Journal of Endodontics Research -
Indirect digital images: limit of image compression for diagnosis in endodontics
By JofER editor
Published on 12/3/2008
M. Siragusa & D. J. McDonnell
Department of Endodontics, School of Dentistry, Universidad Nacional de Rosario, Santa Fe, Rosario, Argentina.

To analyze the impact of the loss of information that results from the compression of a file containing a radiographic image.

The compression ratio calculated as the quotient between the file sizes and the standard deviation of the values corresponding to the image that resulted from digital subtraction may be employed to assess the outcome of the compression process and guarantee adequate quality.

Introduction - Materials and methods.
M. Siragusa & D. J. McDonnell
Department of Endodontics, School of Dentistry, Universidad Nacional de Rosario, Santa Fe, Rosario, Argentina.

Several studies have compared the performance of direct digital intraoral radiographs (DDR) and traditional radiographs (Barbat & Messer 1998, Burger et al. 1999). Other studies have reported the results of the digital processing of images and their value as a clinical tool. These studies demonstrate the adequacy and feasibility of this technique and its advantages over conventional radiography (Mistak et al. 1998, Sullivan et al. 2000, Yousef & Saas 2000). The feasibility of digitizing conventional radiographs employing scanners for visualization and/or data transmission has been evaluated, and the minimum technical requirements for adequate digital processing have been established (Pluss et al. 1998).
Digital radiographs are byte maps where each byte represents the localization and intensity value of each element in the image (Gonzalez & Woods 1996). The quality of the image is directly proportional to the amount of descriptive elements. Most images have areas of slight variations that allow for the use of algorithms to reduce the size of the file. The aim of file compression is to minimize transmission time and memory requirements. The maximum degree of compression that is not detrimental to the quality of diagnosis in endodontics has not yet been established (Wenzel et al. 1996).
The aim of the present study was to analyse the effect of the information loss that results from the compression of a file containing a radiographic image.

Materials and methods.
Fourteen randomly selected intraoral radiographs with different degrees of diagnostic difficulty, including preoperative working length determination or postobturation images, were selected for study. They were digitized with an AGFAARCUS II scanner using the maximum optical resolution without interpolation, i.e. 300 ppi and a grey scale of 8 bytes. The evaluation scale was kept constant and independent of the natural histogram corresponding to each image. A digital image with no saturation or trimmings was obtained. The resulting image (176410-177114 bytes) was stored in Tagged Image File Format (TIFF).
Each of the resulting image files was compressed without loss of information employing WinZip 8.0. Compression with loss of information was performed with Photoshop 5.0 software (Adobe Systems Inc., San Jose, CA, USA) and the Joint Photograph Expert Group (JPEG) format. Each of the images was compressed from its original TIFF format to each of the available qualities (10- 0). The resulting 182 images were grouped as follows: 14 original images, 14 images compressed without loss of information and 154 images compressed with loss of information, in11groups of14 images each.
A qualitative evaluation by a single endodontic expert, to avoid the introduction of interobserver variables and to guarantee uniform assessment criteria, was performed. The aim was to compare the observations that resulted from each of the original TIFF images with those derived from each of the corresponding JPEG images. The observer had no knowledge of the degree of compression applied in each case and was allowed to adjust brightness and contrast freely for the original image. The same values were then applied to all the JPEG images. Observation was performed employing Photo-Paint 8 software (Corel Corporation, Ottawa, Ontario, Canada), the tool in which the observer was best trained. The observed images were scored as follows:
  • Adequate for diagnosis: when the observer could see the image details clearly on the monitor.
  • Adequate for illustration only: when the image details where blurred on the monitor.
  • Inadequate: when the image sharp details where lost.
For the quantitative analysis, each of the JPEG images was digitally subtracted from the corresponding TIFF original image, thus generating a new image in TIFF format as a result of the following operation performed using Adobe Photoshop 5.0. D(n,j).tif ј n.tif_n(quality j).jpg ю125 where n ј1. . .14, is the number of the image, j ј 0. . .10, is the quality applied by Photoshop 5.0 in JPEG format, 125 is a fixed half grey level in the 8 bits scale (125 ј256/2), D(n, j).tif is the digitally subtracted image in TIFF format, n.tif is the original image number n, in TIFF format and n(quality j).jpg is the image number n compressed in JPEG format, with quality j.
The histograms of the grey values corresponding to the 154 new images were statistically processed. The grey values of the pixels followed a normal distribution. The mean and standard deviation were calculated for each case.

Results - Discussion - References.
Compression without loss of information yields an image that is identical to the original, and thus requires no comparison. The average size of the files was 112823 bytes (range 102943-119523) and the average compression was 1.57 (range 1.48-1.71), thus yielding a file that was 68% the size of the original file. The smallest compression was achieved with images 7 and 10; the greatest compression ratio was achieved with image 4 (Table 1).
For JPEG compression (with loss of information), the file sizes are presented in Table 1 and the compression ratios are presented in Table 2. The compression ratio for all the qualities and radiographic images reached a minimum mean value of 2.26 and a maximum of 19.2 with a range of 2.14 for image12 to 22.9 for image 4.
The mean and standard deviation values that resulted from the quantitative statistical analyses are presented in Tables 3 and 4.The results of the qualitative comparative assessment are presented in Table 5. A summary of the present conclusions is presented in Table 6.

Table 1. Size of the image files.

Size of the image files

Table 2. Compression ratios with and without loss of information.

Compression ratios with and without loss of information

Table 3. Mean grey values corresponding to the images that resulted from digital subtraction.

Mean grey values corresponding to the images that resulted from digital subtraction

Table 4. SD values corresponding to the images that resulted from digital subtraction.

SD values corresponding to the images that resulted from digital subtraction

Table 5. Results of the subjective evaluation by the expert.

Results of the subjective evaluation by the expert

Table 6. Summary of results.

Summary of results

Handling large amounts of digital data requires fast, simple and efficient access, a stable communication system and a safe storing procedure within a Picture Archiving and Communications System (PACS). Accepted international standards such as TIFF for images, ASCII for text and DICOM in general must be used. File compression favours speedy access and data transmission and reduces the memory requirements for storage. JPEG compression involves irreversible loss of information, and could be severely detrimental to diagnostic interpretation.
The present study involved the comparative, qualitative and quantitative assessment of indirect digital images obtained from 14 conventional radiographs. They were submitted to a compression process with or without information loss, employing the WinZip 8.0 and Photoshop 5.0 software. The results presented in Table 1 reveal that compression with no information loss reduced the size of the file to 68% of the original, and is a valid alternative to storing original images. The range of compression achieved was below that for compression with information loss.
JPEG compression is a widely accepted standard and runs on multiple platforms. The different edition programs have different interfaces for interaction with the operator. Thus, each of these programs uses a different compression scale. Photoshop uses a compression quality scale of 0-10, whereas PhotoPaint uses a 0-100 scale. Within this context, the final result may be difficult to evaluate.
The analysis of Tables 1, 2 and 3 demonstrates an unexpected outcome, i.e. the JPEG compression qualities 4 and 5 seem to be inverted. The file size for quality 4 was larger than for quality 5. This implies a lower compression ratio for quality 4 that was confirmed, in turn, by a larger standard deviation for quality 5. This finding was confirmed by repeating the procedure to rule out possible errors, and might reveal a flaw in the identification of the JPEG compression algorithm by the Photoshop 5.0 program.
The images compressed to qualities10, 9, 8 and 7 and some of the images compressed to quality 6 did not show interpretation difficulties. Belowt his quality and up to quality level 2, visualization was overall acceptable, but the resulting images were unsatisfactory in terms of guaranteeing accurate diagnosis owing to loss of information on the edges and loss of fine details. Lower quality images were inadequate. The quantitative analysis performed from the viewpoint of Medical Physics employing digital subtraction allowed us to draw the following conclusions:
  • The image that results from the subtraction between the original image and the compressed image is the one that causes visualization difficulties. This becomes more marked as the degree of compression increases.
  • The standard deviation of the grey values of the image that results from the subtraction process indicates the quality of the image and its adequacy for use in diagnosis.
The comparative assessment of both techniques revealed that a JPEGlossy compression six times smaller than the original TIFF is compatible with diagnostic applications. A compression between six and 13 times was not considered suitable for diagnosis. However, the overall quality of the resulting images would be adequate for illustration purposes or when fine details are not diagnostically relevant. The standard deviation of the values corresponding to the image that results from digital subtraction was appropriate to assess the result of the compression process and guarantee adequate quality.


Barbat J, Messer H (1998) Detectability of artificial periapical lesions using direct digital and conventional radiography. Journal of Endodontics 24, 837-42.
Burger C, Mork T, Huntter J, Nicoll B (1999) Direct digital radiography versus conventional radiography for estimation of canal length in curved canals. Journal of Endodontics 25, 260-3.
Gonza. lez R,Woods R (1996) Representacion digital de imagenes. In: Gonza. lez R,Woods R, eds. Tratamiento digital de imagenes. Wesley IberoAmericana S.A.: Addison, 6-7,32-4.
Mistak J, Lousine R, Primack P (1998) Interpretation of periapical lesions comparing conventional, direct digital, and telephonically transmitted radiographic images. Journal of Endodontics 24, 262-6.
Pluss J, SiragusaM, McDonnell J (1998) Hacia la Teleodontologa. Electronic Journal of Dental Research.
Sullivan J, Di Fiore P, KoerberA (2000) RadioVisioGraphy in the detection of periapical lesions. Journal of Endodontics 26, 32-5.
Wenzel A, Gotfredsen E, Borg E, Grondahl H-G (1996) Impact of lossy image compression (JPEG) on accuracy of caries detection in digital images taken with a storage phosphor system. Oral Surgery Oral Medicine Oral Pathology Oral Radiology Endodontics 81, 351-5.
Yousef SaadA, SaasA-N (2000) Radiation dose reduction during endodontic therapy: a new technique combining an apex locator (Root ZX) and a digital imaging system RadioVisio- Graphy). Journal of Endodontics 26, 144-7.