Path B: Computing at Vanderbilt

Meneely used the IBM 1620 in the VU Computer Center to analyze gamma ray spectra from the Whole Body Counter with data entry via IBM punch cards. After the 1620 was replaced, the IBM 7072 was used to analyze punch paper tape or magnetic tape encoded records to deconvolve overlapping spectral data to reveal the amount and kind of radioactive elements in a patient or an inanimate sample. Multichannel analyzers were used to buffer spectral data from the whole body counter, followed in the late 1960s by data capture and analysis in one of a series of DEC PDP 9/15 we acquired for image acquisition and analysis of data from scanners and gamma cameras.

→ Image analysis allowed the enhancement and extraction of features of interest from static and dynamic 2D (planar) and later 3- and 4-D images. Interest in the use of computers dramatically increased in 1964 when David Kuhl demonstrated the first 3-D (tomographic) nuclear medicine patient images.. This was immediately recognized as a revolutionary development. For this application, the computer was for the first time recognized as essential as it was needed to acquire, store, reconstruct and display the multi dimensional data. This greatly stimulated  interest in coupling  computers to medical imaging devices. The development of mathematical algorithms followed taking into account physical and statistical factors needed to compensate for instrument and patient related sources of distortion. These efforts were fully demonstrated using Hounsfield’s system and Cormack’s algorithms   in 1973 used computers to produce the first X-ray computed tomography 3-D images. Note that this was 9 years after Kuhl had demonstrated the first 3-D nuclear images.

Vanderbilt’s first abortive tomography attempt used data collected in a computer from the Picker scanner, with the detector oriented toward the side of the patient’s head. The patient was seated on a barber chair and data collected in the PDP9 computer from successive lateral scans at repeated 6-degree  intervals until 180/360 angular projections were recorded. The goal was to determine locations of stroke defects, but reconstruction methods were not yet available, and our initial attempts in the late 1960s failed. This was done as a part of a Vocational Rehabilitation Agency stroke research grant held by Dick Blanton, a brilliant Vanderbilt Experimental Psychology colleague and collaborator.

→ Nuclear medicine computing: In 1968, we acquired our first Anger Camera. Shortly thereafter we acquired our first programmable computer, a DEC PDP9 with funds from a Computer Assisted Instruction (CAI) grant. The computer capture of data from nuclear devices preceded one’s ability to capture X-ray data, as nuclear data rates were much slower and  more easily captured in real time than from  X-ray systems. For nuclear image capture, we used the patch board programmable data acquisition system that allowed us to adapt the system for use in different configurations and protocols. Elbert Cook, Jon Erickson, and Ed Lagan were the prime innovators. Jon Erickson also developed methods for distortion correction, signal processing and display of data from imaging devices as part of his PhD thesis.

Our first quantitative x-ray imaging studies used a NASA donated video recording system in the early 1970s. With help from Hamamatsu, we were able to capture, store and analyze data in the PDP9 from the bi-plane image intensifiers used in the Cardiac Cath Lab.


COMPUTER ASSISTED INSTRUCTION (CAI)
 In 1968, John Chapman, the Medical Center Dean for Educational programs announced a Computer Assisted Instruction (CAI) grant. We applied and with the award we purchased our first Digital Equipment Corp (DEC) PDP 9 computer. It had 4K 18-bit words of core memory, to which we later added another 4K words of memory. (Each 4K increment cost about $7K, many orders of magnitude higher than costs and performance of contemporary memory). We used methods and tools developed by the Stanford, SUMEX AIM (Artificial Intelligence in Medicine) project, and emulated work done at Mass General (MGH) by Octo Barnett. Students in the Vanderbilt  Engineering School developed tools we used that allowed teachers  to create learning modules. Question and answer scenarios probed different levels of knowledge, presenting remedial materials based on student responses. We collected sample programs developed elsewhere, and generated a few of our own, but none were received favorably by students.. We learned that it was easy to write CAI scenarios but that it was difficult to get talented teachers to spend the time and effort needed to craft effective lessons, in effect decreasing the need for their services in the classroom. CAI was shown elsewhere to be effective for language learning, and other rote memory  tasks. Students taught with the computer modules learned faster and retained foreign language skills longer than those exposed to standard class room lectures (Stanford Univ. report). New developments appear to be changing the situation as many Universities, including Vanderbilt are developing and distributing distant learning courses (Coursera) for different levels of primary through graduate and continuing education instruction. Certificates are awarded, and in some cases, students can earn course credit.


SCIENTIFIC COMPUTING AT VANDERBILT
Physicists had large data handling requirements as they analyzed large numbers of analog films that recorded accelerator-produced nuclear interactions produced at remote accelerators. The task was to identify rare particles based on analysis of kinematic features from bubble chamber tracks recorded on film. Vanderbilt graduate students manually digitized films containing data obtained from Brookhaven National Lab (BNL). The data were analyzed locally on the VU central computer facility which was barely up to the task. Alternative needs were presented by different users. Scientists in Engineering, led by George Haynam in EE/Computer Science and Bill Baker in Biomedical Engineering wanted the new computer to be one with a flexible front end that could be used to control real- time systems. The physicists wanted an IBM number cruncher. A battle between IBM, DEC (the Digital equipment Corp) PDP 10, and Xerox, Sigma 7 ensued. We cast our one vote for the Sigma 7 because of our shared interest in real-time device control. The Sigma 7 was advertised as capable to support such efforts, but this turned out to be wrong, and the Sigma 7 turned out to be a poor choice ,  a work-in-progress that required frequent operating system changes. The only control system tasks were done in Engineering using a small Sigma 5 that Xerox gave to Biomedical engineering. The Physics Department then acquired a PDP 10 dedicated to their needs.
This was at the time that large central computing facilities were being replaced  by increasingly powerful desktop computers. New commercial nuclear medicine imaging systems came with large memory, and special processors that provided the computational power needed for all but the most demanding tasks. For larger tasks shared Institutional resources became available containing clusters of  thousands of low cost Linux processors, as in the Vanderbilt ACCRE facility. For even more massive tasks, powerful super computers are available and can be accessed at ORNL and other National Labs.


MEDICAL COMPUTING DEVELOPMENTS
 Given the diverse growing needs for administrative, scientific and educational computing tasks, top level support was needed and Bill Stead was recruited around 1993 from Duke to lead the effort. In so doing, Vanderbilt escalated computing decisions to cover the College, the Medical Center, and the Hospital, and informatics professionals were recruited to coordinate Medical activities. Bill Stead led the Medical School developments, and in 2001 an Informatics Department was created, first Chaired by Randy Miller from Pittsburg. Miller was well known for diagnostic decision systems (Internist). He brought Kevin Johnson, Dario Rubia and Antoine Geissbuhler with him to Vanderbilt. Dan Masys in 2004,, succeeded Miller mad in 2012 Kevin Johnson became Chair, all having made significant contributions to Informatics as a discipline, while providing strong support for clinical and educational programs at Vanderbilt.. Antoine Geissbuhler, a Swiss physician/scientist, had earlier made significant contributions to nuclear medicine, working with David Townsend (then at CERN) where he wrote the first 3D volumetric code for PET image reconstruction. At Vanderbilt Antoine and Randy Miller led the development of the Wiz Order entry system, a technical and economic Vanderbilt success story. Antoine left Vanderbilt in the mid 1990s, returning to Geneva to lead the Swiss Informatics program from a base in the Cantonal Hospital.

Stead chaired the 2009 National Academy of Sciences report on the Electronic Health Record. Under his leadership Vanderbilt developed strong computer-based tools in support of research and education missions of the Medical School. Knowledge Map provides access to medical information from lectures and accessory information used in the academic and clinical courses. Andrew Spickard III leads many of the advances in educational support tools. A fast computer, IBM Nezza 1000 acquired in 2013 provides a flexible means of connecting genomic and clinical data bases in support of expanding personalized medicine initiatives. In 2013, John Lutz, an IBM systems expert from Canada, was hired to reorganize the cross campus-computing infrastructure. A senior physician/scientist, Blackford Middleton joined the Informatics team in Stead’s former role while Stead’s role was changed to Director of Strategic Planning. The Medical Center has received several important NIH grants to meet Affordable Care Act challenges, and the Medical Center is on target to meet many of the 2015. Hopefully, the success in the Stead days will continue as one attempts to meet the many remaining academic, technical and fiscal challenges.


→ Nuclear Medicine Computing Technology Sharing: We sought a means of collecting and disseminating computing codes from different nuclear medicine University and Industrial groups working to accelerate progress in the field. We learned that Alvin Weinberg, Director of Oak Ridge National Lab, created a series of Information Centers at ORNL, including the Radiation Shielding Information Center (RSIC). We contacted Betty Maskewitz at RSIC, and organized a meeting in Knoxville attended by nuclear medicine colleagues, including Henry Wagner, the Society of Nuclear Medicine (SNM) President. Henry had a keen interest in computing , as he and his Hopkins group were starting to use computers in nuclear medicine. At the meeting, he appointed Randy to Chair a new SNM Committee on Computer Applications. After the Knoxville meeting RSIC initiated a new Information Center, “Biomedical Computing Technology (BTIC) dedicated to the sharing of medical image processing codes. Over the subsequent years, they and Vanderbilt jointly sponsored annual meetings and many computer programs were collected and archived, and results of their use were presented at the annual meetings. In 1979 BCTIC lost its ORAU-based medical support, and BCTIC and its budget was transferred to Vanderbilt. Jon Erickson ran the program until 1983, when the program was discontinued as commercial computing systems had met many of the earlier goals.

NextArrowBackArrow Back to Path B: Device Developments
Next to Path B: Radiation Detectors