# Computational Anatomy and Segmentation of Postmortem Liver

## Introduction

As discussed in Sect.4.4.2.3, increasing training labels that describe the postmortem-specific shape is essential to improve the performance of a postmortem SSM. This section presents a method that solves the abovementioned problem by synthesizing postmortem liver labels, which is inspired by synthesis-based learning [181]. Performance comparisons of SSMs trained using different sets of synthesized postmortem liver labels are presented, followed by a proposal for a postmortem liver segmentation algorithm.

## Postmortem Liver SSMs Using Synthesized Postmortem Labels

Three transformations are developed to simulate the shape deformation from in vivo livers to postmortem livers. They are categorized into a geometrical transformation F_{a} and two statistical transformations, *F _{T}* and F

_{tr}. Details of the methods can be found in [176].

In this study, the transformations yielded five different sets of synthesized postmortem liver labels, or*~D _{T}* from F

_{t},

*~D*from F

_{TR}_{tr},

*~D*from F

_{A}_{a}, ~D

_{at}from F

_{a }followed by F

_{t}, and

*~D*from F

_{ATR}_{a}followed by

*F*respectively. Five postmortem liver SSMs were trained using combinations of the five synthesized liver label sets with original postmortem liver labels,

_{tr},*D*. The relationships between the five SSMs and the five synthesized label sets are summarized as follows. Note that the training label sets are shown in parentheses:

- •
*SSM*model (D and_{D+T}*D*)_{T} - •
*SSM*model (D and_{D+TR}*D*_{TR}) - •
*SSM*model (D and_{D+A}*~D*)_{A} - •
*SSM*model (D and_{D+AT}*~D*)_{AT} - •
*SSM*model (D and_{D+ATR}*~D*_{ATR})

In addition, three conventional SSMs constructed solely from original labels were prepared for comparison.

- • SSM
_{D}model (D only) - • SSM
_{L}model*(L*only) - • SSM
_{D+L}model (D and L)

Figure 4.30 summarizes the performance of the three conventional and five proposed SSMs in terms of the sum of generalization and specificity explained in the previous section. It was found from the figure that most of the proposed SSMs learnt by synthesized postmortem liver labels outperformed conventional SSMs trained without synthesis-based learning. In particular, the D + T model achieved the highest

**Fig. 4.30 ****The sum of generalization and specificity (Fig.7c of [176])**

score which demonstrated the superiority of the proposed statistical transformation for synthesis-based learning.