Mohammed Abdulrahman Alshaikh, Presenter: Nothing to Disclose

Muller Arnaud, Abstract Co-Author: Nothing to Disclose

Capucine Micol, Abstract Co-Author: Nothing to Disclose

Elsa Guillot, Abstract Co-Author: Nothing to Disclose

Pierre-Jean Valette MD, Abstract Co-Author: Nothing to Disclose

To evaluate the performance of a semi-automatic liver segmentation software (Philips, Intellispace Portal) according to the anatomic Couinaud model. The software proceeds by automatic delineation of the liver contour, followed by manual identification of 9 easily recognizable anatomical (vascular and ligamentous) landmarks, and then provides visual representations of the segments

prospective study consisting of localizing 180 tumors into 42 liver enhanced CT scans (336 hepatic segments). Images were analyzed by 4 observers divided in 2 groups: 1-two juniors (3rd year of medicine school) unaware of the Couinaud model and using the software, 2-two seniors (over 5 years experience in abdominal imaging) without the software. In addition, a gold standard was established by an expert in abdominal imaging supported by the software and taken as adjudicator. Kappa statistics tests were used to calculate interobserver agreement between participants into each group (reproducibility), and also the probability of agreement between groups of readers and adjudicator (precision)

1-Kappa values were 0.84 between juniors and 0.88 between seniors indicating a very good reproducibility into each group. 2-Kappa values ranged from 0.80 to 0.81 (substantial to almost perfect agreement) between juniors and gold standard; from 0.84 to 0.91 (almost perfect agreement) between seniors and gold standard. 3-Probability of agreement between all observers and gold standard decreased for some localization: tumors located S1, S2, S6 for juniors and in S3 for seniors. 4-Probability of agreement was independent of tumor size

The concept of liver segmentation based on few landmarks easily recognizable by any CT reader provides adequate and reproducible localization of tumors into the liver. Such approach may be useful for non expert radiologists while facilitating visual representation and treatment planning for surgeons

The propsed computer aided method provides simple anatomical liver segmentation and helps non expert radiologist for better localization of liver tumors aiming to facilate surgical treatement plans . 

Alshaikh, M, Arnaud, M, Micol, C, Guillot, E, Valette, P, Localization of Liver Tumors according to the Couinaud Segmentation: Results of an Innovative Computer Aided Method.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13018758.html