Box counting serial image index (BCSI) versus point counting serial image index (PCSI) in scoring melasma: A comparative, non-interventional validation study

Introduction

Melasma is a chronic condition requiring frequent follow-up visits. Various scoring methods have been developed.^1-8 A comparative and validation study was performed between the Point Counting Serial Image Index (PCSI), and gold standard modified melasma area severity index (mMASI).⁴

PCSI is based on area measurement and intensity of pigmentation (IP). The modifications to PCSI are for area measurement; however, IP remains the same. The principle was image capture and preset grid placement in a manner that touches the margins of melasma.⁹ This achieves consistency in area measurement irrespective of distance and or angle of image capture. Further, we count boxes of the preset grid, with a speck of pigmentation as one box that helps to assess scattered pigmentation. Box Counting Serial Image Index (BCSI) was developed by applying these modifications to PCSI, and then the two were compared.

Methods

A comparative, non-interventional validation study was conducted in the dermatology OPD, JSS Hospital, JSS Medical College, JSSAHER, Mysore over 1.5 years (September 2022 to March 2024). A purposive sampling of melasma involving bilateral malar areas with or without other sites was conducted. Patients aged above 18 years were enrolled irrespective of their sex. patients who were enrolled in baseline visit and who completed at least 1 follow-up visit after 3 weeks were included. Patients with melasma involving only nose, chin, or forehead without bilateral malar areas, chemical peels/ lasers in the past 8 weeks, on topical treatment for more than 8 weeks, and pregnancy were excluded. Sample size was determined by using the formula n = (Z_α+ Z_β)² * σ²/ d², and the estimate was 103.⁴ Where Zα=1.96, Zβ =0.84, σ =25.53, SD value of PCSI scores, d=7 (the effect that is clinically worthwhile to detect).

The training and calibration of evaluators were done to ensure consistency in assessment. A pilot study was performed, and senior consultants validated the scores of the principal investigator (PI) and co-investigator (CI) for 2 months before the initiation of the study.

On the baseline visit, lesions were recorded separately by PI and CI using BCSI and PCSI, respectively. CI was blinded to eliminate the risk of bias. The research/study was approved by the Institutional Review Board at Institutional Ethical Committee JSSAHER, number JSS/MC/PG/IEC-18/2022-23. Serial clinical photographs were taken without revealing the identity, and informed consent was obtained. Images were captured on a smartphone (iPhone 11, model MWL72LL/A) with a 12MP camera. The quality of images was graded based on the visibility of box counting as good (easy counting), fair (counting with difficulty), and poor (cannot be counted).⁷

The steps involved in BCSI and PCSI are illustrated in Figure 1 and 2 respectively. In the BCSI method, the baseline clinical image of the patient was captured. A preset grid was prepared by the computerised superimposition of a standard graph (1*1 cm) over a polyvinyl chloride transparent matte adhesive sticker. The preset grid can fit into any contours of the body. Matte preset grid minimises the glare.

Close

Figure 1:

Illustrates the methodology of the BCSI: (step 1): The image of the melasma patient is captured, (step 2): Preset Grid approximating the size of the melasma lesion is cut from the Grid sheet, (step 3): The adhesive sticker is removed, and the preset grid is superimposed over the melasma lesion making sure the two edges of the preset grid touch the two margins of the melasma lesion to maintain uniformity. One speck of pigmentation within a box is counted as one, (step 4): Box counting is performed (Number of boxes in the instant case is = 7).

Export to PPT

Close

Figure 2:

Illustrates the methodology of the PCSI: (step 1): The image of the melasma patient is captured, (step 2): Transparent polythene sheet is used to trace the hyperpigmented borders of the melasma lesion, (step 3): The polythene sheet is impregnated into a graph sheet using a carbon paper, (step 4): The number of cross-over points lying inside the traced boundary is counted. (Number of crossover points in this instant case=5).

Export to PPT

The preset grid sheet was superimposed in a manner that the two edges of the preset grid touched the two margins of the melasma lesion to maintain uniformity.⁹ Images were captured under standard light settings. The number of boxes showing hyperpigmentation represents the area of involvement [Figure 1]. A speck of pigmentation within a box was counted as one [Figure 3]. The methodology of BCSI is illustrated in Figure 1 and the Video 1.

Close

Figure 3:

To illustrate exactly how a box/ speck looks like a) number of boxes 10 b) number of boxes 12. A speck of pigmentation is counted as 1 box. Red arrows indicate scattered pigmentation, where speck of pigmentation is counted as 1.

Export to PPT

In the PCSI method, area assessments were done by the point counting method of Bahmer.^4,10 The steps involved in PCSI have been illustrated in Figure 2, Video 1.

The standard and established treatment was given. The first image of the pigmented area was compared with the normal skin and recorded as 5, i.e., 100% (baseline image). Patients were evaluated once in 3 weeks up to 2 follow-up visits. The follow-up images were compared with the baseline image, and the intensity of pigmentation (IP) was recorded on a scale of 0-5 (5=100%, 4=80%, 3=60%, 2=40%, 1=20%, 0=0%) for both methods.⁴

In both methods, the score is calculated as the product of area and IP. The differences in the number of boxes versus cross-over points, IP, and time to calculate were noted and compared. The steps followed for standardisation and validation have been illustrated in Figure 4.

Close

Figure 4:

Illustrates the steps followed in the standardisation and its validation of this study. (BCSI: Box counting serial image index, PCSI: Point counting serial image index, mMASI: Modified melasma area severity index). Step1 is based on the reference 4 Shristi et al, Step 3 is based on the reference 9 Anisha et al.

Export to PPT

Statistics were done using mean, standard deviation, and paired sample t-test, with P < 0.05 considered statistically significant. All statistical methods were done using the SPSS 21.0 version for Windows.

Results

A total of 120 patients with mean age of 35-years were enrolled; 101 completed first follow-ups and 19 were lost to follow-up. There were 10 males (9.9%) and 91 females (90.1%).

Baseline: The area measurements using BCSI (PI) and PCSI (CI) showed mean values of 14.5446 and 6.97212, respectively. The difference was statistically significant (p<.0003) [Table 1] (BSCI>PCSI). All the images taken for comparison were of good quality (easy counting of boxes). The total scores for BCSI and PCSI had mean of 72.7228 and 53.7129, respectively, and the difference in results was statistically significant (p<0.0003) [Table 1]. The mean time recorded for BCSI and PCSI were 58.8218 and 62.4554 minutes, respectively, and the difference was found to be statistically significant (p<0.0003). (PCSI>BCSI).

A paired t-test comparing the area, total scores, and the time taken between both methods revealed a statistically significant difference of 0.0003 for the area, total scores (BCSI>PCSI), and time taken (BCSI<PCSI) [Table 1].

First follow-up: The mean area of involvement in both methods was found to be 12.852 and 10.049, respectively. The mean difference was statistically significant (p<0.0003) (BCSI>PCSI) [Table 1]. I.P comparison by PI and CI showed mean values of 3.8218 and 3.9109, respectively, and the paired t-test value was significant for both (P<.035 for both methods) [Table 1]. Total scores by PI and CI showed a mean of 48.6040 and 39.1188, respectively. Paired t-test value showed a statistically significant difference (p<.0003) [Table 1].

Second follow-up: First and second follow-ups were completed by 11 patients. The mean area of involvement of BCSI and PCSI was 15.361 and 12.091, respectively. Paired t-test value was statistically significant (p<.0003) [Table 1]. IP comparison by PI and CI had means 3.8182 and 3.4545, respectively. The paired t-test value was significant (p<0.0347) [Table 1]. BCSI and PCSI total scores showed a mean of 58.8182 and 48.3636, respectively. Paired t-test value was statistically significant (p<0.0057) [Table 1].

Comparison of BCSI: A decreasing trend in the area, IP, and the overall BCSI score between each visit was noted. The paired t–test values for the area, IP, and total scores between the baseline visit and 1st follow-up were found to be statistically significant (p<0.0003). However, the paired t-test values for the area of involvement (p < 0.247), I.P (p<0.114), and total scores (p<0.157) between follow-ups 1 and 2, based on 11 patients, were statistically non-significant [Table 2].

Comparison of PCSI: A decreasing trend in the area, IP, and the overall PCSI score. Paired t-test values showed statistically significant differences (p<0.0003) for all parameters except the IP (p<0.177) and total scores (p<0.0051) between follow-ups 1 and 2, based on 11 patients [Table 2].

Discussion

Our study validated and compared BCSI with the PCSI on the following counts: 1) A decrease in the mean PCSI as well as BCSI scores with treatment in each follow-up with a similar pattern of reduction observed, 2) Paired sample t-test showed that BCSI score is sensitive to change over time with treatment similar to PCSI, 3) The area and total scores obtained by the PI and CI during each visit from baseline to follow-up showed a positive correlation on scatterplot data [Figure 5].

Close

Figure 5:

Scattergram plots of PCSI vs BCSI scores across initial visit and follow-ups in area and total scores. (Note: Blue lines indicate the line of best fit. Red lines were drawn at 45°, dividing the graph into equal triangles to illustrate the degree of deviation between the scores).

Export to PPT

Intraclass correlation coefficients for inter-observer variability was assessed by the same research group and found good agreement.⁹

There was a statistically significant (p<.001) decrease in the total scores between baseline and 1st follow-up and between baseline and 2nd follow-up in both methods. However, the total scores between the 1st and 2nd follow-ups visits in 11 patients in both methods were statistically non-significant (p>0.05). This may be attributed to the slow response to treatment in melasma.

In the second follow-up, only 11/120 patients were available for follow up. The inclusion criteria allowed patients who had completed at least 1 follow-up. We encouraged for a second follow-up (after 6 weeks). Patients reported noticeable improvement in pigmentation after six weeks of treatment. This could be one of the reasons for fewer patients completing second follow-ups.

The total area calculated using the BCSI method showed a decline from baseline to the 1st follow-up visit and between baseline and the 2nd follow-up, which was statistically significant (p<0.05). However, the decrease in area between the 1st and 2nd follow-ups was statistically insignificant (p>0.05). On the other hand, the total area calculated using the PCSI method showed a statistically significant decline in the values between baseline, first, and second follow-ups (p<0.05). A speck of pigmentation is counted as 1 in BCSI; however, only the cross-over point is considered in PCSI. This principle enables BCSI to document scattered pigmentation, which is a limitation of PCSI. Therefore, the number of boxes and BCSI scores are significantly higher than the number of cross-over points and PCSI scores.

For standardisation, IP was assigned a score of 5 for all patients during the baseline visit in both methods. There was significant regression in IP between baseline and first follow-up visits and between baseline and second follow-up visits in both methods (p<0.001). However, the regression between the 1^st and 2^nd follow-up visits was not statistically significant (p>0.05) in both the methods because, though we compared 1st and 2nd follow-up images with baseline photographs, there could be minimal inter-observer variations while assessing the IP since it is subjective.

The mean time recorded by the PCSI method was higher than the BCSI method, which was statistically significant (p<0.001). This is due to the double tracing involved in the PCSI, leading to tracing errors.^11,12 The salient differences between BCSI and PCSI have been summarised in Table 3.

Limitations

Grid placement over bony prominences, facial hair, and partial glare were the factors that interfered with the assessment.

Conclusion

Capture of images in BCSI overcomes the limitations of double tracing in PCSI. It is easy and rapid. BCSI has a pragmatic value in tele-dermatology practice and is therefore recommended for use.