Project 1: The Mouse Brain Library Project 2: Internet Microscopy (iScope) Project 3: Neurocartographer and Segmentation of the MBL Project 4: The Neurogenetics Tool Box


















Principal Investigator/Program Director Williams,Robert W.

  Aim 2: Robotic Slide Handling Systems


Dr. Nissanov and colleagues at Drexel University have spent more than two years designing and building a unique and highly effective slide-handling system that works in conjunction with a motorized microscope stage. This relatively inexpensive slide-handling system consists 1 to 8 carousels, each of which can accommodate 50 2 x 3 inch slides. Slides are delivered from the carousel to a transfer belt that reliably loads slides onto the stage. An optical feedback system ensures that slides are mounted in the correct orientation. In Year 01 we will build a second, improved carousel system and motorized X-Y stage to work in conjunction with one of the inverted microscopes described in Aim 1. This system will allow us to get off to a very rapid start in acquiring high-magnification images stack for the MBL collection. An important part of Aim 2 is to assemble a database of slides and coordinates that will work smoothly in conjunction with both the MBL and the carousel drivers. Drs. Williams, Rosen, and Nissanov will collaborate closely in designing this Slide-and-Coordinate database. Our goal is to deliver web microscopists to any site on any slide in the MBL with a precision of 200 µm. The carousel system will be used throughout the grant as part of our automatic z-axis image acquisition system.


By year 03 the MBL will have grown to more than 5000 slides. We will have gained a great deal of experience with the microscopes, the slide databases, and patterns of use. We will be ready to experiment with a higher-capacity and more flexible slide-handling system. For this reason, in Year 03 we request funds to buy and adapt a commercial robot system to deliver slides to any of four microscope stations and from an arbitrarily large collection of slides. We will need to develop a new slide mounting system. Pairs of slides will be mounted in rigid plastic holders. These holders will improve robotic handling, slide delivery and registration to the stage; in addition, we can apply machine-readable bar codes to the holders. Our goal is to assemble a system that can be easily extended from the 1200 slides now in the MBL to a collection of more than 10,000 slides. The slide system will include a master rack that securely holds the majority of the collection and a cache rack for recently and frequently accessed slides. Our target is to load slides in the cache in less than 1 minute and to load slides in the main rack in less than 2 minutes.
We have constructed a prototype of a robotic slide feeder. It operates with very low rate of jamming (less than 1 jam/100 slides) and no jam result in slide damage. A duplicate of this system will be installed during year 1. It consists of a slide jukebox, a linear conveyor, and an automatic slide clamp. The system is designed to be compatible with both a immobile stage of the macroscopic imager and the 3-axis-motorized stage of the microscope imaging stage.
The jukebox is reminiscent of carousels used in 35mm photographic slide projectors, except that the slides exit and reenter the carousel radially and the ejector mechanism exerts a force from the center outward. The carousel is mounted vertically on a central horizontal spindle. The stationary outside sleeve is fully enclosed except for a single slot while the internal sleeve has a central opening for the ejector arm. The spindle can rotate about its axis to position the desired slide behind the front slot so that an ejector arm from the spindle can push out the slide. Five carousels can be placed on the spindle; each accommodates 60 slides to give a total of 300 slide slots. Translation along the axis of the spindle selects the different carousels.

The ejector is a small lever, actuated by Nitinol (a NiTi shape memory alloy) that pushes the slide from the inner radius of the carousel outward. The linear conveyor (motorized rollers) draws the slide the rest of the way from the carousel and places it partially on the slide clamp, which is mounted on the imaging stage (microscope stage for microscopy). Another small Nitinol-actuated lever drives the slide the remaining distance, gently pressing it against the end stop pins of the slide clamp. A second lever than presses it into a fixed position. Three position sensor obtain accurate (resolution 5 microns) measure of the location of two of the slide edges. From those the virtual upper left corner, the slide coordinate origin, is computed.
After imaging, the slide is released. Another Nitinol lever mounted on the mounting frame ejects the slide, the rollers draw it out of the clamp and drive it most of the way back into the carousel. A final Nitinol lever guarantees the full reinsertion of the slide into the carousel, and the cycle is complete.

A dedicated microcontroller board controls the jukebox, the conveyor, the clamp and all of the levers. A network of optical and other sensors will permit the microcontroller to monitor the progress of a slide through the system. A serial datalink allow scommands and data to be exchanged with a desktop computer, allowing synchronization of the slide manipulation system with the image acquisition system.
To monitor accuracy of the microscope guidance system, we are constructing a slide containing a monochrome 2D-subperfect map is sent through the system. These maps are an A-ary array of size k x l in which each element is one of A possible values and in which all submatrices of size v x w appear exactly once [13] and thus spatial position within the array can be precisely determined by a subfield view. For the present application, we will set A = 5, k = 338, l = 558, and v=w=3. We have developed algorithms to automatically generate these codes [18, 19, 20]. They will be applied to a slide as an array of 100 _m dots with 150 _m dot center to dot center spacing of 5-different optical densities on high-resolution photographic film bonded to a histological slide. In any 10x view of the slide, a 3 x 3 submatrix will be visible. The slide will be produced through a 10:1 photographic reduction of a high-resolution transparency produced by a commercial ink-jet printer. Using already developed software [13, 16, 18], automated processing of acquired images will determine actual acquisition coordinates to a precision of 10 _m. Accuracy of the stage mechanism can be evaluated by comparison of the location to the coordinates commanded. While the performance of the entire guidance process determination of slide coordinates from low resolution images, maintenance of correct coordinates through image rotation as is necessary during alignment, and the high-resolution image acquisition can be tested by passing the slide through all the stages from jukebox feed to the low resolution imager to acquisition of the high-resolution images.

Experimental Results and Methods

During Year 02 and 03 a new robotic system will be designed and built that will accommodate as many as 20000 slides. Instead of a carousel jukebox, the new system will consists of a vertical rack of slides. Each 4 slides will be mounted in a plastic frame. A robotic arm riding on a railing system will move to the correct position in the rack and retrieve the appropriate slide by latching to the slide frame. All 4 slides will be locked into a fixed position on a large XYZ stage mounted under the microscope.

Next Topic

  Aim 3: A Web Archive of Image Stacks and Software for Stereological Analysis.