Technology and Health Care, 2 (1994) 53-60 0928-7329/94/$07.00 © 1994 Elsevier Science B.V. All rights reserved
53
Computer aided densitometric image analysis of bone remodelling after total hip arthroplasty F. Baruffaldi b
b,*,
A. Toni
a,
M. Viceconti
b,
A. Sudanese b, L. Brizio
b,
A. Giunti
a
a Clinica Ortopedica e Traumatologica, Universitli degli Studi di Bologna, Bologna, Italy Laboratorio di Tecnologia dei Materiali, Istituti Ortopedici Rizzol~ Via di Barbiano 1/10, 40136, Bologna, Italy
Received 22 December 1993, accepted 5 February 1994
Abstract By means of a simple and inexpensive system, the computer aided densitometric image analysis (CADIA), we can now evaluate on X-ray film the effect of long term bone remodelling after total hip arthroplasties (THA). The results of CAD lA system applied to the historical archive are burdened by a not always acceptable significance. The main source of error is the variance of radiological parameters. An attempt to retrieve the historical archive, that is of every X-ray film obtained out of any radiological control protocol, can be carried out with the bone density ratio representation.We have evaluated three different specimens with bone density ratio analysis: cementless group (96 cases), cemented group (59 cases) and non prosthetized femurs group (51 cases) as a control, all groups were followed up for more than 24 months.Ratios showed up different behaviour among the three specimens, confirming clinical evidence of a lower bone remodelling for cemented THA. Key words: Total hip arthroplasty; Bone remodelling; Densitometry; Image analysis
1. Introduction Changes in loading conditions can result in large alterations in either shape or mineralization of long bones. We know this adaptive process as bone remodelling or, in the case of total hip arthroplasty (THA) , as stress shielding, (Fig. 1).
* Correspondence to:
Laboratorio di Tecnologia dei Materiali, Istituti Ortopedici Rizzoli, Via di Barbiano 1/10, 40136, Bologna, Italy, tel. 51 6366864, fax 51 6366863, E-mail: [email protected] SSDI0928-7329(94)00006-S
Insertion of a metal stem in the medullary canal of the femur, as in hip prosthetic surgery, always initiates such a process. It is usually slight but sometimes it can be so severe that it becomes a major concern, leading to implant failure [1,2]. To monitor such processes the most common diagnostic technique available in a hospital is the X-ray. However its interpretation is subjective, not quantitative and scarcely repeatable. Other systems [3] such as QCT (quantitative computer tomography), SPECT (single photon emission computer tomography), DPA (dual photon absorptiometry), DEXA (dual energy X-ray
54
absorptiometry) are very expensive and not useful for densitometric research on the older X-ray films. Therefore it became necessary to develop a computer aided densitometric image analysis (CADIA) system as a support to the visual evaluation to monitor the bone density evolution around THA. CAD lA is today widely used in fields like contact microradiograph densitometry [4] or dental X-ray image analysis [5]. We have applied the CAD lA system to describe, morphologically and numerically, the alterations of long bones after THA.
F. Baruffaldi et al. / Technology and Health Care 2 (1994) 53 - 60
2. Experimental procedures
2.1. CADIA system The hardware components of the basic CAD lA system we used are: • video camera Ikegami mod. ICD-290 with 2/3" charge coupled device sensor (CCD); • PC Asem 80386 mod. Thor 9020, with maths coprocessor; • image acquisition AID board PFG, 512 X 512 8 bit-pixel with 256 grey level~ ':lvailable (8 bit systems can give us 2 8 = 256 grey levels);
Fig. 1. Bone remodelling after total hip arthroplasty: on the left 2 months, on the right after 30 months.
F. Baruffaldi et al./Technology and Health Care 2 (1994) 53-60
55
LATERAL ROI
MEDIAL ROI
r-'---'" TRoe (5) SUB TROC (6)
DYAPH (7)
APEX (4)
APEX (8)
o
255
o
RELATIVE BONE DENSITY
55
RELATIVE BONE
02 MONTHS •
44 MONTH~ENSITY
Fig, 2, CADIA evaluation of 8 periprosthetic RO!.
• optical bench Maxi-repro with a 3200 K backlighting system; • light controller Varilight 1000 rheostat; 0
• monitor Mitsubishi colour 19/1, The software governing CADIA system is written in C language and operates with Windows©
300
300..,--------~ 250+---I=~9r--
'in ~
200+----:-----t
~ 150 +---:::!------'I-~ 1OO+-~L-----+,,\O ~ 50~.,.,....----~t_
~ (])
o 150 ~
~
Q)
804--+-fk--T-"--
60+--++~--i¥-- Q) 40+--f+--~-tt-- !I 204---iH---~-=__
O"*" Medial-Lateral profile
Fig, 3, CADIA evaluation on profiles,
, '\
1,
Medial-Lateral profile
-
2 MONTHS
~ ~
V
o J*o,,""
Medial-Lateral profile
~ 1204---+-+---~ 100+--+-I_~t_-
I1
100
a: 50
O~------~
180...--------160+----ft----,~ 140+--........,:1Itp-----
--r-
250
'~ 200
-
-44 MONTHS
300..,..-_ _ _ _ _ __ ~ 'in
~
250+-----,,..-....,.----
200+-----t"---r---
o 1504--.......'""--+---~
'i
1OO~--I-f--+.:Io--
!I
50+--~--"'*"--
Qi
O~~~--~~--
Medial-Lateral profile
56
graphical user interface. This program has been designed with particular tools devoted to densitometric measurements: • templates of prosthesis to scale and to position all the follow-up films of the same patients; • average grey levels on regions of interest (ROn automatically positioned; • grey levels on profiles automatically positioned; • grey levels digital subtraction of different images; • calibrated linear measurements; • false colours representation; • image contrast and brightness elaboration; • images/patients data base management. To realize a comparative study we created a template of the prosthesis from the first post-operative image. Magnitude and positioning of successive images were verified on template to get the best spatial comparison. Two possible ways to semi-automatically evaluate bone remodelling were suggested. The first method concerned the evaluation of 8 periprosthetic ROI, automatically positioned referring to prosthesis templates and measuring the average grey levels evolution during time, (Fig. 2). The second one gave us the grey levels on some profiles, automatically positioned referring to prosthesis templates, with particular information about bone-prosthesis interface evolution, (Fig. 3).
3. CADIA errors analysis The densitometric evaluation of radiographs is affected by many error sources [6]. From the veritable radiological phase, and then through the digital acquisition phase and evaluation via CADIA, effects able to limit the precision of densitometric comparisons can add up. The two major issues, both of them mainly inherited by the radiological phase, involve the geometrical as well as the grey levels comparison of different pictures. Different enlargements and rotation of the same femur are an obstacle to the best geometrical comparison. As for the different enlargements,
F. Baruffaldi et al. / Technology and Health Care 2 (J994) 53 -60
these can be corrected in the acquisition phase by using the camera and with reference templates. In our case the prosthesis templates can be also used to discard the images with excessive rotations. We can therefore control the effects of different enlargements and rotations in our measurements. Concerning grey levels each X-ray film is characterized by different contrast and luminosity related to the different radiological and development parameters. A solution adopted by CADIA systems in the dental field [7] is to assume some radiological markers as densitometric reference, e.g. an Aluminum stepwedge, and to transform the comparison among grey levels in a comparison among equivalent aluminium thickness. This solution excludes all the images that were not obtained with this control procedure because in orthopaedics the X-ray film is almost' always obtained without densitometric references of this kind. As we did not want to lose the huge amount of information contained in every radiological archive, we analyzed the possibility of the assumption of one or more regions of the X-ray film itself as normalizing densitometric sites. For example, tests were carried out on the ischiatic zone or on the prosthesis itself. These two regions often fall out of the CAD lA dynamic window or of the X-ray film linearity zone, making densitometric normalization impossible. An alternative normalization procedure [8] is based on the grey scale every X-ray film has as a whole. This procedure of image processing is defined as 'equalization'. By splitting the image into 512 X 512 points and arranging the grey values on frequency histograms, it is possible to read the information about luminosity and contrast in the obtained distribution. From histogram mean value the average image luminosity can be drawn, and from the dispersion around the mean value, the contrast can be obtained. It is possible to make two different images comparable by altering the respective histogram distribution, as shown in Fig. 4, in order to get their best overlay. In the literature 'equalization' algorithms are divided into parametric [9] and non-parametric [10]. For the characteristics of our images we deemed the non-parametric procedure
F. Baruffaldi et af. / Technology and Health Care 2 (J994) 53 -60
57 Table 1 CADIA samples
, ;xXWg) 'i..iimW,.,."
,,"
53
Computer aided densitometric image analysis of bone remodelling after total hip arthroplasty F. Baruffaldi b
b,*,
A. Toni
a,
M. Viceconti
b,
A. Sudanese b, L. Brizio
b,
A. Giunti
a
a Clinica Ortopedica e Traumatologica, Universitli degli Studi di Bologna, Bologna, Italy Laboratorio di Tecnologia dei Materiali, Istituti Ortopedici Rizzol~ Via di Barbiano 1/10, 40136, Bologna, Italy
Received 22 December 1993, accepted 5 February 1994
Abstract By means of a simple and inexpensive system, the computer aided densitometric image analysis (CADIA), we can now evaluate on X-ray film the effect of long term bone remodelling after total hip arthroplasties (THA). The results of CAD lA system applied to the historical archive are burdened by a not always acceptable significance. The main source of error is the variance of radiological parameters. An attempt to retrieve the historical archive, that is of every X-ray film obtained out of any radiological control protocol, can be carried out with the bone density ratio representation.We have evaluated three different specimens with bone density ratio analysis: cementless group (96 cases), cemented group (59 cases) and non prosthetized femurs group (51 cases) as a control, all groups were followed up for more than 24 months.Ratios showed up different behaviour among the three specimens, confirming clinical evidence of a lower bone remodelling for cemented THA. Key words: Total hip arthroplasty; Bone remodelling; Densitometry; Image analysis
1. Introduction Changes in loading conditions can result in large alterations in either shape or mineralization of long bones. We know this adaptive process as bone remodelling or, in the case of total hip arthroplasty (THA) , as stress shielding, (Fig. 1).
* Correspondence to:
Laboratorio di Tecnologia dei Materiali, Istituti Ortopedici Rizzoli, Via di Barbiano 1/10, 40136, Bologna, Italy, tel. 51 6366864, fax 51 6366863, E-mail: [email protected] SSDI0928-7329(94)00006-S
Insertion of a metal stem in the medullary canal of the femur, as in hip prosthetic surgery, always initiates such a process. It is usually slight but sometimes it can be so severe that it becomes a major concern, leading to implant failure [1,2]. To monitor such processes the most common diagnostic technique available in a hospital is the X-ray. However its interpretation is subjective, not quantitative and scarcely repeatable. Other systems [3] such as QCT (quantitative computer tomography), SPECT (single photon emission computer tomography), DPA (dual photon absorptiometry), DEXA (dual energy X-ray
54
absorptiometry) are very expensive and not useful for densitometric research on the older X-ray films. Therefore it became necessary to develop a computer aided densitometric image analysis (CADIA) system as a support to the visual evaluation to monitor the bone density evolution around THA. CAD lA is today widely used in fields like contact microradiograph densitometry [4] or dental X-ray image analysis [5]. We have applied the CAD lA system to describe, morphologically and numerically, the alterations of long bones after THA.
F. Baruffaldi et al. / Technology and Health Care 2 (1994) 53 - 60
2. Experimental procedures
2.1. CADIA system The hardware components of the basic CAD lA system we used are: • video camera Ikegami mod. ICD-290 with 2/3" charge coupled device sensor (CCD); • PC Asem 80386 mod. Thor 9020, with maths coprocessor; • image acquisition AID board PFG, 512 X 512 8 bit-pixel with 256 grey level~ ':lvailable (8 bit systems can give us 2 8 = 256 grey levels);
Fig. 1. Bone remodelling after total hip arthroplasty: on the left 2 months, on the right after 30 months.
F. Baruffaldi et al./Technology and Health Care 2 (1994) 53-60
55
LATERAL ROI
MEDIAL ROI
r-'---'" TRoe (5) SUB TROC (6)
DYAPH (7)
APEX (4)
APEX (8)
o
255
o
RELATIVE BONE DENSITY
55
RELATIVE BONE
02 MONTHS •
44 MONTH~ENSITY
Fig, 2, CADIA evaluation of 8 periprosthetic RO!.
• optical bench Maxi-repro with a 3200 K backlighting system; • light controller Varilight 1000 rheostat; 0
• monitor Mitsubishi colour 19/1, The software governing CADIA system is written in C language and operates with Windows©
300
300..,--------~ 250+---I=~9r--
'in ~
200+----:-----t
~ 150 +---:::!------'I-~ 1OO+-~L-----+,,\O ~ 50~.,.,....----~t_
~ (])
o 150 ~
~
Q)
804--+-fk--T-"--
60+--++~--i¥-- Q) 40+--f+--~-tt-- !I 204---iH---~-=__
O"*" Medial-Lateral profile
Fig, 3, CADIA evaluation on profiles,
, '\
1,
Medial-Lateral profile
-
2 MONTHS
~ ~
V
o J*o,,""
Medial-Lateral profile
~ 1204---+-+---~ 100+--+-I_~t_-
I1
100
a: 50
O~------~
180...--------160+----ft----,~ 140+--........,:1Itp-----
--r-
250
'~ 200
-
-44 MONTHS
300..,..-_ _ _ _ _ __ ~ 'in
~
250+-----,,..-....,.----
200+-----t"---r---
o 1504--.......'""--+---~
'i
1OO~--I-f--+.:Io--
!I
50+--~--"'*"--
Qi
O~~~--~~--
Medial-Lateral profile
56
graphical user interface. This program has been designed with particular tools devoted to densitometric measurements: • templates of prosthesis to scale and to position all the follow-up films of the same patients; • average grey levels on regions of interest (ROn automatically positioned; • grey levels on profiles automatically positioned; • grey levels digital subtraction of different images; • calibrated linear measurements; • false colours representation; • image contrast and brightness elaboration; • images/patients data base management. To realize a comparative study we created a template of the prosthesis from the first post-operative image. Magnitude and positioning of successive images were verified on template to get the best spatial comparison. Two possible ways to semi-automatically evaluate bone remodelling were suggested. The first method concerned the evaluation of 8 periprosthetic ROI, automatically positioned referring to prosthesis templates and measuring the average grey levels evolution during time, (Fig. 2). The second one gave us the grey levels on some profiles, automatically positioned referring to prosthesis templates, with particular information about bone-prosthesis interface evolution, (Fig. 3).
3. CADIA errors analysis The densitometric evaluation of radiographs is affected by many error sources [6]. From the veritable radiological phase, and then through the digital acquisition phase and evaluation via CADIA, effects able to limit the precision of densitometric comparisons can add up. The two major issues, both of them mainly inherited by the radiological phase, involve the geometrical as well as the grey levels comparison of different pictures. Different enlargements and rotation of the same femur are an obstacle to the best geometrical comparison. As for the different enlargements,
F. Baruffaldi et al. / Technology and Health Care 2 (J994) 53 -60
these can be corrected in the acquisition phase by using the camera and with reference templates. In our case the prosthesis templates can be also used to discard the images with excessive rotations. We can therefore control the effects of different enlargements and rotations in our measurements. Concerning grey levels each X-ray film is characterized by different contrast and luminosity related to the different radiological and development parameters. A solution adopted by CADIA systems in the dental field [7] is to assume some radiological markers as densitometric reference, e.g. an Aluminum stepwedge, and to transform the comparison among grey levels in a comparison among equivalent aluminium thickness. This solution excludes all the images that were not obtained with this control procedure because in orthopaedics the X-ray film is almost' always obtained without densitometric references of this kind. As we did not want to lose the huge amount of information contained in every radiological archive, we analyzed the possibility of the assumption of one or more regions of the X-ray film itself as normalizing densitometric sites. For example, tests were carried out on the ischiatic zone or on the prosthesis itself. These two regions often fall out of the CAD lA dynamic window or of the X-ray film linearity zone, making densitometric normalization impossible. An alternative normalization procedure [8] is based on the grey scale every X-ray film has as a whole. This procedure of image processing is defined as 'equalization'. By splitting the image into 512 X 512 points and arranging the grey values on frequency histograms, it is possible to read the information about luminosity and contrast in the obtained distribution. From histogram mean value the average image luminosity can be drawn, and from the dispersion around the mean value, the contrast can be obtained. It is possible to make two different images comparable by altering the respective histogram distribution, as shown in Fig. 4, in order to get their best overlay. In the literature 'equalization' algorithms are divided into parametric [9] and non-parametric [10]. For the characteristics of our images we deemed the non-parametric procedure
F. Baruffaldi et af. / Technology and Health Care 2 (J994) 53 -60
57 Table 1 CADIA samples
, ;xXWg) 'i..iimW,.,."
,,"
