Desktop version

Home arrow Health

  • Increase font
  • Decrease font

<<   CONTENTS   >>

Premortem vs. Postmortem Body Imaging and Computational Anatomy of Liver


Many medical image interpretation algorithms for different organ systems have been proposed, and some of them are closer to practical use. These algorithms extract organs and/or important radiological findings, e.g., tumors or pneumonia, in a medical image to assist doctors faced with an overwhelming amount of data. A state-of-the-art CT scanner puts out several hundred section images per patient. CT is readily available and easy to use in Japan [173]. Because Ai does not involve radiation dosimetry, a larger number of image sections often result, compared with clinical examinations on living patients. Medical image interpretation algorithms, or computer-aided diagnosis, can help in this instance.

Medical image interpretation algorithms designed for living patients will work to some extent with cadavers, but might fail in some cases. Several types of differences are observed in postmortem CT images. For example, postmortem hypostasis (Fig. 4.27a) that is caused by gravity and increases attenuation or CT value is a significant finding in a CT image of a cadaver. Bronchial branches in a cadaver

Examples of Ai-specific radiological findings

Fig. 4.27 Examples of Ai-specific radiological findings (Original images courtesy of Dr. Yamamoto from Ai Information Center) are often not filled with air (Fig. 4.27b). Increasing CT values in lung tissue is a typical postmortem change (Fig. 4.27c). As time goes on after death, tissue contrast decreases (Fig.4.27d). Severe deformation of organs and bone fractures might be observed in an image of a cadaver (Fig.4.27e, f). Because conventional medical image interpretation algorithms developed for living patients are not designed to deal with these findings, they will fail in segmentation of organs and lesion detection.

This and following sections focus on computational anatomy and segmentation of the liver and lung. As with state-of-the-art segmentation algorithms for a living patient, a CA model or a statistical shape model (SSM)-based approach [174, 175] was employed. The SSM and algorithms are, however, reinforced to deal with postmortem-specific shape variation. This subsection starts with description of differences in shape between an in vivo liver and a postmortem liver.

<<   CONTENTS   >>

Related topics