Yanfang Wang. A thesis submitted to The University of Manchester for the degree. of Doctor of Philosophy in the Faculty of Medical and Human

An investigation of visual field test parameters in glaucoma, patterns of visual field loss in diabetics and multispectral imaging of the optic nerve ...
Author: Brenda Tyler
18 downloads 2 Views 4MB Size
An investigation of visual field test parameters in glaucoma, patterns of visual field loss in diabetics and multispectral imaging of the optic nerve head in glaucoma

A thesis submitted to The University of Manchester for the degree of Doctor of Philosophy in the Faculty of Medical and Human Sciences 2012

Yanfang Wang School of Medicine (Human Development)

1

CONTENTS Title page……………………………………………………………1 Contents……………………………………………………….........2 List of Tables………………………………………………………..9 List of Figures……………………………………………………..10 List of Abbreviations……………………………………………...14 Abstract …………………………………………………………...16 Declaration………………………………………………………...17 Copyright statement………………………………………………17 Acknowledgment……………………………………………...…..19

1. Rationale of the study…………………………………………..20

2. Glaucoma……………………………………………………….24 2.1- Classification of glaucoma……………………………………….........24 2.2 - Clinical assessment in glaucoma……………………………………..27 2.2.1- IOP measurement………………………………………………..27 2.2.2 - Examination of structural and functional loss in glaucoma….28 2.3 - Management…………………………………………………………..32

3. Visual field testing……………………………………………..33 3.1 - Stimuli and background……………………………………………...33 3.2 - Test strategies………………………………………………………….34 3.2.1 - Frequency-of-seeing (FOS) curve and threshold………………34 3.2.2 - Supra-threshold strategy………………………………………..36

2

3.2.3 - Threshold strategy……………………………………………….38 3.2.3.1 - Full threshold, Fastpac and SITA………………………….38 3.2.3.2 - 30-2, 24-2 and 10-2 stimulus distributions…………………41 3.3 - Interpretation of results……………………………………………...42 3.3.1 - Reliable indices………………………………………………….42 3.3.2 - Global indices……………………………………………………44 3.3.3 - Glaucomatous changes and progression analysis……………..47 3.4 - Variability of visual field results………………………………………52 3.5 - Factors related with variability………………………………………..53 3.5.1 - Test parameters …………………………………………………..53 3.5.2 - Patient dependent factors………………………………………...54

4. The pupil……………………………………………………………57 4.1 - Pupillary reflexes………………………………………………………...57 4.2 –Control systems of pupil reaction ………………………………………58 4.3 - Factors affecting pupil size ………………………………………..........60

5. Vigilance……………………………………………………………62 5.1 - Definition ………………………………………………………………..62 5.2 - Measurements of vigilance………………………………………………62 5.2.1 - Subjective measurements………………………………………….62 5.2.2 - Objective measurements…………………………………………..63 5.2.2.1- Electroencephalogram (EEG)……… ……………………......63 5.2.2.1.1- EEG activities related to vigilance………………………64 5.2.2.1.2 - Sleepiness measurement…………………………………65 5.2.2.2 - Pupillography ………………………………………………….66

3

5.2.2.2.1 - Spontaneous pupillary oscillations in the dark………...66 5.2.2.2.2 - Spontaneous pupillary oscillations in the light………...68 5.2.2.2.3 - Development of pupillary parameters………………….68 5.2.2.3 - Comparison between EEG and pupillography………………71 5.3 - Vigilance and perimetric examinations………………………………...72 5.3.1 – Pupil perimetry……………………………………………………..72 5.3.2 - Pupil derived vigilance index ………………………………………75 5.3.3 - Wavelet analysis …………………………………………………….76

6. Optic nerve head …………… …………………………………….81 6.1 Anatomy of the normal optic nerve head (ONH)………………………...81 6.1.1 Structure of ONH……………………………………………………..81 6.1.2 Blood supply to ONH…………………………………………………83 6.1.3 Autoregulation of blood-flow in ONH……………………………….84 6.2 Morphological changes of ONH in glaucoma…………………………….85 6.2.1. Optic disc size………………………………………………………...86 6.2.2. Neuroretinal rim……………………………………………………..89 6.2.3. Retinal nerve fibre layer……………………………………………..91 6.2.4. Parapapillary atrophy……………………………………………….92 6.2.5. Disc haemorrhage……………………………………………………93 6.3 Mechanisms of glaucomatous damage in ONH…………………………..94 6.4 Blood flow and oxygenation measure in ONH and related changes in glaucoma …………………………………………………………………………..97 6.4.1 Measure with Doppler techniques …………………………………..97 6.4.2 Measure with Multispectral imaging ……………………………….98

4

7. Identifying the state of attention during perimetry with pupil dynamics and EEG data……………………………………………101 Contributions………………………………………………………………….101 Publication …………………………………………………………………....101 Abstract………………………………………………………………………...102 7.1 Introduction………………………………………………………………..103 7.2 Methods…………………………………………………………………….104 7.2.1 - Subjects……………………………………………………………....104 7.2.2 - Experimental apparatus……………………………………………105 7.2.3 - Experimental procedures…………………………………………..107 7.2.4 - Data analysis………………………………………………………..108 7.2.4.1 - Pupil data……………………………………………………..108 7.2.4.2 - EEG data……………………………………………………...109 7.2.4.3 - Correlation between pupil and EEG data…………………..111 7.2.4.3.1. Relationships with long term sampling………………..111 7.2.4.3.2. Relationships with short term sampling………………111 7.2.4.4 - Threshold variability………………………………………...114 7.2.4.5 - Correlation between pupil and threshold sensitivity………116 7.3. Results …………………………………………………………………….116 7.3.1 - Threshold sensitivity………………………………………………116 7.3.2 - Pupillary changes………………………………………………….117 7.3.3 - Correlation between pupil dynamics and EEG data……………118 7.3.3.1. Relationships between pupil size and EEG spectral amplitude with long term sampling (1minute)……………………………………………..118 7.3.3.2. Relationships between pupil size and EEG spectral amplitude with short term sampling (2 s)………………………..........................................119 7.3.3.3. PFW amplitude and EEG spectral amplitude with long term sampling (1minute) ……………………………………………………………...121

5

7.3.3.4. PFW amplitude and EEG spectral amplitude with short term sampling (2 s)……………………………………………………………………..122 7.3.3.5. Summary of correlation between pupil dynamics and EEG activity ……………………………………………………………………………124 7.3.4 - Correlation between pupil dynamics and threshold sensitivity...124 7.3.4.1. Pupil size parameters and threshold sensitivity variability..124 7.3.4.1.1. Mean pupil size and threshold variability……………...124 7.3.4.1.2. Slope of pupil size and threshold variability…………...125 7.3.4.1.3. Cumulative miosis and threshold variability…………..126 7.3.4.2. PFW amplitude and threshold variability…………………..127 7.3.4.3. Summary of the relationship between pupil dynamics and threshold variability……………………………………………………………...128 7.4. Discussion.…………………………………………………………………128 7.4.1 - Pupillary changes and EEG activity………………………………128 7.4.2 - Vigilance loss and test-retest threshold variability……………….131

8. Blink frequency and duration during perimetry and their relationship to test-retest threshold variability…………………...133 Contribution…………………………………………………………………..133 Publication……………………………………………………………………133 Abstract……………………………………………………………………….134 8.1 Introduction……………………………………………………………….135 8.2 Methods……………………………………………………………………136 8.2.1 Subjects………………………………………………………………136 8.2.2 Apparatus……………………………………………………………137 8.2.3 Experimental tasks…………………………………………………..137 8.2.4 Data analysis…………………………………………………………138 8.3 Results……………………………………………………………………..139

6

8.3.1 Blink characteristics………………………………………………...139 8.3.2 Blinks and stimulus presentation…………………………………..141 8.3.3 Correlation between blink parameters and threshold variability..144 8.3.4. Blinks during stimulus presentation………………………………145 8.4. Discussion…………………………………………………………………146

9. Diagnostic performance of visual field test using subsets of the 242 test pattern for early glaucomatous field loss…………………...150 Contribution…………………………………………………………………..150 Publication…………………………………………………………………….150 Conferences…………………………………………………………………...150 Abstract………………………………………………………………………..151 9.1 Introduction……………………………………………………………….152 9.2 Methods……………………………………………………………………155 9.3 Results……………………………………………………………………..158 9.3.1 Visual field test performance……………………………………….158 9.3.2. Characteristics of visual field defects detected with PPV-optimized locations…………………………………………………………………………..161 9.3.3. Estimated test time for subsets of 24-2 pattern………………….162 9.4 Discussion…………………………………………………………………162

10. Spatial patterns of central visual field loss in proliferative diabetic retinopathy and related changes after Optos ®-guided Pascal® laser treatment……………………………………………..167 Contribution…………………………………………………………………..167 Publication…………………………………………………………………….167 Conferences…………………………………………………………………...167 Abstract………………………………………………………………………...168

7

10.1 Introduction………………………………………………………………169 10.2 Materials and methods…………………………………………………..172 10.3 Results…………………………………………………………………….178 10.4 Discussion…………………………………………………………………184

11. Changes in the differential light absorption of optic nerve head in glaucoma with reduction of IOP………………………………...188 Contribution………………………………………………………………….188 Conferences…………………………………………………………………..188 Abstract……………………………………………………………………….189 11.1 Introduction……………………………………………………………...190 11.2 Methods…………………………………………………………………..193 11.3 Results……………………………………………………………………200 11.4 Discussion………………………………………………………………..205

12. Summary and Conclusions……………………………………..210 Appendix……………………………………………………...……..214 References…………………………………………………………...216

8

List of Tables Table 9.1 - Test duration of SITA standard visual field test ……………………..153 Table 9.2 - Areas under the ROC curve (AUROC) for the optimized and randomized patterns (R1-R5) and the p-value of the difference………………………….........160 Table 9.3 - Test time estimated for variable subsets of 24-2 test pattern…………162 Table 10.1 - Study major inclusion and exclusion criteria………………………..173 Table 10.2 - Modified Airlie House classification of proliferative diabetic retinopathy………………………………………………………………...177 Table 11.1 - Exposure time for each monochromatic image at different wavelengths………………………………………………………………..196 Table 11.2 - DLA changes in different diagnostic groups based on median grating from five experts………………………………………………………......202

9

List of Figures Fig 2.1- The theoretical relationship between the OCT-estimated retinal nerve fibre layer thickness and visual field sensitivity measured with standard automatic perimetry and relevant variation in normal (95% Confidence Interval)…………....31 Fig 3.1 - FOS curve: threshold is defined at the point of 50% of seeing the stimulus……………………………………………………………………………..35 Fig 3.2 - The 4-2 dB staircase algorithm with two reversals (left) and 3dB staircase algorithm with single crossing (right) for threshold intensity……………………...39 Fig 3.3 - The two probability functions (normal and abnormal) for threshold estimation in Swedish Interactive Threshold Algorithms…………………………..40 Fig 3.4 - Distributions of stimulus presentation in the programme of 30-2, 24-2 and 10-2 patterns…………………………………………………………………….......42 Fig 3.5 - An example of a Bebie curve with local defect…………………………..45 Fig 3.6 - The sectors symmetrical above or below the horizontal meridian in glaucoma hemifield test………………………………………………………...…..47 Fig 3.7 - Glaucoma staging system chart……………………………………….......50 Fig 4.1 - Neural network for pupillary movements………………………...………60 Fig 5.1 - 10/20 system of EEG electrodes arrangement…………………………....63 Fig 5.2 - EEG activities of different frequencies…………………………...............64 Fig 5.3 - Spontaneous pupillary movements recorded from healthy alert and tired subjects in the darkness…………………………………………………………….67 Fig 5.4 - EEG activity, pupil size change and the relative beta activity and percentage of maximum pupil size observed during a 2-minute least stable period……………………………………………………………………………….71 Fig 5.5 - Pupil responses to the On1 stimulus pattern of multifocal pupil perimetry……………………………………………………………………………74 Fig 5.6 - The screening window used in the STFA method………………………..77 Fig 5.7 - Domains obtained by Fourier, STFT and wavelet analysis and the corresponding screen windows (a); Daubiechies family and the scales of wavelet analysis (b)………………………………………………………………………….78 Fig 5.8 - The approximate (An) and detailed (Dn) coefficients obtained at each level in wavelet analysis (a) and reconstructed components of an EEG signal (b)………80

10

Fig 6.1- Ophthalmoscopic appearance of ONH………………………..…………..81 Fig 6.2 - The logarithmic relationship between the retinal ganglion cell axons and age in years…………………………………………………………………………82 Fig 6.3 - Blood supply to the ONH…………………………………………………83 Fig 6.4 - Example of the observation of the scleral ring (dark circle) and assessment of the optic disc size………………………………………………………………...86 Fig 6.5 - Relationship between the optic disc size and cup size (left) and cup/disc ratio (right) in normal……………………………………………………………….87 Fig 6.6 - The disc damage likelihood scale…………………………………………89 Fig 6.7 - The ISNT rule of neuroretinal rim area…………………………………...90 Fig 6.8 - Localized loss of neuroretinal rim in the inferior region (white arrow, left image) and general loss of neuroretinal rim (right image)…………………………91 Fig 6.9 - An example of localized inferior and superior RNFL loss (white arrows) in glaucomatous eye…………………………………………………………………..92 Fig 6.10 - An example of eye with parapapillary atrophy (Black arrows: scleral ring. White arrows: β zone. White arrowheads: α zone)…………………………………92 Fig 6.11 - Splinter-like disc haemorrhage (white arrows)………………………….93 Fig 6.12 - Paradigm of the biomechanical mechanism of glaucoma……………….96 Fig 6.13 - Light absorption of blood components in visible spectrum……………..99 Fig 7.1 - Time frame of data acquisition in Signal 2.05 programme………….…..107 Fig 7.2 - Result from wavelet analysis: 1) raw pupil signal; 2) pupil signal after removal of high frequency noise; 3) extracted PFW; 4) amplitude of PFW..…….109 Fig 7.3 - Amplitude of EEG spectrum from the 3 electrodes……………………..110 Fig 7.4 - Flow chart of calculating intra-test threshold variability………………..115 Fig 7.5 - Examples of pupillary changes during the test……………………….....118 Fig 7.6 - Mean pupil size and mean amplitude of theta, alpha and beta in the first (left) and second test (right) for subject C.D………………………...……………119 Fig 7.7 - Mean relative power of theta (top image) and alpha (bottom image) activities in different pupil stages…………………………………………………121 Fig 7.8 - Mean PFW amplitude and mean amplitude of theta, alpha and beta in the first (left) and second test (right) for subject C.D…………………………………122

11

Fig 7.9 - Mean relative power of theta (top) and alpha (bottom) activities in different PFW stages………………………………………………………………………...123 Fig 7.10 - Relationship between mean pupil size and threshold variability within (top) or between tests (bottom)……………………………………………………125 Fig 7.11 - Relationship between cumulative miosis and test-retest (top graph) and intra-test (bottom graph) threshold variability…………………………………….126 Fig 7.12 - Relationship between mean PFW amplitude and test-retest (top) and intratest (bottom) threshold variability…………………………………………………127 Fig 8.1 - Blink frequency (top), blink duration (middle) and number of microsleeps (bottom) for each patient in the first and second tests…………………………….140 Fig 8.2 - Blink frequency (left) and duration (right) in the first and second test…141 Fig 8.3 - Cumulative plots of blink eye lid closures, time locked to stimulus presentation, for 3 patients ………………………………………………………..142 Fig 8.4 - Average percentage of eye lid closures before and after stimulus presentations within each of the 3 groups ………………………………..……….143 Fig 8.5 - Relationship between blink parameters and threshold variability…...….144 Fig 8.6 - Overall percentage of seen stimuli and the percentage of seen stimuli with various extents of overlap between blinks and stimulus presentation …………....145 Fig 9.1 - Flowchart of Predictive Positive Values (PPVs) calculation……………157 Fig 9.2 - The sensitivity and specificity of the optimized test location pattern in defect group……………………………………………………………………….159 Fig 9.3 - Optimized distributions consisting of 10, 20, 30 and 43 test locations….159 Fig 9.4 - Receiver operating characteristics curves of visual field test with the PPVoptimized and five randomized test location patterns……………………………..160 Fig 9.5 - The Mean Deviation (MD), Pattern Standard Deviation (PSD) and number of defective locations detected with every ten locations following the optimized test location patterns…………………………………………………………………...161 Fig 10.1 - The retinal image and distribution of test locations with 24-2 perimetric test pattern…………………………………………………………………………175

12

Fig 10.2 - Self-organized mapping patterns of SITA 24-2 visual field loss in untreated proliferative diabetic retinopathy (top graph). Mean and standard deviation of MD and PSD for each pattern are displayed in bottom graph…………………180 Fig 10.3 - Number of eyes observed in nine SOM patterns before and after laser treatment………………………………………………………………………….181 Fig 10.4 - The mean difference and standard deviation of MD and PSD between baseline and 3-month post treatment for eyes shifted to less advanced SOM patterns, no change of SOM patterns and shifted to more advanced patterns………………182 Fig 10.5 - Number of locations with TD at p95%), thus their ability to identify early glaucomatous loss is limited (Zangwill et al., 2001, Ford et al., 2003). Additionally, the relationship between the pattern of visual field loss and the ONH is complex and not fully understood. Further research in this topic area may lead to new insights into this disease.

Fig 2.1 The theoretical relationship between the OCT-estimated retinal nerve fibre layer thickness and visual field sensitivity measured with standard automatic perimetry (SAP) and relevant variation in normal (95% CI) (reproduced from (Hood and Kardon, 2007)). OCT=optic coherence tomography, CI=confidence interval, IT=Inferior temporal, SA=Superior arcuate.

31

2.3 Management The goal of glaucoma treatment is to preserve adequate visual function for the patient daily life. To achieve this target there are 3 therapeutic options: medication, laser treatment and surgery. The first line treatment is normally to medically reduce the IOP with beta-blockers and/or prostaglandins (NICE clinical guideline, 2009). Usually a target IOP is established on the basis of the current IOP level, stage of the disease at diagnosis, rate of disease progression and other factors (i.e. age and life expectancy) (European Glaucoma Society, 2008). The beneficial effect of different types of medical IOP reduction treatment was reviewed by Vass who conducted a meta-analysis of trials (Vass et al., 2007). Evidence from randomized controlled clinical trials has shown that patients with decreased IOP could have reduced rates of disease progression (Heijl et al., 2002, CNTGS group, 1998, AGIS, 2000) or a delayed onset of POAG in subjects with increased IOP (Kass et al., 2002). However, there are subgroups who despite successful reduction of IOP to target levels continued to progress and subgroups who without treatment showed no progression. These findings indicate that there are other factors that contribute to the progression of the disease. To further evaluate the effect of medical treatment on glaucoma, a placebo-controlled randomized clinical trial has been undertaking in the UK (Garway-heath et al., 2013).

32

3. Visual field testing

The visual field is the volume of space which you can see when your eyes are open. Normally, the extent of visual field is about 60 degrees superior, 75 degrees inferior, 110 degrees temporal and 60 degrees nasal under bright light conditions (Stamper et al., 1999). A physiological blind spot exits for each eye corresponding to the location of optic disc while pathological defects will exist in conditions that damage the retina and visual pathways (i.e. glaucoma, optic neuritis). Currently, the standard method of testing the visual field is SAP, which can be undertaken with a range of perimeters including Humphrey (Zeiss Meditec, Dublin, California, USA), Octopus (Interzeag, Koeniz, Switzerland) and Henson (Elektron, Cambridge, UK) field analyzer. The most common type used in hospital clinics is Humphrey Field Analyzer (HFA).

3.1 Stimuli and background In a perimetric test, individuals are asked to respond to a series of stimuli (static or kinetic) presented in different locations within their visual field. The static stimuli, which have a constant size but vary in luminance, are used in the vast majority of current tests and this thesis will concentrate on this technique. The SAP stimuli are achromatic light spots displayed on a white background (white-on-white). In a HFA, the size of standard stimulus is Goldmann Ⅲ (~0.5 degrees) and the background

33

luminance is 31.5asb. Other alternative stimuli/backgrounds have also been developed, e.g. the Blue-on-Yellow perimetry, ring perimetry and Frequency Doubling Technology (FDT) (Henson, 2000).

3.2 Test strategies 3.2.1 Frequency-of-seeing (FOS) curve and threshold sensitivity The threshold is usually defined as the stimulus intensity (in decibels) which is seen 50 percent of the time. A good way to estimate the threshold is to obtain an FOS curve by repeated testing at a range of intensities (Chauhan et al., 1993, Olsson et al., 1992). It is an S-shaped curve plotting the probability of seeing stimuli (positive response) against the stimulus intensity (Fig 3.1). When the intensity is dimmer than the threshold value (subthreshold), the probability of seeing the stimulus will be less than 50%. When the intensity of stimulus is greater than the threshold (suprathreshold), it will be greater than 50%. The existence of false negative (FN) and false positive (FP) responses means that the curve never reaches 100 and 0% (Flammer et al., 1984a).

34

Fig 3.1 FOS curve: threshold is defined at the point of 50% of seeing the stimulus. (the figure is reproduced from (Flammer et al., 1984a)). FOS=frequency of seeing.

The slope or standard deviation of the FOS curve is a measure of response variability, in which a steeper curve (smaller standard deviation) means less variability and a shallower curve (larger standard deviation) means greater variability (Henson et al., 2000). Although the FOS curve accurately defines the threshold, it requires a large number of stimuli presentations and cannot be used to monitor transient changes in threshold or to derive measures of threshold from a large number of test locations in a clinical environment.

35

3.2.2 Supra-threshold strategy Supra-threshold perimetry is a fast way to detect visual field defects without quantifying their depth (how much they depart from age corrected normal values). It is frequently used for screening (Heijl, 1976). In this method, the hill of vision is used to establish a stimulus level which is usually about 4~6dB above the expected threshold value. The test intensity may be determined either by an initial quick threshold test at a few test locations (threshold-related) or by the age of the patient (age-related) (Henson, 2000). The threshold-related value is considered to be more accurate. However, while this may be true for normal field locations it may not be for cases where there is a visual field defect. When the primary seed locations (those used to establish the supra-threshold increment) are normal, the threshold-related method was observed to be more accurate than the age-related method. With more than 2 defective seed locations the age-related method was observed to be more accurate (Henson and Artes, 2002).

The conventional criterion for defining a location as defective in supra-threshold perimetry is when it is missed twice out of 2 presentations. When the stimulus is perceived at the first exposure, it will be defined as ‘normal’. When the first presentation is missed, it will be presented a second time. This criterion reduces FP errors, but also reduces sensitivity by a small amount (Artes et al., 2003). Artes looked at various other pass/fail criteria and concluded that more accurate (better sensitivity/specificity) results could be obtained with a fail criteria of 3 misses in up

36

to 5 presentations (Artes et al., 2001, Artes et al., 2003). This criteria was found to detect the glaucomatous damage earlier than the conventional criteria but at the expense of longer test times (Artes et al., 2001).

Apart from the pass/fail criteria, improvements on other aspects of supra-threshold test have also been developed. With the use of multiple rather than single stimuli the sensitivity increases and variability reduces. This may be due to the better control of attention with the multiple stimuli pattern which use verbal responses (Miranda and Henson, 2008). To compensate for the sensitivity difference with eccentricity, most supra-threshold tests use a normal database to calculate the test intensity at each test location which takes account of variations with eccentricity and test location and yields improved test performance (Henson, 2000). ‘HEART’ is an algorithm developed by Henson et al (2000) for establishing the supra-threshold test intensity. It starts at 1dB brighter than the age-related normative value and uses a staircase procedure (1dB steps) terminating after 6 presentations at each of the four seed locations. The final threshold is the mean of the last four presentations at all locations which is then used to provide the final estimation of visual field’s general height. It reduces the effect of defects at seed locations on final estimation by excluding locations with defects over 2.8dB below the age-related normative value and converting to an age-related setting when the final estimation fall 4dB below the age-related mean value. This algorithm was found to be less variable than either the

37

HFA or age-related methods (Henson and Artes, 2002). Using response times to detect and remove FP responses also improves test performance (Artes et al., 2002b).

3.2.3 Threshold strategy 3.2.3.1 Full threshold, Fastpac and SITA Threshold strategies are used to derive an estimate of the depth of any visual field defect. Modern threshold strategies are Full Threshold, Fastpac and Swedish Interactive Threshold Algorithms (SITA).

The Full Threshold strategy is considered to be the standard method and has been widely used throughout the world. In this test procedure, four locations (one in each quadrant) are initially tested and serve as the baseline for adjacent locations. These points will then serve as the baseline for their neighbouring points etc. The intensity of the stimulus follows a repeated up and down staircase algorithm with a step size that changes from 4 to 2dB after the first reversal (Fig 3.2). In this algorithm, the stimuli cross the threshold twice and the final estimation is determined as the last seen value for a given location.

38

Fig3.2 The 4-2 dB staircase algorithm with two reversals (left) and 3dB staircase algorithm with single crossing (right) for threshold intensity.

Fastpac adopts a similar staircase algorithm for the stimulus presentation but uses a 3dB step and a single crossing (Fig 3.2). The result is also determined by the final stimulus presentation. Compared to Full Threshold method, Fastpac is faster but less accurate and more variable (Heijl and Patella, 2002).

SITA was developed in 1990s to provide similar accuracy to the Full Threshold method but with shorter test durations (Bengtsson et al., 1997). It also starts testing at four seed locations and uses a 4-2 dB staircase strategy. During the test, two probability functions for threshold estimation are calculated for each location, one for a normal and one for a defective result (Fig 3.3). These are updated after each stimulus presentation. The testing of each location is terminated when a required level of certainty is obtained (width of the function). The amount of certainty needed for termination is different in SITA standard and SITA fast (Bengtsson and Heijl,

39

1998). The FP rate in SITA is estimated from the patients’ response times rather than using catch trials. This makes the test faster and gives a more consistent estimate of the FP rate (Bengtsson et al., 1997).

Log likelihood Intensity (dB)

Fig 3.3 The probability functions (normal and abnormal) for threshold estimation in Swedish Interactive Threshold Algorithms (Reproduced from Bengtsson et al, 1997). The maximum likelihood of function is updated with patients’ response to stimulus presentations and the abnormal function becomes the most likely one as the test continues in this example.

40

Compared to Full Threshold and Fastpac tests, both SITA standard and SITA fast show a significant decrease of testing time (mean durations: Full threshold, 12.94min; Fastpac, 7.69min; SITA std, 6.56min; SITA fast, 3.84min) and an accuracy that is similar to the Full Threshold method (Artes et al., 2002a, Wild et al., 1999). The test-retest variability was observed to be lower in measurements with SITA standard than the Full Threshold strategy while SITA fast only presented smaller variability in high sensitivity areas (>25 dB) (Artes et al., 2002a). The sensitivity and specificity for glaucoma detection with SITA standard is higher than the fast method (SITA standard: 96% and 98%; SITA fast: 96% and 95% respectively) (Budenz et al., 2002). These advantages make the SITA standard a more suitable strategy for routine testing.

3.2.3.2 30-2, 24-2 and 10-2 stimulus distributions The most common stimulus distributions used in current visual field examinations are central 30-2 and 24-2. The 30-2 programme tests the central 30 degrees of visual field with 76 locations located on a square matrix of 6 degrees displaced from the horizontal and vertical midlines by 3 degrees. Similarly, the 24-2 programme tests a subset of the 30-2 locations, those falling within the central 24 degrees along with two points at 27 degrees in the nasal field. The 10-2 tests 68 locations within the central 10 degrees (Fig 3.4) on a 2 degrees square matrix. In the three types of distribution, 30-2 provides the most information for the central visual fields while 24-2 has a shorter test duration and smaller variability than 30-2 programme

41

(Khoury et al., 1999). Another reason for the routinely use of 24-2 program in most ophthalmic clinics is that the defect found only in the additional 30-2 locations is normally attributed to a lens rim artefact rather than a true glaucomatous defect. In patients with small central fields, the 10-2 programme can be adopted to evaluate the residual visual function with higher spatial resolution.

Fig 3.4 Distributions of stimulus presentation in the programme of 30-2 (the round area in the left plot), 24-2 (the square area in the left plot) and 10-2 patterns (right). (The figures are modified from Heijl and Patella, 2002)

3.3 Interpretation of result 3.3.1 Reliable indices In the printout of a HFA threshold test, there are three indices of reliability, which include the fixation loss (FL), FP and FN. To confirm the fixation of the patient, stimuli are occasionally presented in the blind spot. If the patient’s fixation changed, they may respond to the stimulus which would be recorded as a FL. FLs are presented on the chart as N responded/N presented. FP is the response to a non-

42

existent stimulus while FN is a non-response to a previously seen stimulus. They can be acquired through catch trials in SAP, which are extra tests accounting for a small part (3% - 5%) of total number of stimulus presentations (Fankhauser et al., 1977), and SITA derives an estimate of the FP rate from patient response times. Conventionally, the FN is estimated with the random projection of supra-threshold stimuli in already tested locations in catch trials. In the Octopus perimeter, stimuli are presented at 0dB, while in the HFA they are presented at 9dB brighter than the previous seen stimulus in Full Threshold test. With SITA the increment increases in the defect areas. For FP catch trials, it is traditionally estimated with the rate of response to blank presentations. One shortcoming of catch trials is that they have a low level of precision (Vingrys and Demirel, 1998). The method of using response time window is more repeatable although the technique lacks a good evidence base linking response times to FP rates (Bengtsson and Heijl, 2000).

Among the three reliability indices, FP was observed to be the most stable parameter between tests and FL were the least (Bengtsson and Heijl, 2000). A significant relationship was observed only between the threshold sensitivity and FN rate. Most of the published literatures classify a reliable result as having both FP and FN below 33% and FL below 20% (Heijl and Patella, 2002). There is no evidence to support the use of these cut offs. High rates of FP is considered to be indication of “triggerhappy” patients, while the high rate of FN may be due to inattention of subject and can lead to false defect/recovery of glaucoma progression (Fankhauser et al., 1977).

43

3.3.2 Global indices Flammer et al (1985) proposed that the glaucoma-related visual field defects can be classified into the three categories: overall depression of sensitivity, localized loss and the increased variability of thresholds. The Octopus perimeter thus derived 4 global indices; Mean Defect, Loss Variation (LV) and Corrected Loss Variation (CLV), and Short-term Fluctuations (SF) corresponding to the three categories (Flammer et al., 1985). The Mean Defect is the average difference between the test results and the age-corrected normal values at each location. It gives a result in decibels (dB) with values becoming greater in locations with severe defects (Flammer et al., 1985). LV and CLV are parameters used to measure the local field defects, where LV is the variance of defect values and CLV is the variance adjusted for normal variability component (SF). Additionally, a Bebie curve (or Cumulative defect curve) has been developed in the G1 program to separate the local and diffuse damage of visual field (Bebie et al., 1989). It ranks the sensitivity at all locations from the highest to the lowest after age correction and plots the defect level against the rank order (Fig 3.5). As a result, the overall depression and localized loss are reflected in the shape of the curve: when there is only diffuse depression, the curve presents a constant deviation from the normal distribution; when there is localized defect, an abrupt increase of deviation will be observed at the right-hand side of the curve.

44

Fig 3.5 An example of a Bebie curve with local defect (the bold line). The limiting curve represents absolute defect visual field. Curves with 5 th, 50th, 95th, and 99th represent the corresponding percentile of defect distributions at a given rank in normal visual fields. (Reproduced from Bebie et al., 1989).

In the HFA, the index for diffuse depression is Mean Deviation (MD). It is slightly different to the Mean Defect in the Octopus as it incorporates a weighting factor that increases the significance of the more central test locations (Henson, 2000). The calculation of MD is obtained for all tested locations, except the two possible physiological blind spot ones. It is also presented in dB but with negative values. More negative values signify larger defects. The sensitivity values can be viewed as a greyscale beside the tested threshold plot. The HFA index for localized loss is

45

named Pattern standard deviation (PSD). It is the square root of LV which has been adjusted from the normal fluctuation at each test location (Heijl et al., 1987). It reflects the localized visual defect rather than diffused damage, which makes it more sensitive and specific to the patterns of loss seen in early glaucoma. SF is also obtained in HFA (except SITA algorithm) to estimate the intra-test variability via repeated testing in 10 locations during a single session. When PSD is adjusted for the SF level, it provided a parameter named corrected pattern standard deviation (CPSD), which is similar to CLV in the Octopus perimeter.

In addition to those indices introduced above, the HFA includes a map of Total deviation (TD), pattern deviation (PD) and the results from the Glaucoma hemifield test (GHT) (Heijl and Patella, 2002). TD is a plot of the difference between the tested results and expected normal values at each test location, while PD is the difference adjusted for overall shifts in sensitivity such as those that may occur from lens opacity and small pupil size. Below the plots of TD and PD, corresponding probability values are plotted to indicate how significant the difference is at each location (at the level of

Suggest Documents