Advanced

Artificial intelligence-based versus manual assessment of prostate cancer in the prostate gland : a method comparison study

Mortensen, Mike A. ; Borrelli, Pablo ; Poulsen, Mads Hvid ; Gerke, Oke ; Enqvist, Olof ; Ulén, Johannes LU ; Trägårdh, Elin LU ; Constantinescu, Caius ; Edenbrandt, Lars LU and Lund, Lars , et al. (2019) In Clinical Physiology and Functional Imaging 39(6). p.399-406
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 (SUVmax), mean standardized uptake value of voxels considered abnormal (SUVmean) and volume of abnormal voxels (Volabn). The product SUVmean × Volabn was calculated to reflect total lesion uptake (TLU). Corresponding manual measurements were performed.... (More)

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 (SUVmax), mean standardized uptake value of voxels considered abnormal (SUVmean) and volume of abnormal voxels (Volabn). The product SUVmean × Volabn 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, Volabn (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 Volabn 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.

(Less)
Please use this url to cite or link to this publication:
author
, et al. (More)
(Less)
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
agreement, choline, convolutional neural network, diagnostic imaging, positron emission tomography, prostatic neoplasms
in
Clinical Physiology and Functional Imaging
volume
39
issue
6
pages
399 - 406
publisher
Wiley Online Library
external identifiers
  • pmid:31436365
  • scopus:85071942747
ISSN
1475-0961
DOI
10.1111/cpf.12592
language
English
LU publication?
yes
id
c4ab635d-1147-45d6-bcab-ec8012d49eb5
date added to LUP
2019-09-23 14:06:50
date last changed
2020-01-13 12:53:04
@article{c4ab635d-1147-45d6-bcab-ec8012d49eb5,
  abstract     = {<p>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 <sup>18</sup>F-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<sub>max</sub>), mean standardized uptake value of voxels considered abnormal (SUV<sub>mean</sub>) and volume of abnormal voxels (Vol<sub>abn</sub>). The product SUV<sub>mean</sub> × Vol<sub>abn</sub> 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&lt;0·001). Mean differences (95% CI) between CNN-based and manually derived PET measures of SUVmax, SUVmean, Vol<sub>abn</sub> (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<sub>abn</sub> and TLU correlated with PSA (P&lt;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.</p>},
  author       = {Mortensen, Mike A. and Borrelli, Pablo and Poulsen, Mads Hvid and Gerke, Oke and Enqvist, Olof and Ulén, Johannes and Trägårdh, Elin and Constantinescu, Caius and Edenbrandt, Lars and Lund, Lars and Høilund-Carlsen, Poul Flemming},
  issn         = {1475-0961},
  language     = {eng},
  month        = {08},
  number       = {6},
  pages        = {399--406},
  publisher    = {Wiley Online Library},
  series       = {Clinical Physiology and Functional Imaging},
  title        = {Artificial intelligence-based versus manual assessment of prostate cancer in the prostate gland : a method comparison study},
  url          = {http://dx.doi.org/10.1111/cpf.12592},
  doi          = {10.1111/cpf.12592},
  volume       = {39},
  year         = {2019},
}