During the course of prenatal development, human embryos undergo complicated morphogenetic changes. To understand and analyze such dynamic ontogenetic movements, it is essential to visualize embryonic structures in three- and four-dimensions, including the sequential changes associated with developmental stages.
Classically, drawings and solid reconstruction were used to demonstrate the three-dimensional (3D) structures of the embryonic body. Well-known examples are the wax models of staged human embryos housed in the Carnegie Collection. The wax plate technique of reconstruction was first introduced to human embryology by Born (1883) and later modified in the Carnegie Laboratory in Baltimore by Heard and his colleagues (Heard, 1951, 1953, 1957; Anonymous, 1979). Human embryos were embedded in paraffin or other embedding medium and were carefully serially sectioned. Wax plates were cut faithfully as the enlarged image of each section, and the wax plates were compiled to make the 3D embryonic structures. Some of the historical reconstructed models in the Carnegie Collection are now preserved in the Human Developmental Anatomy Center in Washington, D.C. (http://nmhm.washingtondc.museum/collections/hdac/index.htm). Based on those reconstructed models, numerous accurate drawings of human embryos were produced by skillful artists, including James F. Didusch, which added greatly to the value of the Carnegie Collection and provided detailed 3D information on developing human embryos (O'Rahilly, 1988). Thus, solid reconstruction from serial sections and fine drawings based on the reconstructed models were used as essential methods in classic embryology. However, such reconstruction and drawing methods are time-consuming and require special skills and, therefore, cannot be used in every laboratory.
During the past 20 years, thanks to the advancement of computer science, computer-assisted reconstruction of biological structures became available, which has enabled the reconstruction of various 3D structures from serial sections as well as the manipulation of reconstructed images as desired on the viewing screen. However, the process still needs serial sectioning of specimens and intensive labor to acquire the desirable precision and refinement of the images.
On the other hand, remarkable progress has been made in nondestructive imaging technologies such as computer tomography (CT) and magnetic resonance (MR) imaging. These techniques were originally developed as noninvasive diagnostic tools in clinical medicine, and technological advancement in recent years has enabled detailed imaging and 3D reconstruction of tiny biological structures (Smith, 1999). Such nondestructive imaging techniques do not require sectioning of samples to be examined, and accurate 3D images can be generated with the aid of computer software.
The recent advancement in MR microscopic technology has made it possible to scan and visualize relatively small samples, including mammalian embryos (Smith et al., 1996; Smith, 1999). MR microscopy (MRM) enables tomographic imaging of very small objects, and the digitized data can be manipulated to achieve 3D reconstruction of the samples (Smith, 1999; Haishi et al., 2001). Initially, there were some problems and limitations in MRM imaging. One was the resolution of the MR image, which previously was not sufficiently high. To obtain higher resolution, superconducting magnets (>2 Tesla) have been introduced to MRM (Matsuda et al., 2003), which significantly improved the quality of MR images. Another problem was the speed of imaging. Formerly, it took several hours to scan an embryo specimen that was only 10–20 mm long. This problem has been overcome by the invention of a super-parallel MR microscope (Matsuda et al., 2003), which made it possible to image several specimens simultaneously and facilitated the speed of imaging significantly. In addition, computer graphics (CG) techniques combined with MRM have made it possible to produce detailed 3D images of human embryos and to visualize the sequential morphogenetic movements occurring in the embryo.
Sequential 3D images of human embryos have the potential to serve as a reference and an important data resource for human developmental studies, similar to the Visible Human Database, which contains digitized serial cross-sectional images of the normal male and female human bodies (Ackerman, 1998, 1999) (http://www.nlm.nih.gov/research/visible/visible_human.html). They could also serve as models of network-accessible digital image libraries for medical research and education. Actually, various new projects have emerged from the Visible Human Project such as the invention of Internet-enabled visualization tools and navigation technologies for anatomy education and research, establishing comprehensive atlases of the human body for clinical imaging diagnosis, and virtual reality and surgical simulation for healthcare education and training (Ackerman et al., 2001). Similarly, 3D visualization of human embryonic development should be extremely useful in biomedical education, because such images and movie illustration of sequential developmental changes would help students understand the dynamic morphogenetic movements that occur three- and four-dimensionally in the embryo (Carlson, 2002). There is some commercial software for embryological education, but the embryo images included in that software are often scientifically incorrect and contain errors. Therefore, the illustration of human prenatal development based on actual human specimens is desired greatly.
In the present study, we have constructed a series of 3D images of human embryos, based on the MRM data of human embryo specimens in the Kyoto Collection, with the aid of CG techniques, to illustrate 3D structures and morphogenetic movements in human embryos. Using these 3D images, we then produced movies (Supplementary Movies, which can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat) to show the entire process of morphogenesis in human embryos, from fertilization to the completion of organogenesis. Furthermore, we developed a novel computer-assisted self-learning program of human embryology, using the acquired 3D images and relevant morphological data derived from the Kyoto Collection.
Constructing 3D Models Based on MR Microscopic Images
For embryos at Carnegie stages (CS) 13–23, MRM images of 20–30 well-preserved specimens were acquired for each developmental stage (e.g., Fig. 1A,B). By using the MRM specially prepared for sectional imaging of human embryos as described in the Experimental Procedures section, tomographic MR images of embryos were obtained as shown in Figure 1C. The image shown in Figure 1C is a sectional image acquired by a 1.0 Tesla(T) MR microscope (T1-weighted) represented by 128 × 128 × 128 voxels with the cubic size of 50 × 50 × 50 μm, but we are now routinely using a 2.35 T superconducting magnet that can yield MR images with a significantly higher resolution (see Fig. 7C,D).
The MR images acquired in a 1.0 T permanent magnet included some noise signals. In addition, there were some deformation and irregularities in the surface structure of the embryos caused by fixation and preservation (Fig. 1D). Furthermore, such small structures as fingers and toes were not demonstrated in sufficient detail. Therefore, some graphic modification was needed to visualize fine structures and to produce embryo images approximating the normal form (Fig. 1E).
Because there exists some individual variation in the shape and size among the embryos at a given developmental stage (Fig. 1A,B), efforts were made to generate the “standard” embryo images that represent the embryos at each developmental stage. Graphic designers and embryologists cooperated to carefully modify the details of the initial 3D images to generate the “standard” embryo images. Then final standard models were produced as shown in Figure 1F.
Illustrating Human Embryos at Carnegie Stages 1–12
At the developmental stages between CS1 (fertilized ovum) and CS12 (28 days after fertilization), embryo specimens were too small to acquire MR images. Therefore, their 3D models were reconstructed mainly by CG techniques based on their photographs taken in multiple directions as well as precise measurements. When applicable, histological sections of embryo specimens were used to access details of their fine structures. In the process of constructing CG models, close cooperation was essential between the graphic model designers and embryologists to generate accurate and biologically correct 3D models. Figure 2D,E demonstrates the two aspects of a constructed 3D CG model, illustrating a neurulating embryo at CS10, which was reconstructed based on the gross photographs (Fig. 2A–C) and serial histological sections of CS10 embryos.
Movie Illustration of Embryonic Development
When 3D images were constructed for each developmental stage (Fig. 3), they were continuously arranged to make movies (see Supplementary Movie, which is available at www.wiley.interscience.com/jpages/1058-8388/suppmat) to demonstrate the sequential morphogenetic changes on a viewing screen. The 3D images of different developmental stages were connected with each other by the “morphing” technique. When necessary, additional 3D images were constructed between the images for two adjacent stages to illustrate the morphological shape changes as smoothly as possible.
A sample movie (Supplementary Movie) illustrating a later stage of human embryonic development (6–8 weeks) can be viewed on our video (Supplementary Movie 1). The original version of the movie is in a significantly higher resolution and quality than the sample movie. The movie is approximately 5-min long and includes the developmental processes from fertilization to completion of organogenesis (the end of the 8th week after fertilization). Another movie was prepared to show the embryo images at the same magnification so that the actual growth of the embryo can be visualized (Supplementary Movie 2).
Illustrating Internal Structures
MRM images can also be used for constructing the 3D images of internal structures. Some organs like the brain, spinal cord, heart, and liver could be reconstructed in precise detail, although it was necessary to manually extract the image of relevant structures from each MR sectional image (Fig. 4). Because it was not easy to reconstruct very small structures such as the digestive tube and blood vessels directly from MRM images, their 3D images were constructed by the CG technique based on both MRM images and histological findings, with the aid of published illustrations in standard textbooks. Research is now in progress in our laboratories to digitally and interactively extract specific structures from sectional MRM images of embryos by constructing the “region-based contour trees” (Suwa et al., 2004).
Developing an Interactive Self-Learning Program
Using the CG models together with gross photographs, histological sections, and MR images, we developed a computer-based program for self-leaning and study review of human embryology. It consists of the “General Embryology” and “Special Embryology” sections, and the latter includes the description of the development of each organ system. Each chapter has the following contents: (1) “Lecture” that includes the text description of the sequential events occurring in the embryo or embryonic organ systems, (2) relevant images of embryos (gross photographs, histological sections, MR images, 3D CG images, and movies), (3) “Glossary” of embryological terms explaining key terms relevant to each section, and (4) “Quiz” for study review and self-evaluation.
In the “General Embryology” section, the morphological features of embryos are described for each developmental stage, with pictures and some quantitative data (gestational age, measurements, etc.). Some representative images of embryos are shown and then the names of the body parts are provided (Fig. 5B–D). Key words are explained in the “Glossary.”
In addition to illustrating the development of the whole body, specific parts of the embryonic body (e.g., the face and limbs) were extracted and demonstrated sequentially in 3D models to visualize the process of a specific organogenetic process (Fig. 6A,B). Further details of important structures (e.g., facial processes) and the topographical relationship of specific organs in the whole body are also illustrated (Fig. 6C,D).
If an individual selects a plane in an embryo image and clicks the button, then the histological section of the selected plane is shown (Fig. 7A,B). MR images are also shown for each developmental stage (Fig. 7C,D). Students can scan the embryo image on the viewing screen at any of the transverse, coronal, or sagittal planes and view the sectional MR images of the whole body as desired.
If an individual enters the “Quiz” section in each chapter, they can choose various sets of quizzes such as identification of embryonic structures, yes/no questions, fill-in questions, and multiple-choice questions. When an answer is given, judgment is automatically shown whether the answer is correct or incorrect. The student can choose the questions for the fundamental, standard, or advanced level. For the Quizzes with pictures, figures are first shown that have labels of numbers or alphabetical letters on a test image so that the student can identify the names of the structures in question. The answers are also evaluated automatically. A sample version of the self-learning program for CS13 and CS16 can be viewed on our Web site (http://www.cac.med.kyoto-u.ac.jp/ce.html).
Because the embryonic body undergoes dynamic morphologic changes during the process of organogenesis, it is essential to visualize and analyze the morphogenetic process at least in three-dimensions. 3D and 4D visualization of embryos is essential in both developmental research and the teaching of embryology. Recently, several new techniques have been developed for 3D visualization of various biological structures.
Confocal laser microscopy has been used widely in morphological research. It uses two-photon and multiphoton laser beams and captures fluorescent signals released from target structures. Confocal microscopy can image sections with thickness up to 2 mm, but this is not sufficient to image the whole embryo. Another disadvantage with confocal microscopy is that it can image only fluorescent signals, which makes it difficult to analyze unstained materials and specimens stained with nonfluorescent dyes.
CT has long been used in clinical medicine and is a powerful tool for analyzing hard tissues such as bones and teeth. It has been applied to imaging relatively small structures such as the fetal skull and can image their fine structures (Shibata and Nagano, 1996). However, it should be noted that CT needs X-rays, which makes it difficult to use the CT apparatus in routine laboratories.
Another microscopic imaging technique was developed recently, which is called episodic fluorescent image capturing (EFIC; Ewald et al., 2002; Weninger and Mohun, 2002; Rosenthal et al., 2004). Tissue autofluorescence is captured by a microscope and a CCD camera, and this method can yield accurate high-resolution images without special staining to visualize the tissues. To obtain 3D images, an embedded sample is serially sectioned using a microtome, and the 2D images of the block surfaces are captured after each section is removed. High-resolution images can be obtained by EFIC, but the hardest part with this technique is that the specimen must be cut into slices.
Another technique for cross-sectional imaging was developed by Sharpe et al. (2002). This technique is called optical projection tomography (OPT), and cross-sectional images are produced by scanning the specimen through the diode light beam coupled with a fiber-optic beam splitter. The apparatus is much simpler than CT and MR and can image specimens with a thickness up to 15–20 mm. A weakness of this method is that the specimens to be imaged need to be transparent to allow transmission of light.
MR imaging is widely used in clinical diagnosis, but there used to be some limitations in the MR apparatus that made it difficult to apply the technique to imaging very small specimens like embryos. A pioneering study to visualize human embryos by MRM was undertaken by Smith and his colleagues (Smith et al., 1996; Smith, 1999). Their study introduced the principles of microscopic 3D imaging of embryos. Human embryos can be imaged by an MRM equipped with a 1.0 T magnet, but the resolution is not sufficiently high to identify every structure in detail as shown in Figure 1C. This problem was overcome by developing a MRM equipped with a 2.35 T superconducting magnet, which enabled capturing detailed images of very small human embryos (Matsuda et al., 2003). Some of the images acquired by a 2.35 T MRM are shown in Figure 7C,D. The CG models described in the present study were generated based on MR images acquired by a 1.0 T MRM. Although the resolution of 1.0 T MRM images was not sufficiently high to identify every small structure, the images were useful when CG images of human embryos were generated. Smith and coworkers image tiny embryos in a 9.4 T superconducting magnet and have acquired excellent images (http://embryo.soad.umich.edu/index.html). It is ideal to use a higher-power magnet for imaging tiny samples such as embryos, but special skills are required to optimize the imaging parameters and to eliminate noise images. We have acquired images with sufficiently high resolution in a 2.35 T magnet (Fig. 7C,D) by using exchangeable solenoid coils whose diameter can be optimized for the sample size (Matsuda et al., 2003).
Another shortcoming with MR imaging is the speed of imaging. Usually, it takes several hours to scan an embryo of 20–30 mm crown-rump length, which was a serious limitation when applying the MR technique to imaging numerous samples. To overcome such difficulties, Kose and his colleagues developed a super-parallel MRM and thereby made it possible to image four embryo specimens simultaneously, which significantly facilitated the imaging process (Matsuda et al., 2003). In our laboratories, a project is ongoing to image a large number of human embryo specimens in the Kyoto Collection using a super-parallel 2.35 T MRM.
Some attempts have been made to visualize human embryo development three-dimensionally. A project was carried out at the US National Museum of Health and Medicine (NMHM), which was funded by the National Library of Medicine (Cohen, 2002). This project was called the Visible Embryo project and used human embryo specimens from the Carnegie Embryological Collection. Digitized 3D data sets of staged human embryos were completed and are now partly accessible on the web. After the Visible Human project was discontinued in 2003, the Virtual Human Embryo project was initiated by R.F. Gasser and R.J. Cork, which is supported by the National Institute of Child Health and Human Development (NICHHD) and the National Library of Medicine (http://virtualhumanembryo.lsuhsc.edu/). The project consists of the DREM (Digitally Reproduced Embryonic Morphology) and HEIRLOOM (Human Embryo Imaging and Reconstruction, Library on Line Media) components. The DREM databases contain the high resolution microscopic morphology of human embryos, with the level of each section image shown on 3D reconstructions. The HEIRLOOM project contains 3D reconstructions of the systems of the same Carnegie specimens and assembles an index of terms designed to provide the stage and section locations of many structures in the DREM series. Such embryo visualization programs would provide valuable image data sets of human embryos directly into classrooms of students learning embryology as well as into research laboratories investigating developmental phenomena and mechanisms.
With the recent advancement of scientific research and a greater knowledge of developmental mechanisms, embryology is becoming one of the increasingly important components of the medical curriculum. Another major justification for studying embryology in medical schools is to provide a basis for understanding the genesis of human birth defects (Carlson, 2002). However, the time devoted to anatomy and embryology teaching is decreasing due to the enormous proliferation of new disciplines in medical science (Yates, 1999; Paalman, 2000). Actually, the total course hours for embryology is mostly less than 20 hours (Drake et al., 2002), which is not sufficient for students to acquire a coherent disciplinary knowledge of embryology. In addition, only a small number of the medical schools in the United States (13%) provide a laboratory course of embryology, usually for only 4 or fewer hours (Drake et al., 2002). Therefore, it is important to use the time allocated to the subject efficiently by introducing new teaching materials and self-learning programs.
Various attempts have been made to develop novel teaching methods for embryology, including computer-assisted or Web-based learning (Watt et al., 1996; Aiton et al., 1997; Carlson, 2002; Puerta-Fonolla et al., 2004; Arroyo-Jimenez Mdel et al., 2005). In those programs, animations and morphing techniques greatly help students understand the dynamic developmental processes in which 3D morphological changes take place over time. In addition to animations for illustrating the sequential developmental process in human embryos and fetuses, their actual images could be viewed if photographs and MRM images are included. Because histological sections of human embryos are not easily available in many medical schools, digitized MRM images of human embryos could be incorporated in embryology teaching. The resolution of MRM images is approaching that of light microscopic pictures (Smith, 1999; Matsuda et al., 2003; Puerta-Fonolla et al., 2004), and could allow students to access to the tomographic and 3D images of human embryos at various developmental stages and to appreciate the spatial vision of human morphogenesis.
Furthermore, it is important for students to undertake self-learning and study review programs. In the present study, we have demonstrated part of our program for student self-learning of embryology, which includes lectures, case images, glossary of terms, and various types of quizzes. If excellent teaching materials are shared among medical schools, such a program could make students all over the world to access to invaluable human specimens and excellent teaching materials and could supplement conventional lectures and laboratory courses in embryological education.
Approximately 44,000 human embryos and fetuses have been collected and stored at Kyoto University over the past four decades, with the aid of several hundred obstetricians. In the majority of the cases, pregnancy was terminated for social reasons during the first trimester of pregnancy (the Maternity Protection Law of Japan) and healthy embryo generally were derived from the pregnancies. The pregnancies were mainly terminated by dilatation and curettage. Some specimens were derived from spontaneous or threatened abortions. Because the attending obstetricians did not examine the aborted materials, the collection of embryos was not biased by their outcome (normal or abnormal, live or dead, etc.) and the embryo collection can be considered to be representative of the total intrauterine population in Japan. Further details of the Kyoto Collection of Human Embryos have been described in previous reports (Nishimura et al., 1968; Nishimura, 1975; Shiota, 1991; Yamada et al., 2004). The embryo collection now comprises over 44,000 specimens and approximately 20% are undamaged well-preserved embryos. When the aborted materials were brought to our laboratory, the embryos were measured, staged, and examined for gross external abnormalities and signs of intrauterine death using a dissecting microscope. The developmental stage of the embryos (CS) was determined according to the criteria proposed by O'Rahilly and Müller (1987). In this study, human embryos at CS13 to CS23 were selected for MR microscopic imaging. For each developmental stage, 20–30 embryos were scanned, and representative cases were chosen for 3D reconstruction.
MR Microscopic Imaging
For MR imaging of human embryonic specimens, we first used a compact 1.0 T MRM equipped with a permanent magnet made of a high performance Nd-Fe-B magnetic material and a compact MRI console (54 cm(W) × 77 cm(H) × 60 cm(D), 80 kg in weight; Haishi et al., 2001). The console consisted of a PC computer, a radiofrequency (RF) transceiver unit (modulator and detector) with an RF frequency of 35–65 MHz, a three-channel gradient driver, and an RF transmitter. An RF coil unit was attached to the probe box from above, and a solenoid coil was placed at the center of the RF shield probe box. The diameter of the solenoid RF coil was optimized for the sample size (5–20 mm). In the present study, a four-turn solenoid coil was used for the 100 MHz signal excitation and detection frequency. The pulse sequence was a T1-weighted 3D spin-echo sequence (TR = 100 msec, TE 8 msec), and the resolution of the images was 120 μm3. The image matrix was consistently 128 × 128 × 128. When embryo specimens were scanned, each specimen was stored in an MR test tube filled with 10% formaldehyde solution. The diameter of the test tube was selected as close as to the size of the specimen to avoid the movement during image acquisition. The duration for image acquisition was 7.5 hr.
We also used a super-parallel MRM with a 2.35 T super-conducting magnet for imaging a large number of embryo specimens, which yielded MR images with higher resolutions. The details of its specification and imaging conditions have been described elsewhere (Matsuda et al., 2003).
Construction of a 3D Model Sequence of Embryonic Development
The body surface of each specimen was extracted as isoforms of MRM images using the software program MATLAB (http://www.mathworks.com/). The body surface was described in 3D polygon meshes. The isosurface was extracted using the threshold values that were variable, depending on the developmental stage and noises were reduced by 3D smoothing (Mizuta et al., 2002).
Initial 3D surface models thus constructed were modified to some extent based on the gross photographs of embryo specimens, because small structures such as fingers and toes could not be clearly visualized in 1.0 T MRM images (Fig. 1D). In addition, many of the specimens were deformed to a variable extent due to fixation and preservation, as seen in Figure 1D. Therefore, some cosmetic modification was needed to generate smooth-surfaced embryo images (Fig. 1E). Furthermore, considerable individual variation was noted in the posture and general contours of specimens among the cases at a given developmental stage (Fig. 1A,B). Therefore, we made further modification to eliminate the individual variation as much as possible and constructed a “standard” image of embryos for each developmental stage. Such modification was necessary for generating standard or representative images of embryos for educational purposes and for making a movie (see Supplementary Movie) to show smoothly the sequential morphogenetic changes occurring in the embryo. For making movies of embryonic development, the generated 3D images were processed using a CG software “SoftImage” (http://www.softimage.com/).
For movie animation of embryo development between CS13 and CS23, additional 3D images were constructed between two adjacent stages to illustrate the transformation of the body and parts of the body as smoothly as possible. For animation of development between CS1 and CS12, 3D models were constructed by CG artists, based on the photographs and histological sections of the specimens in our laboratory with the aid of some illustrations in standard textbooks. The images of major internal organs (the central nervous system, the digestive and respiratory systems, and the heart and great vessels) were also reconstructed from MRM images and histological sections, although some manual extraction of images of the organs was required.
Developing an Interactive Self-Learning Program
An interactive self-learning program of human embryology was developed using the photographs, histological sections, MRM images, and CG images of staged human embryos. Necessary text explanation and a glossary of embryological terms were also given. The program is composed of the “Lecture” and “Quiz” (practice) sections. In the “Lecture” section, text explanation is given together with embryo pictures for each developmental stage. On the photographs and other images of embryos, names were given for important parts of the embryo. The definition and meaning of embryological terms were described in the “Glossary” of terms. The “Quiz” section provides various types of questions such as yes/no questions, fill-in questions, multiple-choice questions, and matching pairs questions. The Web browsing of the program was developed using commercial software including Macromedia Dreamweaver (http://www.macromedia.com/), Adobe Photoshop, and Adobe Illustrator (http://www.adobe.com/).
Critical reading of the manuscript by Dr. Murray Smith, School of Medical Sciences, the University of New South Wales, Sydney, and helpful comments of Dr. Gary Schoenwolf, the University of Utah, are gratefully acknowledged. We thank Dr. Hiroo Imura, Former President of Kyoto University, for his support and encouragement. K.S. was funded by grants-in-aid from the Japanese Ministry of Education, Science, Sports and Culture, and by the BIRD grant from the Japan Science and Technology Agency.