A new computer-aided diagnostic tool for non-invasive characterisation of malignant ovarian masses: results of a multicentre validation study

Eur Radiol. 2010 Aug;20(8):1822-30. doi: 10.1007/s00330-010-1750-6. Epub 2010 Mar 20.

Abstract

Objectives: To prospectively assess an innovative computer-aided diagnostic technology that quantifies characteristic features of backscattered ultrasound and theoretically allows transvaginal sonography (TVS) to discriminate benign from malignant adnexal masses.

Methods: Women (n = 264) scheduled for surgical removal of at least one ovary in five centres were included. Preoperative three-dimensional (3D)-TVS was performed and the voxel data were analysed by the new technology. The findings at 3D-TVS, serum CA125 levels and the TVS-based diagnosis were compared with histology. Cancer was deemed present when invasive or borderline cancerous processes were observed histologically.

Results: Among 375 removed ovaries, 141 cancers (83 adenocarcinomas, 24 borderline, 16 cases of carcinomatosis, nine of metastases and nine others) and 234 non-cancerous ovaries (107 normal, 127 benign tumours) were histologically diagnosed. The new computer-aided technology correctly identified 138/141 malignant lesions and 206/234 non-malignant tissues (98% sensitivity, 88% specificity). There were no false-negative results among the 47 FIGO stage I/II ovarian lesions. Standard TVS and CA125 had sensitivities/specificities of 94%/66% and 89%/75%, respectively. Combining standard TVS and the new technology in parallel significantly improved TVS specificity from 66% to 92% (p < 0.0001).

Conclusions: Computer-aided quantification of backscattered ultrasound is a highly sensitive for the diagnosis of malignant ovarian masses.

Publication types

  • Evaluation Study
  • Multicenter Study
  • Validation Study

MeSH terms

  • Adolescent
  • Adult
  • Aged
  • Aged, 80 and over
  • Algorithms*
  • Artificial Intelligence
  • Europe
  • Female
  • Humans
  • Image Enhancement / methods
  • Image Interpretation, Computer-Assisted / methods*
  • Imaging, Three-Dimensional / methods
  • Middle Aged
  • Ovarian Neoplasms / diagnostic imaging*
  • Pattern Recognition, Automated / methods*
  • Reproducibility of Results
  • Sensitivity and Specificity
  • Software Validation*
  • Software*
  • Ultrasonography, Doppler / methods*
  • Young Adult