Malaysian J Path01 1990; 1 Z(2): 101 - 106

A dBASE 111 SYSTEM FOR MANAGING 35 MM SLIDES IN PATHOLOGY

KT WONG, MBBS. MPath and K S CHAN, MBBS.

Department of Pathology, Faculty of Medicine, University of Malaya. Summary We describe the design and Management of a 35 mm slide using a menu-driven dBASE HIT* PLUS programme and a in a large department of pathology that also caters for the individual pathologist. Existing systems described in the literature are geared towards slides of general medicine and do not address the needs of the individual pathologist. A total of 11,481 slides in fhe Department of Pathology, Faculty of Medicine, University of Malaya, were filed into a single database with each record representing one slide. Nine fields which comprised the slide accession number, reference number, slide category, SNOMED codes, and a description of the slide in natural language, seemed adequate for slide definition. The menu-driven programme had functions which included the abilities to add, delete, edit and back-up records, and to search for desired slides. Although slides may be searched for in various fields, we found that searches using natural language alone were both comprehensive and efficient, provided a standard format of description was adhered to and data entries scrutinized carefully for errors. We believe therefore, that for the pathologist working alone, coded language fields are not absolutely necessary, as manual coding and additional data entry can be time consuming. As expected, for larger than 10,000 slides, 3 80286 microprocessor-based microcomputer was more efficient. We are of the opinion that a system such as ours is very useful for a large department of pathology or the individual pathologist to file and retrieve 35 mm slides.

Keywords: 35 mm slide, microcomputer, dBASE IlITM PLUS, database, pathology, SNOMED, natural language.

INTRODUCTION The 35 mm slide remains an important visual RAM, a fixed (hard) disc and a floppy disc aid in the dissemination of medical drive, a monochrome monitor and a dot-matrix knowledge.' y2 This certainly applies to the printer. field of pathology where the use of such slides is often crucial in the lecture hall, research MATERIALS AND METHODS laboratory, seminars and clinico-pathological Database conferences. As the collection of such slides grows, A single database (SLIDE.dbf) was used there is a real need for an efficient system to to file 11,481 slides of the Department of file and re.trieve them. Several authors have Pathology, Faculty of Medicine, University written about filing and retrieval systems for of Malaya. Each slide was filed as a single 35 mm slides of general medicine using micro- record. New slides acquired were filed as soon c~rn~uters.~~~~~We are, however, not aware as possible while the filing of backlog slides of any published report on the use of was done as an on-going process, generally microcomputer systems to manage 35 mm but not necessarily in reverse order of their slides in a large department of pathology, or slide numbers. which addresses the specific needs of the Two academic staff wrote down the entries individual pathologist in such a task. for the various fields of each record, after We, therefore, describe the design of a reviewing each slide individually. A standard database for 35 mm slides of a largely format of data entry and recording was adhered pathological content, and the problems and to at all times. The lists were then handed to solutions inherent in the management of such a typist to enter into the computer. All slides, with particular reference to the alphabetical data were entered automatically individual pathologist. We us$ easily available in uppercase (capital) letters using a custom computer software and hardware. The minimum screen. Entries were only considered final hardware requirements are an 8088 IBM after at least 2 checks for mistakes by the compatible microcomputer with 640 KB same academic staff.

Address for reprlnt requests: Or K T Wong. Department of Pathology. Faculty of Mediclne, University of Malaya, 59100 Kuala Lumpur. Malaysia. Malaysian J Path01 December 1990

Software and hardware case number of the specimenlpatient of which the slide was taken. This could be the autopsy, The commercially available software surgical, cytology or other reference numbers. programmes used were PC DOS version 3.3 Apart from making it possible to obtain more (International Business Machines Corp., Boca information abut the case than can be Raton, FI,, USA), dBASE I11 PLUS version 1.1 provided by the DESCRIBE field, this reference (Ashton Tate, Torrance, CA, USA) and number provided a !ink between slides WordPerfect version 5.0 (WordPerfect Corp., belonging to the same specimenlpatient that Orem UT, USA). The hardware consisted of were taken and filed at different times. a 4.77 MHz, 8088 microprocessor-based IBM- Five axes of the Systematised Nomenclature compatible microcomputer with a 20 MB of Medicine (SNOMED) codes,6 provided fixed disc, a floppy disc drive, 640 RAM, additional slide definition (Table 1). The codes standard keyboard, CGA colour monitor, based on procedure and occupation were not and a dot matrix printer. We also used a 16 included. Since most of our slides had only MHz, 80826 microprocessor-based, IBM- one diagnosis, only one field for each of the compatible microcomputer with a 20MB five axes was needed. For all newly acquired fixed disc and a floppy disc drive to test our slides, we entered the SNOMED codes after system. manually looking up the appropriate codes. In a majority of cases, it was necessary to Programme design enter only the topography and morphology All files, including SLIDE.dbf, were kept codes. As most of our backlog slide collection in the fixed disc. A menu-driven programme have yet to be coded, we chosed not to code (written by K S Chan) provided easy access them when entering data initially. This was to the database. Its main programme necessary as coding was very labour intensive. (SL1DE.prg) and other programme files were Only after entry of all backlog slide data written in dBASE language using WordPerfect (other than codes) into the computer was instead of the in-built dBASE 111 word completed, did we start to enter the codes. processor as the former has a superior text The field CATEGORY (Table I) recorded handling capability. the type of slide and this was designated by Three sub-directories were created: 2 sub- a single letter or a combination of 2 letters directories to keep PC DOS and dBASE I11 e.g. "G" = Gross specimen or patient; "H" PLUS respectively, and the third contained = Histology or microscopic appearance; "C" all other programme files including SLIDE. = Cytology, and "EH" = Electron microscopy dbf, SLIDE.prg, etc. and Histology, and so on. The menudriven programme consisted of a RESULTS main menu and 4 submenus (Fig. l). From the main menu, submenus were selected by pressing The structure of SLIDE.dbf is show11 in a single key which would then replace the main Table 1. Nine fields were used to define each menu on-screen. Likewise, choices made on record. Theoretically, 99,999 slides can be the submenus would reveal other screens or accommodated in the database. The SLIDENO further submenus. The important functions field contained the unique accession number performed by our programme included the given to each slide at the time of acquisition. abilities to add, delete (or cancel), edit (or SL1DE.dbf was indexed by this accession change), back-up (or duplicate into diskettes) number. records, and to search for required slides. In the DESCRIBE field (Table l), natural Several pages of documentation provided help language keywords were used to record suitable to the novice user. descriptions for the slide. This usually meant The search function allowed for slide listing the organ or tissue, diagnosis, searches using slide numbers, case numbers, pathological features, and special stains, in this SNOMED codes, or keywords. In practice, order. British spelling was used. As far as since most of our slides were not coded, most possible, we used nouns, e.g., a slide showing slide searches were done using keywords. lung involved with a metastatic lesion This, we have found to be entirely satisfactory. from cervical squamous carcinoma would be A maximum of 3 keywords may be searched entered as "LUNG, METASTASIS, for at one go. In addition, an option to exclude SQUAMOUS CARCINOMA, CERVIX". Short one word from the search was provided. In forms were generally u'sed only for special instances where the length of the word steins e.g. "PAS" for "periodic acid-schiff". exceeded the 20 spaces provided, part of the The CASENO field (Table l) contained the word usually sufficed. 35 MM SLIDE dBASE 111 SYSTEM

TABLE 1 STRUCTURE OF DATABASE (SLIDE.dbf)

Field Field Name TYpe Width * Content

1 SLIDENO Numeric 5 Unique accession number

DESCRIBE Character 60 Organ diagnosis, special features, etc.

3 CASENO Character 10 Reference number

4 CODET Character 5 SNOM ED code: topography

Character 6 SNOMED code: morphology

6 CODED Character 5 SNOMED code: disease

7 CODEE Character 4 SNOMED code : aetiology

8 CODEF Character 5 SNOMED code: function

9 CATEGORY Character 2 Type of slide

* Total width = 102

For slide searches using two or more search could be terminated by pressing the keywords, the combination of words must all escape key without having to search through be matched before the record was selected, the entire database. i.e. a conjunctional (logical and) search. We In terms of speed, excluding one word was did, not provide for an alternation (logical equivalent to searching for an additional word, inclusive or) search because we felt this would while the length of the word did not alter the complicate the search procedure. Moreover, speed appreciably. There was no difference this type of search can easily be done by in speed even if words were entered in different repezted searches using single keywords. orders, e.g. in searching for "OVARY", Using the 8088 microcomputer, and "CYSTADENOMA" and "BORDERLINE", allowing the entire database to be searched, entries made in this order were the same as about 9, 10, and 13 minutes were needed to the reverse order. We tested the search function search 1, 2, or 3 words respectively (the time with the 80286 microcomputer and the speeds interval being measured from the completion were only about 2 minutes in all instances. of keyword/s entry to the completion of The search output list took 2 optional search). The time needed to search for 3 forms (the BRIEF or the COMPLETE report words and exclude I word was about 20 formats). In most instances, we used the minutes. In practice, however, less time was BRIEF format (fields SLIDENO, CASENO often needed, as our database contained many CATEGORY, DESCRIBE only) as the, sldes of the same subject which were of a information in these fields was usually adequate similar quality so that once the user felt he had for slide selection (Fig. 2). There was the an adequate number of desired slides, the added advantage that the on-screen display MalrryJirm J Parhol December 1990

Quit Slide System mI 1 Search by Keywrds I 2 Search by Slide No. r) 3 SearchbyCueNo. 4 Search by SMHED Codes 1 X ~xitmis h

Database Maintenance Menu 3hSlide Systen Main Menu I Quit Slide System O Cmit Slide System 1 Search for Slides 1 2 Browse Records 2 Maintenance of Database l 3 Edit Records 3 Tools and Utilities 4 Re-create Idex Files 4 Help Back up and Restore Database I 5 I Enter your option ->

Tools and Utilities Menu

0 Quit Slide System 1 View or Print File 2 List Duplicated Slide hnber 3 Char\geReportFonmt&I.Me 4 Print Records in Retural Order Fig. 1 : On-screen main menu and submenus. X Exit'IhisMenu

of search output lists was simple and In the context of anatomical pathology unclutte d. For the purpose of checking data databases, it has been suggested that the use entries. the COMPLETE format (all fields of both coded language (e.g. SNOMED codes) displayed) was used. and natural language is probably ideal for recording diagnoses.899Consequently, we used With the search list (Fig. 2) as a guide, both coded and natural language for slide suitable slides were retrieved from a storage definition in our database. unit consisting of slide boxes, where slides If the SNOMEDcoding system is used, were arranged according to the accession S axes (topography, morphology, aetiology, numbers. In addition to a print-out, our function and disease) are adequate for most programme offered an option to save the search diagnostic details.' Although our experience list into an ASCII file on a floppy disc for with the SNOMED code for slides is limited, future reference. as most of our slides have yet to be coded, we believe that the multiaxial SNOMED code DISCUSSION is at least as good as, if not better than, the One of the most important function any other coding system which has been used by slide filing system should have is the ability others for slide definition, viz., the WHO to retrieve desired slides expeditiously and International Classification of effectively because valuable teaching slides are In recent years, controversy has arisen as good as lost if they cannot be regarding the relative advantages of using coded To achieve this, accurate slide description or language and natural language in computerized definition is a prerequisite. medical records. Some authors1 ' were of the 35 MM SLIDE dBASE III SYSTEM

Word( S) Searched : PYELONEPHRITIS

Page No. 1

Slide# Case# Category and Description of Slides

KInNEY, DIABEITS MEU,ITUS, CHRONIC PYELONEPHRITIS KIDNEY, C) RONIC PYELOIIJEPHRITIS KIDNEY, ACUTE PYEZONEPHRITIS KIDNEY, ACUTE PYELONEPHRITIS KIDNEY, ACU1'E P?XLONEPFIRITIS KIDNEY, ACUTE PYnONEPHRITIS KIDNEY, ACUTE PYnONEPHRITIS KIDNEY, I'YELONEPHRITIS WITH ABSCESS KIDNEY, P'ELONEPHRITIS, ATROPHY PERICAKI)IUM, SEPTIC PERICARDITIS, PYELONEPHRITIS KIDNEY, ACUT'F: PYEXONEPFBITIS, HYDRONEPKROSIS KIDNEY, ACUTE PYf3ONEPf-RITIS KIDNEY, ACUTE PYESLONEPINITIS KIDNEY, CHRONIC PYF:L.ONEPHRITIS KIDNEY, CHRONIC PYELONEPt-IKI'TIS KIDNEY, PYELONEPHRITIS, GRAM -VE SEmICAEMIA KIDNLT, CWONIC PYWNEPHRITIS, HYDRONEPHROSIS KIDNEY, CHRONIC PYELQNEPHRITIS, HYDRONEPIBOSIS KIDNEY, CHRONIC PYELONEPHRITIS, HYDRONEPHROSIS KIDNEY, CHRONIC PYELONEPHRITIS, HYDRONEPHROSIS Fig. 2: An example of a search output list. opinion that the latter will assunie a greater database with a large pathological content wc role, while Friedman8 suggested that natural recommend its inclusion in the slide database language searches could sonietimes be more design. Thus, we believe the minimal number efficient than SNOMED searches. of fields in any pathological slide database In situations like ours, where coding lagged should include - in the context of our database behind natural language slide description, we - the fields SLIDENO, CASENO, CATEGORY found the use of natural language alone in and DESCRIBE. New fields may be added slide searches to be entirely satisfactory, whenever the need arises. provided a standard format of description One problem that has to be addressed in was strictly adhered to. Moreover, the using natural language description is that DESCRIBE field. wheri: natural language slide preferred clloices of keywords must be definition was entered, was flexible enough communicated to other users in a large to accommodate such diverse slide descriptions department. A list of commonly used keywords as "vascular invasion" or "hyperchromatic made available to all users would partially nuclei" that would otherwise be inipossible overcome this problem. For the single-user with established coded language. Likewise, database, this problem would not arise if multiple diagnoses and cross-indexing were preferred keywords are used consistently all easily accommodated. the time. We believe, therefore, that for the individual As one's database grows, the time needed pathologist operating alone, or who has scarce to search for slides naturally increases. For clerical resources at his disposal, it might be a database of about 10,000 records, using a advantageous not to include a coded language 4.77 MHz, 8088 niicrocomputcr is probably field or fields in the database. This is because too slow to be practical. An 80286 micro- manual coding and additional data entry computer, on the other hand, would be more involved can be very time consuming. suitable since slide searches by natural language As the CATEGORY field seems to be very keywords can be achieved in only about 2 useful for further slide classification in a slide minutes. Thus, in choosing computer hardware, Malaysian J Pathol December l990 the projected size of the database should be REFERENCES borne in mind. Typographical errors during data entry 1. Barker VF, Harden RM. The storage and would undoubtedly hamper slide search using retrieval of 35 mm slides. Med Educ any field and Inay even render the slide 1980; 14 (l): 59-7 l. inaccessible. This is particularly true for codes.9 2. Hedley AJ, Morton R. The clinical slide Thus, careful checking of data entries cannot library: A valuable learning resource in be overemphasized. continuing medical education. Med Biol Although the use of a menu-driven Illust 1976; 26: 203-7. programme generally simplifies access into the 3. Gilson CC, Collins JM. Use of the database, it may inipose certain limjtations microco~nputerin a department of medical to the experienced user well versed with dBASE illustration for retrieval of clinical teaching language. In searching for slides e.g., one Inay slides. J Audiovisual Media Med 1982; be able to execute a more specific slide search 5: 130-4. "manually" than may be possible with a rnenu- 4. Brown SE. Ashton Tate's dBASE 11 as a driven programme. Nevertlieless, for a patient-index manager. J Audiovisual collection of slides which caters for many Media Med 1985; 8: 5-8. users, some of whom Inay be computer 5. Robertson SJ. 'Cardbox' - an ideal illiterate, a menu-driven program~ne offers program for clinical filing. J Audiovisual a practical solution. Media Med 1985; 8: 19-2 l. Automated coding if incorporated into 6. Systematized Nomenclature of Medicine the system, will enable coding to be done (SNOMED), ed 2. Skokie, 111, College of speedily and remove the tedium of manual American Pathologists, 1979. coding. Significant advances in optical disc 7. Nayler J. The CPHA's ICD-9.CM as a technology such as WORM (Write Once. Read coding source for niedical slide filing. Mainly) has enabled mass storage of J Audiovisual Media Med 1990; 13: information and if available, can be used to 55-60. store slide iniages12 so that viewing of slides 8. Friedrnan BA. The impact of new features for selection becomes much easier as the of laboratory information systems on image can be brought on-screen. quality assurance in anatomic pathology. We wish to stress that our system is a low- Arch Pathol Lab Med 1988; 112: cost, stand alone system, operable on a 1189-91. relatively unsophisticated microcomputer. We 9. Ulirsch RC. Status of anatornic pathology find our con~puterisedsystem efficient in the data management systems. Arch Pathol management of 35 mm slides in a large Lab Med 1984; 108: 884-7. department of pathology. It offers the speed 10. Cote RA, Robboy S. Progress in medical and co~nprehensivenessneeded to cut down information management. Systematised the time -pent searching for desired slides, Nonienclature of Medicine (SNOMED). thereby reeillg the pathologist to concentrate JAMA 1980; 243 (8): 756-62. on more productive work. The individual 11. Moore GW, Hutchins GM, Miller RE. pathologist. too, should have no problem using Strategies for searching niedical natural it to build up a 35 mm slide database, manage language text. Am J Pathol 1984; 115: and retrieve slides needed for teaching and 36-4 1. demonstrations. 12. Aller RA. Surgical pathology cornputer ACKNOWLEDGEMENT systems have colne of age. Arch Patliol We are thankful to Dr Jayaranee for her Lab Med 1988; 1 12: 1 186-7. assistance in listing and checking data entries and to MS Sariniall for entering data into the computer. We also would like to thank Prof. L M Looi, Head, Department of Pathology, University of Malaya for pern~issionto publish this article.