Model report

DRAFT: Creation of Anatomically segmented canonical human model data -- starting materials for customized anatomical models for the public

INTRODUCTION

 * The recent advancement in information technologies are reshaping the whole healthcare communication; one way dissemination among stratified readers are being reshaped into sharing of information across patients and experts with dense bi-directional communications and information archiving -- collectively called health2.0.
 * Since thousands of years ago (Calkins, Franciosi, and Kolesari 1999)in healthcare communications, anatomical illustrations and diagrams are frequently accompanied to fill gaps in vocabulary, convey spatial relations, and to enable thinking with the data (Card, Mackinlay, and Shneiderman 1999),(Hienert et al. 2011),, similarly to geographical maps. ; anatomical diagrams are the hall mark of healthcare communications.
 * Accordingly, in health2.0 environment, the needs for anatomical diagrams assisting communication or making sense of archived information will be greater than ever.
 * With the advance in digital technologies, anatomical diagrams have also undergone various enhancements.
 * Most of illustrations are born digital with drawing applications and disseminated as image files in digital books.
 * Moreover we have spectrum of 3D human anatomical interactive atlases for education or surgical simulation available commercially or freely(ref).
 * Nevertheless, two things remained unchanged: firstly, only experts with anatomical knowledge and special skills can create good diagrams; secondly, reuse of or modification on elaborated diagrams are usually prohibited by the copyright.
 * These two features limits the production of anatomical diagrams to small population of experts and they precluded possible bigger contribution to digital communication and innovative development of medical drwaing itself.
 * From such point of view, we started to build a set of 3D anatomical human model with new functions that allows it to serve as starting materials for further creations and innovations in downstream.
 * Here we report mostly about the methodologies in data creation and the features of the data to inform the downstream creators in the community.

Rationale (can be a part of introduction)

 * To be used as materials for creation of models and images for various purposes, we provided the following features to the model data.


 * 1) To be served in wider application, we made a model as canonical model representing the common features in highly variable physical structure among people.
 * 2) To allow extensive customization in 3D model at various levels, we made the data as a collection of independent 3D wire frame model.
 * 3) To help selection of body parts by users, file names of body parts are mapped on to conceptIDs in anatomical ontology -- a whole-part and is a kind of relation data among all anatomical concepts.
 * 4) For the ease of downstream manipulation of the models, we disseminate the model after transforming into wavefront obj format from the original voxel format which is more tractable in creating 3D forms.
 * 5) To eliminate licensing effort by users, we used Creative Commons license to clearly show some rights we retained.

Modeling plan

 * To create one canonical body in 3D shape is a challenge because, canonical model exists only as a knowledge represented across desperate materials in various forms; we do not have a template nor a set of physical specifications of canonical human body.
 * So, our task was to gather physical specifications for canonical body parts, essential features among their relations and model those knowledge into one coherent set of 3D models.
 * We did this process as follows....
 * 1) Use real instance based skeleton model and align them as textbook says.
 * 2) Using skeleton as basis, we create muscle and tendinous tissues adhering to skeleton.
 * 3) Gather clinical 3D images or sectional atlases as templates for internal organs and augment the internal structures of those organs according to the textbook and atlas descriptions.
 * 4) With proportional scaling up and down within the range of reported normal sizes, we made internal organs to fit in the cavity made by muscle and skeleton.
 * 5) We made vessels and peripheral nerves that connect those organs and tissues mostly from full scratch as according to the textbook descriptions and atlases.

Modelling environment
reasons は？
 * We used Free Form Modeling Plus (FFMP) (Geomagic), and Intage volume editor for volume rendering of digital imaging and communication in medicine (DICOM) files, among various applications and appliance that support creation of 3D models from several reasons.


 * While our modelling proceeds, we encountered various limitations in original software in recommended environments.
 * mostly because free from is not supposed to be manage extensive numbers of files such as anatomical body parts. ? 日本語でいいので以前三橋さんが外注してくれたカスタム化の内容を
 * In addition, it took some minutes to change the tools within the application when many components are the objects for the tools.
 * Accrodingly, we enhanced the machine environment from recommendation by the manufacturer to; windows 7 professional, CPU Intel(R) Xeon(R)5680 3.33GHz X 2, 48.0GB memory, 64bit, Quadro6000x2　for graphic cards.

Skeleton
Mark for the sagittal plane (fig.3) - the gravity line which pass through the dens of the axis cervicothoracic junction, the thoracolumbar junction, the hip joint, the knee join, and the ankle joint -the angle of 60 degrees for the anatomical conjugate - 172cm from the origin for stature For the coronal plane (fig.4) -the normal physical ratios. -the line of the midsagittal plane which passes between the middle point of the clavicle, and the middle of the superior anterior iliac spine and pubic joint. - the angle of 126 degrees of between the hip joint and the femur. For transverse plane (fig.5) - about 60 degrees for the acromioclavicular joint - the angle of 23 degrees between the transverse axis of the lower tibia and the transverse axis of the lower tibia - the angle of 45 degrees for the longitudinal axis of the foot of the right and the left foots
 * We started to create skeletal structure because they provides solid framing of the human machine made up with flexible materials in movable anchoring and suspensions.
 * Because of this nature, skeletal system provides references to describe normal positions of other organs in anatomical textbooks; we used such descriptions as sources of canonicity.
 * We started to make bony parts one by one.
 * To make them, we used the bone images from the front, side and top views in a textbook.
 * The images were imported to sketch planes from each direction, X, Y, Z axes in FFMP, and outlined by the Freehand Curve (figure1).
 * The cube made from the Basic Shape was cut away from the closed profiles on sketch planes by the Wire Cut from each direction.
 * The rough shape of the bone part was accomplished at this point. (figure2).
 * The details of objects with FFMP are visualized by embossing irregularities, so the Sculpting and the Deforming tools were used to make details such as fossa, facet crest, and groove of the bone.
 * The shapes of the skull’s parts such as a sphenoid bone were very complicated, so it was difficult to create it from a scratch. We used anatomical mock-up by scanning to look at contours as the base of the model.
 * The scanned mock-up needed to remodel to fit the model and to revise distorted deformation of each part of the skull. The cranial bones were commonly asymmetrical in human, but we created with idealized, symmetrical skull bone.
 * With the same method of making bone parts, we created the teeth and added the portions such as the styloid process to accomplish the skull.
 * After the bone parts were created, we assembled them and built a skeletal structure according to descriptions over various sources.
 * We set up sagittal, coronal and transverse planes from the origin to mark the insides of the bone which were significant to fix an upright posture.
 * By using the Reposition piece, we moved all skeletal parts to the certain position according to the above marks. Then, symmetric bones were mirrored by the sagittal plane and the skeletal structure was fixed.

The Muscles

 * After the skeletal structure was finished, we created skeletal muscles in order to make the boundary of cavities for internal organs.
 * All muscles were made from the scratch.
 * Based on textbook descriptions we marked the origin and the insertion area for each muscle in the skeleton models and connected the these area to make a set of boundary curves for a quadrangle (Fig.6).
 * A quadrangle curve could make a patch and it could convert a piece of clay with a certain thickness according to the every muscle by the Convert to Clay. The details of muscles were created by embossing with Sculpting and deforming(fig.7). (??)

The gastrointestinal tract (GI tract)

 * As physiological ontogeny goes, we started to simply create one pipe.
 * Then, we draw a sketch curve from the esophagus to the large intestine on each sketch plane from the front and the side views with the Freehand Curve.
 * By using the function "Intersect Sketches", a 3D curve was created at the intersection of two projected sketch curves which contained characteristics of both sketches (fig.9).
 * Then, 3D curve was converted to a piece of clay tube with the Ridge.
 * With this pipe clay as an index, the detail of each organ such as the stomach, duodenum, and intestine was elaborated by deforming and sculpting repeatedly (fig10).

The Heart

 * We used several resources　to create heart.
 * We scanned a plastic model for anatomical education.
 * We deformed the scanned heart to meet textbook descriptions relative to our skeletal model.
 * The first step was to modify the heart to fit the outline of the rib cage from the front view.
 * We draw curves according to descriptions in textbooks.
 * The height of the cardiac upper limit of right side is as the same height as the third costal cartilage and its left side is the second intercostal space.
 * The right margin of the heart passes range from the right third costal cartilage to the right sixth costal cartilage.
 * The left margin of the heart begins from the second intercostals space to the fifth intercostal space near midclavicular line where the apex of the heart is located.
 * The inferior border of the heart lengthens from the end of the right sixth costal cartilage to the apex of the heart (fig.11).
 * Then, the heart was deformed to fit the outline.
 * We could not find skeletal indices for the lateral view in textbooks.
 * Therefore, we referred to sliced CT images by thoracic vertebrae in the textbook, Human Sectional Anatomy (Adrian 2001).
 * The images were imported to the sketch planes on the transverse sections, and traced outlines of the heart respectively with the Freehand Curve (fig.12).
 * Then, contours of the heart were tugged and deformed.
 * To compensate for the form among vertebrae, it was beneficial to make a rough shaped solid from the traced curves.
 * The contour of the heart was roughly finished at this point.
 * Then, we made cavities of ventricles and atriums.
 * Since the thickness of the muscles in each chamber was roughly determined, the cavities were made by removing the thickness of the wall from the solid heart.
 * The following rules are applied to make chambers.
 * The muscle of the left ventricle is three times thicker than the right, and it is usually about three millimeters for the right and about nine millimeters for the left.
 * For atrial walls, it is usually about three millimeters for left, and two millimeters for the right.
 * To confirm the thickness of the walls, cross sections were verifying by activating the Cut Away from menu bar (fig.13).
 * At the same time, the positions, angles and dimensions of the valves by ribs were also ascertained from frontal and transverse views.
 * It became clear anatomical restriction in this way, so we elaborate details of the heart with scarping and deforming (fig.14).

The Brain

 * We used some resources to make brain.
 * We made brain from scratch based on　the images of cross sections and published segmented CT data.
 * The telencephalon and the cerebellum were created based on the SPL PNL Brain Atlas (Talos 2008), and the diencephalon, the allocortex and the brain stem were created to be base on the images of cross sections in The Human Central Nervous System (Nieuwenhuys 2008).
 * Since the model from the SPL PNL Brain Atlas did not fit well with the existing skull model we made, we retrofitted them by deforming the model with the Tug.
 * For modeling the diencephalon, the brain stem and the allocortex, the images of the coronal and the horizontal sections in specific points of the brain were imported on the planes (fig.13).
 * The outlines of the slices were traced on the planes and converted to 3D curves.
 * The curves enclosed by a set of boundary curves made possible to create a patch.
 * The each path by the plane stitched together to make the solid and it was converted to clay.
 * Although the white matter had not been made yet, we completed to create the detail of the cerebral cortex and the deep parts of the brain at this point (fig16).
 * To create the white matter, we duplicated the solid cerebrum, and then the cerebral cortex, allocortex and the diencephalon were removed from it.
 * The details were created with the same methods we did on the other organs. (fig17).
 * Finally, we compare the model data with structural features in descriptions and illustrations in neuroanatomy textbooks to augment and modify the detail of brain model.

Tips for FFMP users
- Repeat editing 3D curves when a project contains more than three hundred independent conoutrs. - Combination use of the Curves with the Deform, the Tug Area, and the Patches. To avoid trouble with a file, it is the best to reduce the number of clays to lessen the file size. -Close the number of open eyes on the object list (OL) to reduce the memory usage. -Lower the resolution of the Clay Coarseness. -Delete unnecessary planes, curves, and patches. -Delete the small particles of clay which remain during splitting a piece of clay. -Smooth the surface of the clay to minimize surface area. -Convert objects which you finished modeling from clay to mesh
 * In FFMP, a model is made up of voxels which are grains like sand. Generally, the data made of them become much heavier than the data based on surface　modelers.
 * This is a disadvantage to work with FFMP, a voxes based modeler, especially managing many independent contour files.
 * As a file in one project gets heavy in number?, FFMP becomes unstable and a file operation becomes slow.
 * To enhance operability of the modeling, we examine the tendencies of the modeler as a user because we can not look into the commercial sources.
 * Among various programs--refereed to "tools" in FFMP --those with much computational complexity such as the "Deform Clay" and the "Emboss" has tendencies to freeze and crush as the numbers of files in a project gets bigger.
 * The Curves tool also gets heavy and causes them under the following conditions.
 * However, it is necessary to put one file with many parts to look at the connection between organs and the overall balance as much as a work advances.
 * In such case, it is able to reduce the size of files by the following work.

Model dissection

 * We set the the granularity of model dissection that determine the freedom in customization to the standard anatomical terminology of international society of anatomy (Terminologia Anatomica: TA) (Federative Committee on Anatomical terminology 1998).
 * But, as a concept source to label the polygon files we depended on conceptIDs in an anatomical ontology, Foundation Model of Anatomy (FMA) of Washingtong University (Rosse and Jr 2003)from several reasons which will be described elsewhere.