MT - 11.08


On Missing Organs in Multi-organ Segmentation with Atlas-Guided Approach


Miyuki Suzuki

Oral Defence Date: 



HH 301


Profs. Kaz Okada. Hui Yang and Ilm1 Yoon


Atlas-guided segmentation is one of the current state-of-the-art multi-organ segmentation algorithms. However, previous studies have not discussed and handled cases with surgically removed organs (missing organs). From the clinical aspect, demands for handling such cases are high because surgical removals are common procedures for cancers or organ failures and those populations need more attentions for follow-up examination. Therefore, the clinical applications could not ignore the populations. This project aims to analyze the anatomies of the missing organ cases, cope with those cases in atlas-guided segmentation and improve segmentation accuracy for such difficult cases. Two novel data-driven missing organ detection methods are proposed and by combining the methods, it achieves 93% of accuracy and 0.92 of AUC (Area under curve) of ROC (Receiver operating characteristic) analysis. Integrating the combined missing organ detection into atlas-guided segmentation results in improvements in segmentation accuracy and automation of the overall procedure.