Z Gastroenterol 2009; 47 - P193
DOI: 10.1055/s-0029-1241443

Computer-aided classification of colorectal polyps based on vascular patterns

JJW Tischendorf 1, S Gross 1, 2, R Winograd 1, H Hecker 3, R Auer 2, A Behrens 2, C Trautwein 1, T Aach 2, T Stehle 2
  • 1University Hospital Aachen (RWTH), Medical Department III (Gastroenterology, Hepatology and Metabolic Diseases), Aachen, Germany
  • 2RWTH Aachen University, Institute of Imaging & Computer Vision, Aachen, Germany
  • 3Medical School of Hannover, Institute of Biometry, Hannover, Germany

Background and study aims: The aim of the present study was to develop and evaluate a computer-based method for automated classification of colorectal polyps on the basis of vascularization features.

Patients and methods: In a prospective study with 128 patients who underwent a zoom NBI colonoscopy, 209 detected polyps were imaged and subsequently removed for histological analysis. The proposed computer-based method consists of several steps: preprocessing, vessel segmentation, feature extraction and classification. The results of the automatic classification were compared to those of human observers, who were blinded to the histological gold standard.

Results: The consensus decision between the human observers resulted in a sensitivity of 93.8% and a specificity of 85.7%. Making a safe decision in case of a discrepancy yielded a sensitivity of 96.9% and a specificity of 71.4%. The overall correct classification rates were 91.9% and 90.9% for the consensus and safe decision, respectively. With suitable parameterization for the computer-based approach a sensitivity of approximately 90% and a specificity of approximately 70% was achieved. In this case the overall correct classification rate was 85.3%. To establish a direct comparison of the human and automatic results, we selected computer-based results with sensitivities identical to the results of consensus (93.8%) and safe decision (96.9%). The computer-based classification resulted in a specificity of 61.2% for a sensitivity of 93.8%, respectively 53.1% for a sensitivity of 96.9%.

Conclusions: The study has shown that automatic classification of colon polyps based on NBI vascularization features is feasible. Further research needs to improve the promising performance of the automated classification system.