Employing Raman Chemical Imaging - ChemImage
Automation of Ingredient-Specific Particle Sizing Employing Raman Chemical Imaging
TM
Ryan J. Priore, Oksana Olkhovyk, Oksana Klueva and Michael Fuhrman ChemImage Corporation, Pittsburgh, PA , USA
Raman Chemical Imaging (RCI) has been studied as a potential measurement tool for determining the particle size distribution of a corticosteroid in the formulation matrix (3). RCI combines the chemical identity and selectivity of Raman spectroscopy with digital imaging to determine the particle size distribution and association/ agglomeration of micronized drug in aqueous suspension formulations of nasal sprays. Improved automation across both hardware and software in wide-field chemical imaging is enabling RCI to measure larger populations of ingredient-specific particles as well as incorporate repetitive sampling methods. We have applied RCI to characterize the budesonide API particle size in a sample of Rhinocort® AquaTM nasal spray employing an automated method of data collection and analysis of Raman chemical and brightfield imaging. The advantages of automated software processing of particles on a particle by particle basis are compared to the current global method for determining the particle size distribution of a corticosteroid.
Experimental Polystyrene microsphere suspensions (Polysciences, Inc., Warrington, PA), 0.99 ± 0.03 µm and 5.34 ± 0.26 µm in diameter (mean ± standard deviation), were used to develop the concept of local processing of individual fields of view and ultimately individual particles. Brightfield reflectance and Raman chemical images over the 1000 cm-1 band were collected over individual fields of view until a minimum of 100 microspheres of each diameter were obtained, then these fields of view were concatenated together to form a dataset. A Rhinocort® AquaTM nasal spray (32 mcg budesonide, AstraZeneca, Wilmington, DE) sample was prepared by shaking, priming (four actuations each) and spraying in an upright position onto an inverted aluminum-coated glass microscope slide positioned approximately 15 cm above the spray nozzle. The microscope slide was then immediately turned upright and the nasal suspension droplets were allowed to dry. Actuated samples were analyzed to include actuation device influence as opposed to bulk samples. 1. Sample preparation
2. Raman spectroscopy of pure components
3. Optimize spectral range for Raman Chemical Imaging
A
Global processing uses one threshold for all particles based on pixel intensity distribution. Secondary scattering and the reliance upon spectral normalization to flat-field the chemical image typically leads to RCI oversizing medium and large particles when using global processing methodologies, while small particles are sometimes lost. Local processing treats the identified particles individually resulting in a more accurate representation of the ingredient-specific particle size distribution while also retaining smaller particles through the following data processing routine: 1. Individual API particles are initially identified by their specific Raman spectrum 2. Each particle is treated with a unique intensity threshold unlike in global 3. processing where a single threshold value is utilized to binarize the Raman 4. chemical image and thus yielding a particle map 3 A feedback loop is initiated which confirms the chemical identity against the 5. 6. Raman spectrum and validates the particle size against the brightfield optical 7. image A representative set of polystyrene particle data is illustrated in Figure 3 to demonstrate the use of RCI information to size the microspheres. A
C
B
EXCIPIENT #4 EXCIPIENT #5
5. Data processing
2. Bias Correction 3. Vector Normalization 4. API Peak Frame extraction
2600
2800
7. Boundary Particles Removed
Total Particles
50 40
Standard Deviation
3.1 µm
D10
1.1 µm
D50
3.5 µm
D90
8.5 µm
30 20 10 0
2
313
4
6
8 10 12 14 Max Chord (µm)
16 18
Figure 1. Ingredient-Specific Particle Sizing analytical process The ISPS analytical process is a six-step process from a raw sample to statistical results as shown in Figure 1. Optical microscopy, Raman dispersive spectroscopy and RCI were used to measure the particle size distribution of polystyrene spheres as well as the budesonide API in an actuated nasal suspension. All data was collected using a FALCON II™ Wide-Field Raman Chemical Imaging System (ChemImage
Percentage
0% Maximum Chord (µm)
C
D10 (µm)
1.1
2.2
30%
D50 (µm)
3.5
3.8
Global Processing
D90 (µm)
8.5
7.8
Local Processing
Standard Deviation (µm)
3.1
2.7
20%
Figure 6. Ingredient-Specific Particle Sizing of budesonide in a Rhinocort® AquaTM actuated droplet: (A) Brightfield reflectance / Raman fusion image after local processing, (B) Budesonide particle size distribution histogram for global and local processing and (C) Summary table of particle size distribution for global and local processing
15%
B
Statistic
Actual
Global Processing
Local Processing
106
107
Total Microspheres Mean (µm)
0.99
1.34
1.03
Standard Deviation (µm)
0.03
0.32
0.10
Figure 4. NIST-traceable particle sizing standard measurement of 1 µm polystyrene microspheres: (A) Equivalent circle diameter particle size distribution histogram for global and local processing and (B) Summary table of particle size distribution for global and local processing The same approach was used for the formulated nasal spray, Rhinocort® AquaTM. Raman spectra were obtained for the API (budesonide) and all excipients present in the nasal spray formulation as shown in Figure 5. The dispersive Raman spectrum of the budesonide exhibits a characteristic C=C feature at 1657 cm-1 that can be used to discriminate it from excipients in the formulated sample including the potassium sorbate excipient with a C=C feature at 1648 cm-1.
960 60
BUDESONIDE
1010
1020
Figure 3. Ingredient-Specific Particle Sizing of polystyrene microspheres using the local processing algorithm: (A) Brightfield reflection optical image of polystyrene microspheres, (B) Raman chemical image extract at 1000 cm-1, (C) Brightfield reflection image with overlaid particle outline from Raman chemical image and (D) Normalized polystyrene Raman spectrum
B
C
10 µm
D
1635
1645
1655
1665
1675
1685
Figure 7. Single field of view of budesonide particles in a Rhinocort® AquaTM droplet: (A) Brightfield reflectance optical, (B) Global processed Raman Chemical Image, (C) Local processed Raman Chemical Image and (D) Normalized Raman spectra of the locally processed particles.
POLYSORBATE 80
POTASSIUM SORBATE
Conclusion
DEXTROSE
AVICEL
EDTA 1200
1400
1600
1800
2000
2200
2400
2600
2800
3000
Figure 5. Raman dispersive spectra of Rhinocort® AquaTM formulation components. The vertical dashed lines define the spectral boundaries used for the Raman Chemical Imaging measurements 980 990 1000 Raman Shift (cm-1)
A
Raman Shift (cm-1)
Raman Shift (cm-1)
970
A comparison of global and local processing at the single field of view level is shown in Figure 7. In this field of view, both processing methods resolved the larger, budesonide particles, but only the local processing identified the smaller budesonide particles. Raman spectra of all four detected budesonide particles are shown to demonstrate the validity of retaining the additional particles lost during global processing.
1625
1000
D
Normalized Intensity
6. Brightfield Guided Binarization
60
5%
35%
3000
6. Statistical analysis
Frequency
5. Gaussian Filter
Raman Chemical
2400
10 µm
1. Cosmic Filter
Brightfield Reflectance
2000 2200 Raman Shift (cm-1)
Normalized Intensity
4. Brightfield Reflectance & Raman Chemical Imaging
1800
10%
363
Offset Intensity
EXCIPIENT #3
1600
15%
313
Equivalent Circle Diameter (µm)
API
1400
Local Processing
Total Particles
0%
Results & Discussion
Global Processing
Local Processing
5%
EXCIPIENT #2
1200
20%
Global Processing
10%
Figure 2. Ingredient-Specific Particle Sizing using a ChemImage FALCON II™ Chemical Imaging System
B
Statistic
25%
EXCIPIENT #1
1000
B 25%
100 µm
Formulation
Pure components
A
Normalized Intensity
No validated method currently exists for characterizing the active pharmaceutical ingredient (API) particle size distribution in nasal aerosol and spray products despite the request of such data for bioequivalence (BE) testing for new drug applications (NDAs) and abbreviated new drug applications (ANDAs) (1). A qualitative and semi-quantitative estimation of drug and aggregated drug particle size distribution is recommended based on optical microscopy, but insoluble suspending agents found in nasal spray formulations typically complicate the Ingredient-Specific Particle Size (ISPS) measurements (2).
Percentage
A comparison of local and global processing results for 1 µm polystyrene microspheres is shown in Figure 4. A total of 107 and 106 microspheres were observed using the local and global processing approaches respectively. Local processing yielded a narrower equivalent circle diameter compared to global processing: 1.03 ± 0.10 µm and 1.34 ± 0.32 µm respectively. Additional global processing was performed on the entire concatenated data sets of 1 and 5 µm microspheres resulting in the loss of all 1 µm microspheres when attempting to accurately size the 5 µm spheres according to a brightfield guided binarization of the 5 µm microspheres.
Offset Intensity
Introduction
Corporation, Pittsburgh, PA) with 532 nm laser excitation. The FALCON II is shown in Figure 2. Brightfield reflectance and Raman chemical images were collected over a15 x 15 matrix or 225 total fields of view yielding a sampling area of 565 x 565 µm2 in size. Imaging data was analyzed using the ChemImage Xpert™ software package (Version 2.3.1, ChemImage Corporation, Pittsburgh, PA) yielding both the Raman / brightfield fusion images as well as the budesonide particle statistics.
The budesonide particle size distribution in a Rhinocort droplet was determined via global and local processing of the brightfield reflectance and Raman chemical images. The maximum chord, longest distance across a particle, was used as the metric for evaluating particle size distributions since the drug particles are not spherical. Figure 6 illustrates global and local processing results for the budesonide particle size distribution in the acquired data.
RCI is capable of producing an ingredient-specific particle size distribution of multiple components in a complex corticosteroid aqueous nasal spray formulation. Budesonide particles were clearly distinguished from those of excipients in a dried Rhinocort® AquaTM spray. Local processing incorporating multiple imaging modalities and image fusion detected and sized smaller budesonide particles initially lost after global processing. Thorough validation of RCI for ingredientspecific particle size distribution requires significant additional research; RCI coupled with optical imaging shows promise as a method for particle size analysis and shape characterization of corticosteroids in aqueous nasal spray suspension formulations which can directly benefit the BE requirements for NDAs and ANDAs.
References 1. Food & Drug Administration, “Critical Path Opportunities for Generic Drugs,” available at http://www.fda.gov/oc/ initiatives/criticalpath/reports/generic.html#sprays (2007). 2. Food & Drug Administration, “Draft Guidance for Industry Bioavailability and Bioequivalence Studies for Nasal Aerosols and Nasal Sprays for Local Action,” available at http://www.fda.gov/cder/guidance/5383DFT.pdf (2003). 3. Doub, W.H. et al. (2007), “Raman Chemical Imaging for Ingredient-Specific Particle Size Characterization of Aqueous Suspension Nasal Spray Formulations,” Pharm. Research, Vol 24, No 5, pp 934-45.
TM
Ryan J. Priore, Oksana Olkhovyk, Oksana Klueva and Michael Fuhrman ChemImage Corporation, Pittsburgh, PA , USA
Raman Chemical Imaging (RCI) has been studied as a potential measurement tool for determining the particle size distribution of a corticosteroid in the formulation matrix (3). RCI combines the chemical identity and selectivity of Raman spectroscopy with digital imaging to determine the particle size distribution and association/ agglomeration of micronized drug in aqueous suspension formulations of nasal sprays. Improved automation across both hardware and software in wide-field chemical imaging is enabling RCI to measure larger populations of ingredient-specific particles as well as incorporate repetitive sampling methods. We have applied RCI to characterize the budesonide API particle size in a sample of Rhinocort® AquaTM nasal spray employing an automated method of data collection and analysis of Raman chemical and brightfield imaging. The advantages of automated software processing of particles on a particle by particle basis are compared to the current global method for determining the particle size distribution of a corticosteroid.
Experimental Polystyrene microsphere suspensions (Polysciences, Inc., Warrington, PA), 0.99 ± 0.03 µm and 5.34 ± 0.26 µm in diameter (mean ± standard deviation), were used to develop the concept of local processing of individual fields of view and ultimately individual particles. Brightfield reflectance and Raman chemical images over the 1000 cm-1 band were collected over individual fields of view until a minimum of 100 microspheres of each diameter were obtained, then these fields of view were concatenated together to form a dataset. A Rhinocort® AquaTM nasal spray (32 mcg budesonide, AstraZeneca, Wilmington, DE) sample was prepared by shaking, priming (four actuations each) and spraying in an upright position onto an inverted aluminum-coated glass microscope slide positioned approximately 15 cm above the spray nozzle. The microscope slide was then immediately turned upright and the nasal suspension droplets were allowed to dry. Actuated samples were analyzed to include actuation device influence as opposed to bulk samples. 1. Sample preparation
2. Raman spectroscopy of pure components
3. Optimize spectral range for Raman Chemical Imaging
A
Global processing uses one threshold for all particles based on pixel intensity distribution. Secondary scattering and the reliance upon spectral normalization to flat-field the chemical image typically leads to RCI oversizing medium and large particles when using global processing methodologies, while small particles are sometimes lost. Local processing treats the identified particles individually resulting in a more accurate representation of the ingredient-specific particle size distribution while also retaining smaller particles through the following data processing routine: 1. Individual API particles are initially identified by their specific Raman spectrum 2. Each particle is treated with a unique intensity threshold unlike in global 3. processing where a single threshold value is utilized to binarize the Raman 4. chemical image and thus yielding a particle map 3 A feedback loop is initiated which confirms the chemical identity against the 5. 6. Raman spectrum and validates the particle size against the brightfield optical 7. image A representative set of polystyrene particle data is illustrated in Figure 3 to demonstrate the use of RCI information to size the microspheres. A
C
B
EXCIPIENT #4 EXCIPIENT #5
5. Data processing
2. Bias Correction 3. Vector Normalization 4. API Peak Frame extraction
2600
2800
7. Boundary Particles Removed
Total Particles
50 40
Standard Deviation
3.1 µm
D10
1.1 µm
D50
3.5 µm
D90
8.5 µm
30 20 10 0
2
313
4
6
8 10 12 14 Max Chord (µm)
16 18
Figure 1. Ingredient-Specific Particle Sizing analytical process The ISPS analytical process is a six-step process from a raw sample to statistical results as shown in Figure 1. Optical microscopy, Raman dispersive spectroscopy and RCI were used to measure the particle size distribution of polystyrene spheres as well as the budesonide API in an actuated nasal suspension. All data was collected using a FALCON II™ Wide-Field Raman Chemical Imaging System (ChemImage
Percentage
0% Maximum Chord (µm)
C
D10 (µm)
1.1
2.2
30%
D50 (µm)
3.5
3.8
Global Processing
D90 (µm)
8.5
7.8
Local Processing
Standard Deviation (µm)
3.1
2.7
20%
Figure 6. Ingredient-Specific Particle Sizing of budesonide in a Rhinocort® AquaTM actuated droplet: (A) Brightfield reflectance / Raman fusion image after local processing, (B) Budesonide particle size distribution histogram for global and local processing and (C) Summary table of particle size distribution for global and local processing
15%
B
Statistic
Actual
Global Processing
Local Processing
106
107
Total Microspheres Mean (µm)
0.99
1.34
1.03
Standard Deviation (µm)
0.03
0.32
0.10
Figure 4. NIST-traceable particle sizing standard measurement of 1 µm polystyrene microspheres: (A) Equivalent circle diameter particle size distribution histogram for global and local processing and (B) Summary table of particle size distribution for global and local processing The same approach was used for the formulated nasal spray, Rhinocort® AquaTM. Raman spectra were obtained for the API (budesonide) and all excipients present in the nasal spray formulation as shown in Figure 5. The dispersive Raman spectrum of the budesonide exhibits a characteristic C=C feature at 1657 cm-1 that can be used to discriminate it from excipients in the formulated sample including the potassium sorbate excipient with a C=C feature at 1648 cm-1.
960 60
BUDESONIDE
1010
1020
Figure 3. Ingredient-Specific Particle Sizing of polystyrene microspheres using the local processing algorithm: (A) Brightfield reflection optical image of polystyrene microspheres, (B) Raman chemical image extract at 1000 cm-1, (C) Brightfield reflection image with overlaid particle outline from Raman chemical image and (D) Normalized polystyrene Raman spectrum
B
C
10 µm
D
1635
1645
1655
1665
1675
1685
Figure 7. Single field of view of budesonide particles in a Rhinocort® AquaTM droplet: (A) Brightfield reflectance optical, (B) Global processed Raman Chemical Image, (C) Local processed Raman Chemical Image and (D) Normalized Raman spectra of the locally processed particles.
POLYSORBATE 80
POTASSIUM SORBATE
Conclusion
DEXTROSE
AVICEL
EDTA 1200
1400
1600
1800
2000
2200
2400
2600
2800
3000
Figure 5. Raman dispersive spectra of Rhinocort® AquaTM formulation components. The vertical dashed lines define the spectral boundaries used for the Raman Chemical Imaging measurements 980 990 1000 Raman Shift (cm-1)
A
Raman Shift (cm-1)
Raman Shift (cm-1)
970
A comparison of global and local processing at the single field of view level is shown in Figure 7. In this field of view, both processing methods resolved the larger, budesonide particles, but only the local processing identified the smaller budesonide particles. Raman spectra of all four detected budesonide particles are shown to demonstrate the validity of retaining the additional particles lost during global processing.
1625
1000
D
Normalized Intensity
6. Brightfield Guided Binarization
60
5%
35%
3000
6. Statistical analysis
Frequency
5. Gaussian Filter
Raman Chemical
2400
10 µm
1. Cosmic Filter
Brightfield Reflectance
2000 2200 Raman Shift (cm-1)
Normalized Intensity
4. Brightfield Reflectance & Raman Chemical Imaging
1800
10%
363
Offset Intensity
EXCIPIENT #3
1600
15%
313
Equivalent Circle Diameter (µm)
API
1400
Local Processing
Total Particles
0%
Results & Discussion
Global Processing
Local Processing
5%
EXCIPIENT #2
1200
20%
Global Processing
10%
Figure 2. Ingredient-Specific Particle Sizing using a ChemImage FALCON II™ Chemical Imaging System
B
Statistic
25%
EXCIPIENT #1
1000
B 25%
100 µm
Formulation
Pure components
A
Normalized Intensity
No validated method currently exists for characterizing the active pharmaceutical ingredient (API) particle size distribution in nasal aerosol and spray products despite the request of such data for bioequivalence (BE) testing for new drug applications (NDAs) and abbreviated new drug applications (ANDAs) (1). A qualitative and semi-quantitative estimation of drug and aggregated drug particle size distribution is recommended based on optical microscopy, but insoluble suspending agents found in nasal spray formulations typically complicate the Ingredient-Specific Particle Size (ISPS) measurements (2).
Percentage
A comparison of local and global processing results for 1 µm polystyrene microspheres is shown in Figure 4. A total of 107 and 106 microspheres were observed using the local and global processing approaches respectively. Local processing yielded a narrower equivalent circle diameter compared to global processing: 1.03 ± 0.10 µm and 1.34 ± 0.32 µm respectively. Additional global processing was performed on the entire concatenated data sets of 1 and 5 µm microspheres resulting in the loss of all 1 µm microspheres when attempting to accurately size the 5 µm spheres according to a brightfield guided binarization of the 5 µm microspheres.
Offset Intensity
Introduction
Corporation, Pittsburgh, PA) with 532 nm laser excitation. The FALCON II is shown in Figure 2. Brightfield reflectance and Raman chemical images were collected over a15 x 15 matrix or 225 total fields of view yielding a sampling area of 565 x 565 µm2 in size. Imaging data was analyzed using the ChemImage Xpert™ software package (Version 2.3.1, ChemImage Corporation, Pittsburgh, PA) yielding both the Raman / brightfield fusion images as well as the budesonide particle statistics.
The budesonide particle size distribution in a Rhinocort droplet was determined via global and local processing of the brightfield reflectance and Raman chemical images. The maximum chord, longest distance across a particle, was used as the metric for evaluating particle size distributions since the drug particles are not spherical. Figure 6 illustrates global and local processing results for the budesonide particle size distribution in the acquired data.
RCI is capable of producing an ingredient-specific particle size distribution of multiple components in a complex corticosteroid aqueous nasal spray formulation. Budesonide particles were clearly distinguished from those of excipients in a dried Rhinocort® AquaTM spray. Local processing incorporating multiple imaging modalities and image fusion detected and sized smaller budesonide particles initially lost after global processing. Thorough validation of RCI for ingredientspecific particle size distribution requires significant additional research; RCI coupled with optical imaging shows promise as a method for particle size analysis and shape characterization of corticosteroids in aqueous nasal spray suspension formulations which can directly benefit the BE requirements for NDAs and ANDAs.
References 1. Food & Drug Administration, “Critical Path Opportunities for Generic Drugs,” available at http://www.fda.gov/oc/ initiatives/criticalpath/reports/generic.html#sprays (2007). 2. Food & Drug Administration, “Draft Guidance for Industry Bioavailability and Bioequivalence Studies for Nasal Aerosols and Nasal Sprays for Local Action,” available at http://www.fda.gov/cder/guidance/5383DFT.pdf (2003). 3. Doub, W.H. et al. (2007), “Raman Chemical Imaging for Ingredient-Specific Particle Size Characterization of Aqueous Suspension Nasal Spray Formulations,” Pharm. Research, Vol 24, No 5, pp 934-45.
