Multimodal imaging risk factors predictive of small choroidal melanocytic lesion growth to melanoma: An educational study and pictorial guide

Abstract

Background

Risk factors for small choroidal melanocytic lesion growth to melanoma have been redefined using multimodal imaging. We explored provider ability to recognize risk factors for small choroidal melanocytic lesion growth to melanoma before and after image-based education and with and without multimodal imaging.

Methods

Providers were invited to participate in a survey assessing ability to identify risk factors for small choroidal melanocytic lesion growth to melanoma using either fundus imaging or multimodal imaging. Risk factors included thickness >2 mm on ultrasonography, subretinal fluid on optical coherence tomography, presence of orange pigment by autofluorescence, acoustic hollowness by ultrasonography, and diameter >5 mm by fundus imaging. Performance was assessed before and after reviewing an educational PowerPoint providing pictorial examples of risk factors. Comparison between groups was conducted using two-tailed Fisher’s exact test.

Results

Thirty and 26 providers completed the pre-education and post-education assessments, respectively. Post-education participants were more accurate within ±1 risk factor for lesions with zero risk factors (77% vs. 100%, p = 0.01) or two risk factors (79% vs. 91%, p = 0.03). Following education, participants presented with multimodal imaging more often correctly identified lesions with four (12% vs. 42%, p = 0.03) or five (4% vs. 39%, p = 0.004) risk factors, demonstrated lower mean level of concern for lesions with zero risk factors (2.0 vs. 1.4, p < 0.001), and expressed higher level of concern for lesions with 5 risk factors (2.4 vs. 3.6, p < 0.001).

Conclusion

Use of multimodal imaging may be more beneficial than education itself to improve accuracy of risk factor identification for small choroidal melanocytic lesions.

Introduction

Choroidal nevus presents as a deep, pigmented lesion found in up to 4.7% of adults in the United States [1]. Transformation of choroidal nevus to malignant melanoma is rare and differentiating a benign small choroidal melanocytic lesion from small choroidal melanoma can be challenging. This differentiation is important as choroidal melanoma poses a profound risk to morbidity and mortality, with up to 50% of cases metastasizing to distant organs [2]. Once metastasized, median survival time for extrahepatic metastases is 19–28 months, and median survival time for hepatic metastases is 6 months [3, 4]. Early detection of choroidal melanoma, when the lesion is small, is the best defence for preventing metastatic disease.

Multimodal imaging features can be used to identify risk factors for choroidal nevus transformation to melanoma and can be remembered by the mnemonic “To Find Small Ocular Melanoma Doing IMaging” (TFSOM-DIM) [5–7]. The risk factors represented by this mnemonic are as follows: lesion thickness >2 mm, presence of subretinal fluid, vision loss to 20/50 or worse by Snellen acuity (symptoms), presence of orange pigment, melanoma acoustic hollowness, and lesion diameter >5 mm. Risk of transformation to melanoma over a 5 year period has been shown to increase with each additional risk factor [5].

For small choroidal melanocytic lesions at high-risk for growth to melanoma, referral to a subspecialist is indicated for detection and treatment of melanoma at the earliest stage. Therefore, providers should be aware of and be able to identify risk factors for small choroidal melanocytic lesion growth to melanoma to make appropriate referral decisions. However, DeSimone et al., in a recent survey of 60 ophthalmologists, found that only 48% of respondents were using the TFSOM-DIM criteria for small choroidal melanocytic lesion screening [6]. This could be related to a lack of provider trainer on these new criteria.

Education on melanoma risk factors has been shown to increase provider ability to recognize probable melanoma [8], and recent literature suggests that image-based education improves provider ability to apply imaging features to risk stratify melanocytic lesions [9–12]. However, previous studies have not explored the impact of TFSOM-DIM education or the importance of multimodal imaging compared with fundus photography alone for risk stratification of small choroidal melanocytic lesions. Herein, we explore provider ability to recognize risk factors for small choroidal melanocytic lesion growth to melanoma using TFSOM-DIM criteria before and after image-based education and with and without multimodal imaging.

Materials and Methods

This study was determined to be exempt by the Institutional Review Board at The Mayo Clinic in Rochester, MN. All images used in this study were anonymized prior to utilization.

Participants

Optometrists, ophthalmologists, ophthalmology residents, and fellows within the Mayo Clinic Health Systems were invited to participate in a survey assessing ability to identify risk factors for small choroidal melanocytic lesion growth to melanoma with and without multimodal imaging (Fig. 1). Assessment was conducted using a secure RedCap survey that could only be accessed on Mayo Clinic servers [13]. Survey invitations contained information regarding the risks and benefits of participation in this study, and all surveys explicitly asked for participant consent before they continued with their assessment. Participants were asked to provide non-identifiable usernames to match pre- and post-education survey responses. Current practice setting (residency programme, fellowship programme, academic centre, or private practice), postgraduate year (1–6 or greater, not applicable), and subspecialty training (medical or surgical retina, ocular oncology, not applicable) were recorded for each participant. Responses were anonymized and stored on RedCap under a secure Mayo Clinic server.

Fig. 1. Methodology in survey content and distribution of ophthalmology provider cohort.

Section 1 of each survey consisted of questions related to cases of small choroidal melanocytic lesions, and section 2 consisted of questions related to cases of lesion growth to melanoma or presumed melanoma on initial presentation.

Educational materials

After a pre-education survey, participants were sent an educational PowerPoint providing pictorial examples of TFSOM-DIM risk factors for small choroidal melanocytic lesion growth to melanoma, as well as additional features of retinal pigment epithelium trough (RPE trough) and documented growth (Fig. 2). Lesion thickness of >2 mm was shown using B-scan ultrasonography images. Subretinal fluid was demonstrated using optical coherence tomography. Symptoms, defined as visual acuity loss to 20/50 or worse measured by a Snellen acuity chart, were not assessed. Orange pigment accumulation was shown on colour fundus and fundus autofluorescence imaging. Acoustic hollowness and solidness were shown using B-scan ultrasonography with accompanying A-scan demonstrating low or high internal reflectivity. Lesion diameter of >5 mm was illustrated using fundus photographs. RPE trough was discussed as an indicator of chronic subretinal fluid, with examples on colour fundus and fundus autofluorescence imaging. Lesion growth was highlighted as a feature for malignant transformation independent of TFSOM-DIM risk factors, with fundus photograph, autofluorescence, OCT, and B-scan images shown as examples. Participants were given one month to review these educational materials between initial and follow-up surveys.

Fig. 2. Representative images of risk factors for small choroidal melanocytic lesion growth to melanoma.

Examples of images used to demonstrate TFSOM-DIM imaging features including: A lesion thickness >2 mm using B-scan ultrasonography, B subretinal fluid using optical coherence tomography, orange pigment accumulation using C colour fundus and D fundus autofluorescence imaging, melanoma acoustic hollowness using E B-scan and F A-scan ultrasonography, G lesion diameter >5 mm using fundus photographs, H small choroidal melanocytic lesion with I growth to melanoma over 5 months, and retinal pigment epithelium trough using J colour fundus and K autofluorescence imaging.

Risk factor survey assessment

The pre-education survey was broken into two parts: the first involving 10 cases of small choroidal melanocytic lesion and the second involving 3 cases: 2 cases of small choroidal melanocytic lesion with documented growth before-and-after 12-month follow-up, and one of choroidal melanoma. These cases were selected for inclusion from the Ocular Oncology Service at the Mayo Clinic in Rochester, MN, and all images in educational materials were reviewed for risk factors by ocular oncologist consensus agreement. In the first section, participants viewed a single funduscopic image of a small choroidal melanocytic lesion and were asked their level of concern (not at all concerned, a little concerned, moderately concerned, or highly concerned) and how many risk factors for growth to melanoma were present (0–5). Participants were then asked the same questions when presented with a multimodal imaging panel of the same lesions, consisting of colour fundus photograph, autofluorescence, OCT, and A- and B-scan ultrasonography.

In the second part of the pre-education survey, participants were presented with colour fundus images of small choroidal melanocytic lesion at initial presentation and at 12-month follow-up, or with a single image of a lesion diagnosed as melanoma at initial presentation. They were then asked their level of concern and when the patient should see a specialist (within the next 2 months, within 4–6 months, within the next 12 months, or does not need to see a specialist). Participants were asked the same questions when presented with a multimodal imaging panel of the same lesions at initial presentation and 12-month follow-up for lesions with growth or at initial presentation only for lesions initially diagnosed as melanoma.

After one month with the educational materials, participants were invited to complete a follow-up survey. Questions were identical to the previous survey. However, different cases were used in the pre- and post-education surveys. The post-education survey contained 10 cases of fundus and multi-modal imaging of small choroidal melanocytic lesion in the first section, with the second section consisting of one case of small choroidal melanocytic lesion with documented growth to melanoma before-and-after a 12-month follow-up, and one case of choroidal melanoma.

As part of a sub-analysis investigating which TFSOM-DIM risk factors were most often missed, participants were invited to complete a second post-education survey 13 months after the initial survey. This survey featured the same 10 image-based cases as section 1 of the first post-education survey, however, participants were prompted to respond with which TFSOM-DIM risk factors (tumour thickness >2 mm, subretinal fluid, presence of orange pigment, acoustic hollowness, tumour diameter >5 mm, or none of the above) were present on both fundus and multi-modal imaging.

Statistical Analyses

Accuracy identifying TFSOM-DIM risk factors was compared to the predetermined “correct” answers as defined by ocular oncologist consensus agreement. A percentage of correct responses among all participants was provided for lesions with each number of risk factors. Level of concern was expressed as a mean (median, range) among all participants. Referral timeframes were represented as percentage frequency of each response.

Comparison between groups was conducted using two-tailed Fisher’s exact test. Primary outcomes compared: pre- and post-education accuracy in number of TFSOM-DIM risk factors for each case using both single fundus and multimodal imaging; differences in mean level of concern; and differences in respondent referral timeframe recommendations. Sub-analysis compared responses with a single fundus image to those with multimodal imaging presented on the pre-education survey and again on the post-education survey. All statistical tests were performed using the latest version of RStudio. Statistical significance was defined as p < 0.05.

Results

Of the 121 ophthalmology providers who received the initial survey, 41 participated and 30 completed the entire image-based assessment. Of those who completed the assessment, 4 (13%) were in residency, 4 (13%) were in fellowship, 16 (60%) worked in an academic centre, and 4 (13%) worked in private practice. Of those 8 participants currently in a post-graduate programme, 2 (7%) were in their first year, 2 (7%) in their third year, and 4 (13%) in their fifth year. There were 6 (20%) respondents who trained in medical/surgical retina, and 24 (80%) indicated no retina or ocular oncology subspecialty training. Of the 41 providers who received the follow-up survey after educational intervention, 30 participated and 26 completed the assessment. Of those who completed the assessment, 4 (15%) were in residency, 3 (12%) were in fellowship, 18 (69%) worked in an academic centre, and 1 (4%) worked in private practice. Of those 7 currently in a post-graduate programme, 1 (4%) was in their first year, 1 (4%) in their second year, 2 (8%) in their third year, 3 (12%) in their fifth year, and 1 (4%) in year six or greater. There were 5 (19%) respondents who trained in medical/surgical retina, and 21 (80%) indicated no retina or ocular oncology subspecialty training. Among the 110 ophthalmology providers who received the sub-analysis survey, 14 participated and 10 completed the entire image-based assessment. Of those who completed the assessment, 4 (40%) were in residency, 1 (10%) was in fellowship, and 5 (50%) worked in an academic centre. Of those 5 participants currently in a post-graduate programme, 2 (40%) were in their second year, 1 (10%) in their third year, 1 (10%) in their fourth year, and 1 (10%) in their fifth year. There were 2 (20%) respondents who trained in medical/surgical retina, and 8 (80%) indicated no retina or ocular oncology subspecialty training.

A pre- vs. post-education comparison (Table 1) revealed that, when presented with a single fundus image, post-education participants more frequently responded incorrectly by identifying at least one “risk factor” in cases with zero risk factors (70% vs. 8%; p < 0.001). A similar comparison revealed that, when presented with multimodal imaging, post-education participants more frequently responded incorrectly by identifying at least one “risk factor” in cases with zero risk factors (67% vs. 31%, p = 0.02) and were less likely to correctly identify number of risk factors in lesions with one risk factor (48% vs. 23%, p = 0.03). When presented with single fundus imaging, post-education participants had greater mean level of concern for cases with zero (1.3 vs. 2.1, p < 0.001) and three (3.0 vs. 3.2, p = 0.04) risk factors. No significant change in level of concern between cases was noted pre- versus post-education when presented with multimodal imaging. When presented with multimodal imaging of choroidal melanoma, post-education participants reported a higher level of concern (3.5 vs. 3.9, p = 0.002) and more frequently recommended subspecialty referral within 2 months (77% vs. 100%, p = 0.03). A similar comparison revealed no significant change in level of concern or recommended time to subspecialty referral when presented with either single fundus or multimodal imaging of small choroidal melanocytic lesions with documented growth to melanoma before-and-after a 12-month follow-up. Similar results were reflected after removing responses from those with medical/surgical retina training (Supplemental Table 1). When giving respondents a ± 1 margin of error in accuracy of assessing TFSOM-DIM risk factors (Supplemental Table 2), a comparison (pre- vs. post-education) revealed that post-education participants were more frequently accurate within ±1 risk factor for lesions with zero risk factors (77% vs. 100%, p = 0.01) or two risk factors (79% vs. 91%, p = 0.03).

Table 1.

Participant responses using single fundus and multimodal imaging before and after educational intervention.

	Pre-education (N = 30)a [n/N (%)]	Post-education (N = 26)a [n/N (%)]	p-values
Lesion with single fundus image
Number of risk factors	Correct responses (%)	Correct responses (%)
0	21/30 (70)	2/26 (8)	<0.001
1	19/60 (32)	8/26 (31)	0.99
2	39/90 (43)	34/78 (44)	0.99
3	32/90 (36)	25/78 (32)	0.74
4	3/30 (10)	3/26 (12)	0.99
5	N/A	1 /26 (4)	N/A
Level of concern per number of risk factors	[Mean (median, range)b]	[Mean (median, range)b]
0	1.3 (1, 1–2)	2.0 (2, 1–3)	<0.001
1	2.4 (2, 1–4)	2.4 (2, 1–4)	0.56
2	3.0 (3, 1–4)	2.7 (3, 1–4)	0.08
3	2.97 (3, 1–4)	3.2 (3, 2–4)	0.04
4	3.3 (3, 2–4)	3.2 (3, 1–4)	0.75
5	N/A	2.4 (2.5, 1–4)	N/A
Lesion with multimodal imaging
Number of risk factors	Correct responses (%)	Correct responses (%)
0	20/30 (67)	8/26 (31)	0.02
1	29/60 (48)	6/26 (23)	0.03
2	35/90 (39)	20/78 (26)	0.07
3	36/90 (40)	29/78 (37)	0.75
4	9/30 (30)	11/26 (42)	0.41
5	N/A	10/26 (39)	N/A
Level of concern per number of risk factors	[Mean (median, range)b]	[Mean (median, range)b]
0	1.3 (1, 1–3)	1.4 (1, 1–2)	0.20
1	2.03 (2, 1–3)	2.0 (2, 1–3)	0.72
2	2.81 (3, 1–4)	2.8 (3, 1–4)	0.77
3	3.27 (3, 1–4)	3.3 (3, 1–4)	0.22
4	3.63 (4, 2–4)	3.4 (3.5, 2–4)	0.36
5	N/A	3.6 (4, 2–4)	N/A
Documented growth to melanoma with single fundus image
Level of concern [Mean (median, range)b]	3.3 (4, 1–4)	3.5 (4, 2–4)	0.42
Recommend time to referral (months)	Responses [n/N (%)]	Responses [n/N (%)]
Within 2 months	45/60 (75)	23/26 (88)	0.65
4–6 months	10/60 (17)	3/26 (12)
12 months	4/60 (7)	0/26 (0)
Referral not indicated	1/60 (2)	0/26 (0)
Documented growth to melanoma with multimodal imaging
Level of concern [Mean (median, range)b]	3.4 (4, 2–4)	3.7 (4, 2–4)	0.31
Recommend time to referral (months)	Responses [n/N (%)]	Responses [n/N (%)]
Within 2 months	45/60 (75)	25/26 (96)	0.12
4–6 months	12/60 (20)	1/26 (4)
12 months	2/60 (3)	0/26 (0)
Referral not indicated	1/60 (2)	0/26 (0)
Melanoma with single fundus image
Level of concern [Mean (median, range)b]	3.1 (3, 2–4)	2.7 (3, 1–4)	0.08
Recommend time to referral (months)	Responses [n/N (%)]	Responses [n/N (%)]
Within 2 months	11/30 (37)	7/26 (27)	0.57
4–6 months	16/30 (53)	14/26 (54)
12 months	3/30 (10)	5/26 (19)
Melanoma with multimodal imaging
Level of concern [Mean (median, range)b]	3.5 (4, 2–4)	3.9 (4, 3–4)	0.002
Recommend time to referral (months)	Responses [n/N (%)]	Responses [n/N (%)]
Within 2 months	23/30 (77)	26/26 (100)	0.03
4–6 months	5/30 (17)	0/26 (0)
12 months	2/30 (7)	0/26 (0)

^aTotal denominator is given for each number of risk factors, as some pertained to more than one question in the survey.

^b1 = not at all concerned, 2 = a little concerned, 3 = moderately concerned, 4 = highly concerned.

Bold values indicate significant p-value.

A subanalysis comparison between survey responses with single fundus versus multimodal imaging before education (Table 2) revealed that, when presented with multimodal imaging, participants had a lower level of concern for cases with one risk factor (2.4 vs. 2.03, p = 0.002) and a higher level of concern for cases with three risk factors (2.97 vs. 3.27, p = 0.001). Following education, a comparison (single fundus vs. multimodal imaging) revealed that, when presented with multimodal imaging, participants were less likely to accurately identify number of risk factors in lesions with two risk factors (44% vs. 26%, p = 0.03) but more likely to correctly identify lesions with four (12% vs. 42%, p = 0.03) or five (4% vs. 39%, p = 0.004) risk factors. A comparison (single fundus vs. multimodal imaging) following education revealed that multimodal imaging presentation was associated with lower mean level of concern when assessing lesions with zero risk factors (2.0 vs. 1.4, p < 0.001) and higher level of concern when assessing lesions with 5 risk factors (2.4 vs. 3.6, p < 0.001).

Table 2.

Comparison of participant responses when presented with single fundus versus multimodal imaging.

	Single Fundus Image [n/N (%)]	Multimodal Imaging [n/N (%)]	p-values
Pre-Education
Number of risk factors	Correct responses (%)	Correct responses (%)
0	21/30 (70)	20/30 (67)	1.00
1	19/60 (32)	29/60 (48)	0.09
2	39/90 (43)	35/90 (39)	0.65
3	32/90 (36)	36/90 (40)	0.65
4	3/30 (10)	9/30 (30)	0.10
Number of risk factors	Level of Concern mean (median, range)a	Level of Concern mean (median, range)a
0	1.3 (1, 1–2)	1.3 (1, 1–3)	0.77
1	2.4 (2, 1–4)	2.03 (2, 1–3)	0.002
2	3.0 (3, 1–4)	2.81 (3, 1–4)	0.27
3	2.97 (3, 1–4)	3.27 (3, 1–4)	0.001
4	3.3 (3, 2–4)	3.63 (4, 2–4)	0.08
Post-Education
Number of risk factors	Correct responses (%)	Correct responses (%)
0	2/26 (8)	8/26 (31)	0.08
1	8/26 (31)	6/26 (23)	0.76
2	34/78 (44)	20/78 (26)	0.03
3	25/78 (32)	29/78 (37)	0.61
4	3/26 (12)	11/26 (42)	0.03
5	1/26 (4)	10/26 (39)	0.004
Number of risk factors	Level of Concern mean (median, range)a	Level of Concern mean (median, range)a
0	2.0 (2, 1–3)	1.4 (1, 1–2)	<0.001
1	2.4 (2, 1–4)	2.0 (2, 1–3)	0.20
2	2.7 (3, 1–4)	2.8 (3, 1–4)	0.19
3	3.2 (3, 2–4)	3.3 (3, 1–4)	0.28
4	3.2 (3, 1–4)	3.4 (3.5, 2–4)	0.79
5	2.4 (2.5, 1–4)	3.6 (4, 2–4)	<0.001

^a1 = not at all concerned, 2 = a little concerned, 3 = moderately concerned, 4 = highly concerned

Bold values indicate significant p-value.

Subanalysis for risk factor estimation (Table 3) revealed that, before education, participants often overestimated number of risk factors in low-risk lesions using both single fundus (30% overestimated for zero risk factors, 67% for one risk factor) and multimodal imaging (33% overestimated for zero, 43% for one risk factor). Similar findings were noted post-education using both single fundus (92% overestimated for zero, 65% for 1 risk factor) and multimodal imaging (69% overestimated for zero, 62% for one risk factor). Before education, participants often underestimated number of risk factors for high-risk lesions when given only single fundus images (54% underestimated for three risk factors, 87% for four risk factors), but accuracy improved with multimodal imaging (28% underestimated for three, 37% for four risk factors). Similar findings were observed following education when given only single fundus images (50% underestimated for three, 85% for four, and 96% for five risk factors), but accuracy improved for high-risk lesions with multimodal imaging (23% underestimated for three, 46% for four, and 61% for five risk factors).

Table 3.

Over-estimation and under-estimation of the number of risk factors using single fundus and multimodal imaging before and after educational intervention.

	Single Fundus			Multimodal
	Correct	Over-Estimated	Under-Estimated	Correct	Over-Estimated	Under-Estimated
Pre-Education
Number of risk factors	Responses [n/N (%)]	Responses [n/N (%)]	Responses [n/N (%)]	Responses [n/N (%)]	Responses [n/N (%)]	Responses [n/N (%)]
0	21/30 (70)	9/30 (30)	n/a	20/30 (67)	10/30 (33)	n/a
1	19/60 (32)	40/60 (67)	1/60 (2)	29/60 (48)	26/60 (43)	5/60 (8)
2	39/90 (43)	37/90 (41)	14/90 (16)	35/90 (39)	38/90 (42)	17/90 (19)
3	32/90 (36)	9/90 (10)	49/90 (54)	36/90 (40)	29/90 (32)	25/90 (28)
4	3/30 (10)	3/30 (10)	26/30 (87)	9/30 (30)	10/30 (33)	11/30 (37)
Post-Education
0	2/26 (8)	24/26 (92)	n/a	8/26 (31)	18/26 (69)	n/a
1	8/26 (31)	17/26 (65)	1/26 (4)	6/26 (23)	16/26 (62)	4/26 (15)
2	34/78 (44)	18/78 (23)	26/78 (33)	20/78 (26)	44/78 (56)	14/78 (18)
3	25/78 (32)	14/78 (18)	39/78 (50)	29/78 (37)	31/78 (40)	18/78 (23)
4	3/26 (12)	1/26 (4)	22/26 (85)	11/26 (42)	3/26 (12)	12/26 (46)
5	1 /26 (4)	n/a	25/26 (96)	10/26 (39)	n/a	16/26 (61)

Subanalysis for which risk factors were most often missed is presented in Supplemental Table 3. Overall, using single fundus images, each risk factor was correctly identified in less than two thirds of cases, with worst performance for subretinal fluid, which was correctly identified only 57% of the time. When using multimodal imaging, participants had the most difficulty correctly identifying orange pigment, which was only correctly identified 70% of the time. A comparison (single fundus vs. multimodal imaging) revealed that, when compared to single fundus imaging, participants viewing multimodal images more accurately identified tumour thickness >2 mm (60% vs. 88%, p < 0.001), subretinal fluid (57% vs. 75%, p = 0.011), acoustic hollowness (61% vs. 78%, p = 0.014), and tumour diameter >5 mm (64% vs. 79%, p = 0.018).

Discussion

The risk of transformation of small choroidal melanocytic lesions to choroidal melanoma can be estimated based on the number of risk factors present. Multimodal imaging features can be used to precisely estimate risk for any given lesion, which can guide decisions for monitoring frequency or referral. Remembered by the mnemonic TFSOM-DIM, multimodal imaging risk factors for choroidal nevus transformation to melanoma include thickness >2 mm, subretinal fluid, symptoms of vision loss, orange pigment, melanoma acoustic hollowness, and diameter >5 mm [7, 14]. Herein, we explored the impact of image-based education on provider ability to correctly apply TFSOM-DIM criteria for small choroidal melanocytic lesion evaluation and secondarily assessed the importance of multimodal versus single fundus imaging for risk factor identification.

Image-based education has proven effective in increasing participant ability to identify high-risk imaging findings among the general population and medical providers. Maganty et al., in a study of 60 dermatology patients, found participants receiving an online image-based education intervention demonstrated superior sensitivity for detecting features of melanoma compared to those who received only an educational pamphlet [9]. Similarly, Goulart et al., in a study of 130 second-year medical students, demonstrated improved proficiency in classifying skin lesions as benign or malignant following an image-based lecture on melanocytic risk factors [10]. Clinical images combined with a systematic method of evaluation have been effective in distinguishing low risk small choroidal melanocytic lesions from melanoma. Regarding specific applications to ocular oncology, Roefels et al. and Salehi et al. demonstrated that MOLES criteria had a sensitivity of 99.8% and specificity of 97%, respectively, for identification choroidal nevi with high-risk of transformation to melanoma [11, 12]. Flanagan et al., in a study of 39 optometrists, found that optometrists could correctly identify probable melanomas from choroidal nevi with 95.8% sensitivity and 64.1% specificity using the MOLES system along with clinical images, which included colour fundus photographs, fundus autofluorescence, OCT, and B-scan ultrasound images [8]. Considering the effectiveness of image-based education in improving provider ability to reliably apply high-risk imaging features to small choroidal melanocytic lesions, we sought to investigate if this method would remain successful using TFSOM-DIM criteria.

In the present study, participants more frequently identified at least one risk factor in cases with zero risk factors following education compared to pre-education participants regardless of presentation with a single fundus image or multimodal imaging. This result might indicate that the image-based education increased bias toward finding at least one risk factor. When given a ± 1 margin of error in accuracy for assessing risk factors, post-education participants were more frequently accurate within ±1 risk factor for lesions with zero or two risk factors compared to pre-education participants. Following education, participants were more likely to correctly identify number of risk factors in lesions with four or five risk factors using multimodal imaging, suggesting a greater awareness of key features in high-risk lesions. Regardless of education or image modality, the participants tended to overestimate the number of risk factors for lesions with one or no risk factors. This opposes the conjecture that errors in low-risk lesion assessment were solely due to post-education bias to identify more risk factors.

On the post-education survey, multimodal imaging (compared to a single fundus image) was associated with lower mean level of concern for lesions with zero risk factors and higher level of concern for lesions with 5 risk factors, suggesting the importance of multimodal imaging in appropriate triage of low- versus high-risk lesions. A combination of education and use of multimodal imaging might increase detection and appropriate follow-up for high-risk small choroidal melanocytic lesions while avoiding over-referral of low-risk lesions.

When cases had three or more risk factors, both pre- and post-education participants had higher accuracy using multimodal imaging compared to a single fundus image. Despite these findings, this study showed overall poor ability of participants to accurately determine the precise number of risk factors for a given lesion by TFSOM-DIM criteria. This could be due to the difficulty utilizing this system, as TFSOM-DIM was intended for use by subspecialty experts, whereas the MOLES system was intended for non-experts to determine the likelihood of malignancy [15]. Additionally, our subanalysis revealed that, when utilizing single fundus or multimodal imaging, presence of subretinal fluid and orange pigment were less frequently identified correctly compared to other TFSOM-DIM risk factors, suggesting a potential area for educators to specifically target when teaching risk factor identification for uveal melanoma.

Limitations of this study included the small sample size with variable education stages and practice settings of participants. A larger sample size would better confirm results and improve generalizability. Due to the timing of the study, participants in training had progressed to the next post-graduate year between the pre- and post-education surveys. Additional education as training progresses could account for some differences in survey results, but the short one-month time-period between surveys was used to limit this effect. Although the study intended to collect participant-selected “usernames” to allow paired comparison of pre- and post-education survey results while maintaining survey anonymity, several participants forgot their usernames prior to taking the follow-up survey, which precluded paired comparisons. The pre- and post- education surveys asked participants to identify the number of risk factors and not which risk factors were present. As a result, we could not determine which specific risk factors were most frequently correctly recognized and which were more often missed.

Early detection of choroidal melanoma is the best defence against metastatic disease. Therefore, it is critical that physicians recognize risk factors for lesion growth to melanoma. Image-based education on TFSOM-DIM risk factors combined with multimodal imaging improves provider recognition of high-risk melanocytic lesions. However, this education might also increase provider concern for presence of risk factors in low-risk lesions. Overall, use of multimodal imaging might be more beneficial than education alone to improve accuracy of risk factor identification. Further research is needed to better understand whether and to what extent image-based educational material and multimodal imaging contribute to the correct identification of risk factors for small choroidal melanocytic lesion growth to melanoma.

Summary

What was known before

• Identification of risk factors for small choroidal melanocytic lesion malignant transformation is critical for early detection of small melanoma.

• Identifying these risk factors can be challenging for many providers due to a lack of training and low frequency with which choroidal melanoma is seen in a comprehensive eye care practice.

What this study adds

• In this study, image-based education on risk factors for small choroidal melanocytic lesion growth to melanoma improved provider ability to identify high-risk lesions, while also raising levels of concern for all melanocytic lesions, including those deemed low-risk.

• The addition of multimodal imaging to single fundus photographs appeared to be more beneficial than education itself to improve accuracy of risk factor identification for small choroidal melanocytic lesion.

• These findings provide inconclusive evidence on the importance of image-based education for small choroidal melanocytic lesion risk factors.

• However, study results underscore the importance of multimodal imaging for accurate identification of high-risk small choroidal melanocytic lesions and small melanoma.

Author contributions

LAD and AAT were responsible for identifying the knowledge gap investigated in this manuscript, designing the survey questions, choosing representative images for both survey and review materials, determining an appropriate population to recruit from, and providing input for both data analysis and writing of this manuscript. RAC was responsible for design of pre- and post-education surveys, recruitment of participants, data analysis, drawing conclusions from the analysis, and writing of the manuscript. TYC was responsible for drawing conclusions from data analysis and contributed meaningfully to manuscript writing. OMH and AM provided feedback and impactful edits during the revision process.

Funding

This publication was made possible through the support of the Leonard and Mary Lou Hoeft Career Development Award Fund in Ophthalmology Research, Grant Number P30 CA015083 from the National Cancer Institute, and CTSA Grant Number KL2 TR002379 from the National Center for Advancing Translational Science (NCATS). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.

Data availability

More detailed data are not publicly available due to ethical reasons. Additional data requests can be submitted to the corresponding author. Lauren A. Dalvin, M.D. has had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Competing interests

This manuscript was submitted as a poster presentation for The Association of Research in Vision and Ophthalmology 2023 Annual Meeting. This manuscript has not been published elsewhere. The authors declare that they have no conflict of interest.

Supplementary information

The online version contains supplementary material available at 10.1038/s41433-023-02782-8.