Results: The mean (range) weight of the prostate specimens was 44 g (20–109), while CNN-estimated volume was 62 ml (31–108) with a mean difference of 13·5 g or ml (95% CI: 9·78–17·32). The two measures were significantly correlated (r = 0·77, P<0·001). Mean differences (95% CI) between CNN-based and manually derived PET measures of SUVmax, SUVmean, Vol abn (ml) and TLU were 0·37 (−0·01 to 0·75), −0·08 (−0·30 to 0·14), 1·40 (−2·26 to 5·06) and 9·61 (−3·95 to 23·17), respectively. PET findings Vol abn and TLU correlated with PSA (P<0·05), but not with Gleason score or stage. Conclusion: Automated CNN segmentation provided in seconds volume and simple PET measures similar to manually derived ones. Further studies on automated CNN segmentation with newer tracers such as radiolabelled prostate-specific membrane antigen are warranted.

OriginalsprogEngelsk
TidsskriftClinical Physiology and Functional Imaging
Vol/bind39
Udgave nummer6
Sider (fra-til)399-406
ISSN1475-0961
DOI
StatusUdgivet - nov. 2019

Fingeraftryk

Prostate
Prostatic Neoplasms
Neoplasm Grading
Weights and Measures
human glutamate carboxypeptidase II
Positron Emission Tomography Computed Tomography

Citer dette

Mortensen, Mike Allan ; Borrelli, Pablo ; Poulsen, Mads Hvid ; Gerke, Oke ; Enqvist, Olof ; Ulén, Johannes ; Trägårdh, Elin ; Constantinescu, Caius ; Edenbrandt, Lars ; Lund, Lars ; Høilund-Carlsen, Poul Flemming. / Artificial intelligence-based versus manual assessment of prostate cancer in the prostate gland : a method comparison study. I: Clinical Physiology and Functional Imaging. 2019 ; Bind 39, Nr. 6. s. 399-406.
@article{95b8ca8db41f44deb683ebb29fcc0917,
title = "Artificial intelligence-based versus manual assessment of prostate cancer in the prostate gland: a method comparison study",
abstract = "Aim: To test the feasibility of a fully automated artificial intelligence-based method providing PET measures of prostate cancer (PCa). Methods: A convolutional neural network (CNN) was trained for automated measurements in 18F-choline (FCH) PET/CT scans obtained prior to radical prostatectomy (RP) in 45 patients with newly diagnosed PCa. Automated values were obtained for prostate volume, maximal standardized uptake value (SUV max), mean standardized uptake value of voxels considered abnormal (SUV mean) and volume of abnormal voxels (Vol abn). The product SUV mean × Vol abn was calculated to reflect total lesion uptake (TLU). Corresponding manual measurements were performed. CNN-estimated data were compared with the weighted surgically removed tissue specimens and manually derived data and related to clinical parameters assuming that 1 g ≈ 1 ml of tissue. Results: The mean (range) weight of the prostate specimens was 44 g (20–109), while CNN-estimated volume was 62 ml (31–108) with a mean difference of 13·5 g or ml (95{\%} CI: 9·78–17·32). The two measures were significantly correlated (r = 0·77, P<0·001). Mean differences (95{\%} CI) between CNN-based and manually derived PET measures of SUVmax, SUVmean, Vol abn (ml) and TLU were 0·37 (−0·01 to 0·75), −0·08 (−0·30 to 0·14), 1·40 (−2·26 to 5·06) and 9·61 (−3·95 to 23·17), respectively. PET findings Vol abn and TLU correlated with PSA (P<0·05), but not with Gleason score or stage. Conclusion: Automated CNN segmentation provided in seconds volume and simple PET measures similar to manually derived ones. Further studies on automated CNN segmentation with newer tracers such as radiolabelled prostate-specific membrane antigen are warranted.",
keywords = "agreement, choline, convolutional neural network, diagnostic imaging, positron emission tomography, prostatic neoplasms",
author = "Mortensen, {Mike Allan} and Pablo Borrelli and Poulsen, {Mads Hvid} and Oke Gerke and Olof Enqvist and Johannes Ul{\'e}n and Elin Tr{\"a}g{\aa}rdh and Caius Constantinescu and Lars Edenbrandt and Lars Lund and H{\o}ilund-Carlsen, {Poul Flemming}",
note = "This article is protected by copyright. All rights reserved.",
year = "2019",
month = "11",
doi = "10.1111/cpf.12592",
language = "English",
volume = "39",
pages = "399--406",
journal = "Clinical Physiology and Functional Imaging",
issn = "1475-0961",
publisher = "Wiley-Blackwell",
number = "6",

}

Artificial intelligence-based versus manual assessment of prostate cancer in the prostate gland : a method comparison study. / Mortensen, Mike Allan; Borrelli, Pablo; Poulsen, Mads Hvid; Gerke, Oke; Enqvist, Olof; Ulén, Johannes; Trägårdh, Elin; Constantinescu, Caius; Edenbrandt, Lars; Lund, Lars; Høilund-Carlsen, Poul Flemming.

I: Clinical Physiology and Functional Imaging, Bind 39, Nr. 6, 11.2019, s. 399-406.

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

TY - JOUR

T1 - Artificial intelligence-based versus manual assessment of prostate cancer in the prostate gland

T2 - a method comparison study

AU - Mortensen, Mike Allan

AU - Borrelli, Pablo

AU - Poulsen, Mads Hvid

AU - Gerke, Oke

AU - Enqvist, Olof

AU - Ulén, Johannes

AU - Trägårdh, Elin

AU - Constantinescu, Caius

AU - Edenbrandt, Lars

AU - Lund, Lars

AU - Høilund-Carlsen, Poul Flemming

N1 - This article is protected by copyright. All rights reserved.

PY - 2019/11

Y1 - 2019/11

N2 - Aim: To test the feasibility of a fully automated artificial intelligence-based method providing PET measures of prostate cancer (PCa). Methods: A convolutional neural network (CNN) was trained for automated measurements in 18F-choline (FCH) PET/CT scans obtained prior to radical prostatectomy (RP) in 45 patients with newly diagnosed PCa. Automated values were obtained for prostate volume, maximal standardized uptake value (SUV max), mean standardized uptake value of voxels considered abnormal (SUV mean) and volume of abnormal voxels (Vol abn). The product SUV mean × Vol abn was calculated to reflect total lesion uptake (TLU). Corresponding manual measurements were performed. CNN-estimated data were compared with the weighted surgically removed tissue specimens and manually derived data and related to clinical parameters assuming that 1 g ≈ 1 ml of tissue. Results: The mean (range) weight of the prostate specimens was 44 g (20–109), while CNN-estimated volume was 62 ml (31–108) with a mean difference of 13·5 g or ml (95% CI: 9·78–17·32). The two measures were significantly correlated (r = 0·77, P<0·001). Mean differences (95% CI) between CNN-based and manually derived PET measures of SUVmax, SUVmean, Vol abn (ml) and TLU were 0·37 (−0·01 to 0·75), −0·08 (−0·30 to 0·14), 1·40 (−2·26 to 5·06) and 9·61 (−3·95 to 23·17), respectively. PET findings Vol abn and TLU correlated with PSA (P<0·05), but not with Gleason score or stage. Conclusion: Automated CNN segmentation provided in seconds volume and simple PET measures similar to manually derived ones. Further studies on automated CNN segmentation with newer tracers such as radiolabelled prostate-specific membrane antigen are warranted.

AB - Aim: To test the feasibility of a fully automated artificial intelligence-based method providing PET measures of prostate cancer (PCa). Methods: A convolutional neural network (CNN) was trained for automated measurements in 18F-choline (FCH) PET/CT scans obtained prior to radical prostatectomy (RP) in 45 patients with newly diagnosed PCa. Automated values were obtained for prostate volume, maximal standardized uptake value (SUV max), mean standardized uptake value of voxels considered abnormal (SUV mean) and volume of abnormal voxels (Vol abn). The product SUV mean × Vol abn was calculated to reflect total lesion uptake (TLU). Corresponding manual measurements were performed. CNN-estimated data were compared with the weighted surgically removed tissue specimens and manually derived data and related to clinical parameters assuming that 1 g ≈ 1 ml of tissue. Results: The mean (range) weight of the prostate specimens was 44 g (20–109), while CNN-estimated volume was 62 ml (31–108) with a mean difference of 13·5 g or ml (95% CI: 9·78–17·32). The two measures were significantly correlated (r = 0·77, P<0·001). Mean differences (95% CI) between CNN-based and manually derived PET measures of SUVmax, SUVmean, Vol abn (ml) and TLU were 0·37 (−0·01 to 0·75), −0·08 (−0·30 to 0·14), 1·40 (−2·26 to 5·06) and 9·61 (−3·95 to 23·17), respectively. PET findings Vol abn and TLU correlated with PSA (P<0·05), but not with Gleason score or stage. Conclusion: Automated CNN segmentation provided in seconds volume and simple PET measures similar to manually derived ones. Further studies on automated CNN segmentation with newer tracers such as radiolabelled prostate-specific membrane antigen are warranted.

KW - agreement

KW - choline

KW - convolutional neural network

KW - diagnostic imaging

KW - positron emission tomography

KW - prostatic neoplasms

U2 - 10.1111/cpf.12592

DO - 10.1111/cpf.12592

M3 - Journal article

VL - 39

SP - 399

EP - 406

JO - Clinical Physiology and Functional Imaging

JF - Clinical Physiology and Functional Imaging

SN - 1475-0961

IS - 6

ER -