Intelligent Microscopy Systems Kristóf Kovács University of Veszprém, Dept. of Silicate- and Materials Eng., Central Laboratory Introduction Development of microscopy methods has gained new pace with the birth of user friendly computers and operating systems. Interaction between computers and microscopes was first restricted to processing of structural and compositional data as well as image information. The availability of high speed analog to digital converters has rendered data acquisition to one of the tasks performed by computers. Today processing contains information feedback and the burden of microscope operation and setup is also taken over by the hardware and software invisibly operating behind the instrument panel. Artificial intelligence, adaptive systems, and cellular neural networks open previously unforeseen perspectives to end users and to engineers involved in microscope development as well. Imaging Systems Basic operation principles of microscopes are identical to all analytical instruments. These procedures have some features in common: properties (most frequently structure and composition) of the unknown matter are revealed by properly exciting the specimen. Signals produced during the interaction between the exciting energy and specimen will provide all information relevant to the properties. Various analytical procedures differ from each other in methods generating the excitation energy, interactions with matter, as well as signal processing. Analytical microscopy provides additional information through the point to point correspondence between specimen and image: visual image (quality) and analysis results (quantity) are observed at once in the very same instrument and hence provide complex morphological and compositional data. Artificial Intelligence By definition intelligence is the adaptive answering capability to new and unknown situations. “Crystalline” and “fluid” intelligence (knowledge and flexible thinking) are supporting and enhancing each other. Knowledge provides “ready-to-eat” schemes among which one selects with flexible thinking: Unknown situations are solved by creatively using the existing building blocks and adding new schemes to the existing ones. Artificial intelligence is using computational methods to simulate features of human intelligence such as learning, intuition, deduction, self-correction. Recent methods pass the limits of ordinary computers by setting up cellular neural networks in order to model the structure and function of human neurons. Intelligent systems attempt to follow the human intelligence as close as possible by using the tools of artificial intelligence. With the available tools and techniques there is still a lot to do in terms of achieving good results, however a significant development is observed in visual applications such as image processing, pattern recognition, artificial vision etc. Connection of microscopes to computers Due to the operation principles conventional microscopes provide analog images. These images are easily viewed by visual observation, however quantitative evaluation and automation of individual evaluation steps is difficult, requires sophisticated experience and is time consuming. The computer behaves like an elegant servant [1], but the master is still spending most of his energy to build up the hardware and software connecting the microscopes to computers. Conventional light microscopes and transmission electron microscopes provide images as a whole at once using parallel imaging. These analog images are interpreted on the whole image plane, the corresponding digital image is valid for individual image points, pixels only. The naked eye operation is also based on parallel processing although the image is interpreted at discrete points, corresponding to the nerve cells. Parallel image information shall be converted to serial data flow and analog data of each pixel converted to digital information before processing by a conventional computer. Operation of scanning electron microscopes and scanning probe microscopes is serial, i.e. only one image point is produced at a time, the next image point is observed later, shifted by the sampling time. This sort of serial imaging is almost ready for computer processing, there is only one signal to handle. Sampling rate and optimum resolution [2] are determined by taking into account the resolution of human vision and resolution of microscope imaging system (information theory, Shannon). Not surprisingly the result is reflecting the title of von Neumann’s classic: How to make a reliable system from unreliable parts. Application examples There are many examples to demonstrate all steps leading from conventional microscopes towards intelligent measuring systems. One cannot declare an “intelligent” system being superior to that of without “intelligence” since by virtue both can be equally excellent and useless. Keyword is quality. Artificial intelligence serves our ease only, and in spite of providing many additional, previously not available information plays a secondary role. Some of the imaging systems such as scanning probe microscopes or confocal microscopes cannot operate without an on-line computer. Scanning electron microscopy and on-line energy dispersive X-ray analysis might act as a role model for other techniques by establishing intelligent operation as a whole. A first step on the long way might be the gathering and processing of X-ray maps [3]. The interface connecting the X-ray analyzer to the computer is rather simple, does not even contain analog to digital converters, and in spite of this it will select automatically the optimum sampling rate as a function of magnification and count rate, as well as averaging of successive frames. State-of-the-art scanning electron microscopes are all equipped with some sort of feedback providing many automated functions such as autofocus, automatic astigmatism correction, etc. Multi-user laboratories provide equal access to all users through networked intelligent microscopes and provide a real expert system database [4]. There are endless possibilities provided by the Internet. Users and instruments located thousands of miles far away from each other are easily connected through the high speed connection lines establishing intelligent operation [5]. References [1] Ralston: A számítástudomány elmélete, Műszaki Könyvkiadó, Budapest, 19 [2] Russ, J.C.: Computer Assisted Microscopy, Plenum Press, New York, 1990, p23 [3] Kovács, K., Mátz, A.: Proc. XIIth Int. Congr. Electron Microscopy, Supplements, Seattle, 1990, p7 [4] Cross, R.: Proc. 4th. Multinational Congress on Electron Microscopy, Veszprém, Hungary, p9, publ. University of Veszprém, Ed. K. Kovács [5] Zaluzec, N.J.: Proc. 4th. Multinational Congress on Electron Microscopy, Veszprém, Hungary, p19, publ. University of Veszprém, Ed. K. Kovács Copyright © 2000 |