Skip to main content
Open Access Publications from the University of California

Dermatology Online Journal

Dermatology Online Journal bannerUC Davis

Clinical pathway for melanoma detection using comprehensive cutaneous analysis with Melanoscan®

Main Content

Clinical pathway for melanoma detection using comprehensive cutaneous analysis with Melanoscan®
Rhett J Drugge MD, Chi Nguyen, Luciana Gliga, Elizabeth D Drugge PhD
Dermatology Online Journal 16 (8): 1

Dermatology Imaging Center, Stamford, Connecticut


The usefulness of a comprehensive cutaneous photography system (Melanoscan®) was tested using the following parameters: 1) decision to screen pathway, 2) clinical pathway, 3) clinical outcome, and 4) patient acceptance. The results indicate that 55 percent of those with criteria for scanning were reimbursed by insurance (AMA CTP category 1 code status 96904 for total body photography). In this model of whole body scanning, the ratio of time demand on physicians, patients, and technicians is 1:8:12. In 53 patients, 394 lesions of concern were identified. Of these lesions, 48 (12.31%) were scars, 306 (78.46%) were changed, and 36 (9.23%) were new. The decision to biopsy was made for 18 of the 394 lesions analyzed in the follow-up studies. Sensitivity of the process in determining malignant lesions is 75.00 percent and specificity is 73.70 percent. Preliminary results suggest that change detection analysis reduces the number of biopsies and improves diagnostic accuracy. Assessment of survey results revealed a high degree of patient satisfaction with ease of following Melanoscan directions (81.25%), as well as overall satisfaction with the process (73.44%). These results suggest that whole body screening using the Melanoscan provides a device in which accuracy of lesion tracking, patient confidence in lesion documentation, and clinician time are optimized.

I. Background

The continued increases in melanoma incidence [1], the poor prognosis of advanced melanoma, the economic burden of advanced cases [2, 3], and the lack of a definitive non-invasive melanoma detection technique underscore the critical need for an effective melanoma screening system. Currently available melanoma detection modalities outlined and reviewed by Goodson and Grossman [4] have limited value when used individually; successful screening necessitates an integrated approach.

Studies of store and forward teledermatology in the 1990s demonstrated the power of both digital images and historical data in conveying clinical information to the consulting dermatologist [5]. Likewise, studies of whole body photography for melanoma screening have demonstrated the key benefit of detection at thinner Breslow depth [6]. However, in spite of the preponderance of evidence for clinical photography [7, 8, 9, 10], dermatology has been hampered by a lack of access to high quality, reproducible clinical photography, possibly because of earlier methods. Previous whole body photography methods for melanoma screening potentially add to physician examination time [11, 12]. Fortunately, in the last two decades we have witnessed advances in photography techniques and culminating in the application of a multi sensor photo booth to whole body photography.

Lesion change detection with whole body photography is a valuable predictor of melanoma [13, 14] and has been associated with improved outcomes (lower Breslow depth) [6, 8, 9]. It is possible that improved outcomes result from unproductive practices: long examination times, increased numbers of biopsies, and low patient acceptance. We sought to address this possibility by evaluating Melanoscan screening with serial whole body photography using the following parameters: 1) decision to screen pathway, 2) clinical pathway, 3) clinical outcome, and 4) patient acceptance.

II. Methods

Figure 1Figure 2
Figure 1. Melanoscan

Figure 2. Melanoscan (top view)

Figure 3
Figure 3. Melanoscan body poses

Scanner. A novel hybrid photo imaging and therapy device was placed in a phototherapy room in a private dermatology practice. The phototherapy booth was augmented by a 25 digital camera array (Figures 1 and 2). In addition, the booth was serviced by white LED lighting and detailed with forward and backward footprints. The computer-synchronized system captured the patients in three choreographed positions driven by a sequence of audiovisual cues (Figure 3).

Patient Evaluation. The decision pathway for scanning under the AMA CPT category I code status (96904) for total body photography outlines hard criteria: more than four dysplastic nevi, as assessed by clinical diagnosis by a pathologist or dermatologist, previous personal history of melanoma, or history of melanoma in a first degree family member. Other criteria (soft criteria) that were considered were numerous moles, more than fifty, multiple non-melanoma skin cancers (basal cell or squamous cell carcinomas), and immunosuppressive disorders (lymphomas, leukemias, infectious disease or immunosuppressive drugs). Insurance companies reimbursing under CPT code 96904 include: Health Net, Connecticare, PHCS, Aetna, Medicare, and Guardian. Insurance Companies not covering 96904 include: Cigna, Oxford, United Healthcare, Anthem BCBS (Blue Care, Century Preferred), and Empire.

Study Design and Analysis. The study group was randomly selected from scanned patients over the course of a three-month period. All patients included in the study were asked to fill out a consent form as well as a patient satisfaction survey and were managed identically to all other scanned patients in the practice. Captured images were extracted, encrypted, and reviewed for quality assurance. None of the 64 scans required repeated capture caused by flaws in image quality or other technical problems. There were two types of studies: baseline and follow-up. For a baseline study, the whole body scan was reviewed by a technician for quality assurance, followed by mapping of the images with the following: lesions of patient concern, prior surgical locations, and atypical appearing lesions. Dermoscopy images are obtained from these locations and annotated to the baseline map. Follow-up studies involve comparison of images with baseline images obtained at least one year apart. Data was entered and analyzed using Stata®, version 11.0, StataCorp, 2001, College Station, Texas.

Figure 4Figure 5
Figure 4. Visual representation of image comparison

Figure 5. Annotation dialog for lesion of concern

Software. A specialized software system was built to analyze the comprehensive cutaneous imagery. For a follow-up study, changed or new spots of concern were identified by trained technicians. The visualization tools used to enhance comparisons include scaling and color registration, time-lapse montage, and flicker techniques. In addition, images were linked to current dermoscopy and prior clinical information, such as surgical procedures and pathology reports (Figure 4). Lesions were annotated with clinical information, enhanced by three controlled vocabularies ICD 9, CPT, and DermLex [15] (Figure 5). These annotations combined with information drawn from the associated electronic medical record allow for indexing of the studies within the SNOMED (Systematized Nomenclature of Medicine) vocabulary.

Figure 6
Figure 6. Report showing images of lesions of concern from 01/03/208 and 05/11/2009 with dermoscopy images and clinical notes below

Patient Evaluation. Whole body scan review occurred in a clinical staff area. Trained technicians analyzed follow-up images to discover changes and captured dermoscopy images of new lesions or lesions of concern. Complete reports were presented to the dermatologist in the clinical staff area for review (Figure 6). Preliminary clinical management decisions were determined and communicated to the support staff. The physical examination was enriched by a stepwise approach that included listening to the patient’s concerns, confirming report findings, and discussing findings. The evaluation and management decisions were then presented to the patient for informed consent.

III. Results

Decision to Screen Pathway. Patients (n=64) were randomly selected from those patients scanned (n=312) from August 1, 2009 to December 1, 2009. The mean age of the group was 55.38 ± 16.48 years, 53.5 years, and ranging from 20 to 83 years. Of the 64 patients, 38 were male (59.38%) and 24 were female (40.63%). Of those patients with a prior history of melanoma 38.89 percent were reimbursed by insurance for scanning. Of those patients with a family history of melanoma, 33.34 percent were reimbursed by insurance. When additional hard and soft criteria were considered, the percent of patients reimbursed by insurance increased to 55.81 percent (Table 1).

Additional selection criteria for scanning were based on skin cancer risk assessment scores derived from a questionnaire [16, 17]. The 14 question, score range (1-33), risk assessment tool, tested in over 8,000 patients, indicates low risk (0-9), high risk (10-19), and very high risk (20-33). The overall risk of the group we tested (n=64) was 17.92 ± 5.87, median = 17, and range of 8 to 33. The distribution was as follows: Risk Score 0-9: 1 person (1.56%); Risk Score 10-19: 39 people (60.94%); Risk Score 20-33: 24 people (37.50%).

Documented criteria were scanned into the patients’ charts to facilitate future processing. Patients with documented hard criteria and insurance coverage were scanned directly. Patients with soft criteria and insurance coverage first had to see the dermatologist to qualify for a scan. Those interested in scanning without coverage, whether qualified or not, as well as qualified and covered patients who wished to be scanned more than once a year, were responsible for the cost of the scan and were scheduled directly.

Figure 7
Figure 7. Scanning Process

Patient Flow. Once a patient was scanned, images were extracted and analyzed as depicted in Figure 7. Analysis involved comparison of all 56 new images to baseline images. Lesions of concern that were new or changed and previous biopsy sites were then included in the report. The average number of lesions of concern in this study group was 7.3 ± 3.5 (n=53). Next, the patient was brought back into the clinic and dermoscopy images of the lesions of concern were taken and added to the report. The physician then added diagnoses on the report before entering the clinic. Diagnoses were then confirmed in person and the traditional skin exam was performed, with special attention to areas missed by the scan (especially the bottom of the feet and hair-bearing areas).

Figure 8
Figure 8. Time requirements for analysis of a patient with 6 lesions

Scan Timing. The time requirements of scanning have been broken down in terms of patient time, system time, technician time, and dermatologist time (Figure 8). Automated scanning that included 2 minutes of instruction and patient preparation averaged 3.2 ± 1.2 minutes (n=64). The actual time in the booth was 2 minutes 47 seconds directed by automatic instruction and time to extract and encrypt images was 7 minutes. The time for the technician to analyze the 56 images per person ranges from 15 to 25 minutes, depending on the patient history, number of lesions, and skin cancer risk assessment score. The time to take dermoscopy images of suspicious lesions was 30 seconds per lesion. The time to import the dermoscopy images into the Melanoscan and to prepare the report for the dermatologist was 10 seconds per lesion. Physician time to look at the report, using flicker analysis and to make a diagnosis was 10 seconds per lesion. Physician time to consult with the patient and confirm diagnosis was 20 seconds per lesion being considered for biopsy. Finally, the physician time to assess areas not covered by the Melanoscan and perform the entire skin exam including hair-bearing areas and bottom of feet was approximately 2 minutes.

Clinical outcome. Of the 64 patients included in the study, 11 were baseline (first time scans). The remaining 53 patients’ images were compared to previous images taken at various time intervals. The mean time between scans in this patient group is 2.15 ± 0.96 years. Of the 390 identified lesions of concern, 48 were scars tracked from previous biopsies that are kept under surveillance. Three hundred six were changed lesions and 36 were new lesions. The decision was made to biopsy 18 lesions. Of these, 12/18 (66.67%) had strict agreement between change-detection impression and pathological diagnosis (Table 2). In addition, 15/18 (83.34%) were in agreement based on benign or malignant change-detection impression vs. pathology diagnosis (Kappa value: 0.57 + 0.21, p< 0.05). Furthermore, 16/18 (89.90%) were matches between impression and diagnosis based on the dominant cell type (e.g., lichenoid and actinic keratosis would both be classified as keratinocytic lesions). Analysis of inter-rater correlation based on cell type of origin resulted in a Kappa value of 0.78 ± 0.23, p< 0.05.

Eleven of the patients were given baseline (first-time) scans and diagnoses were made based on clinical impression of which 13 were concordant pairs, malignant or benign, as identified by clinical impression and pathological diagnosis. Of these 11 patients, 5 lesions were biopsied based on complete physical exam by physician and image surveillance by technician. Although the sample size is too small for comparison, preliminary results suggest concordance between impression and pathology diagnosis using strict agreement of 3/5 (60.00%); using benign or malignant impression vs. diagnosis of 2/5 (40.00%); and using cell origin of 4/5 (80.00%) as shown in Table 2. Analysis of benign vs malignant change detection based on impression vs. pathology diagnosis resulted in a 100.00 percent sensitivity and 80.00 percent specificity. Similar analysis performed on the biopsies taken without change detection resulted in a 0 percent sensitivity and 50.00 percent specificity. When compared to non-sequential evaluations, the biopsy rate decreased from 45.00 percent (5/11 baselines) to 34 percent (18/53).

Figure 9
Figure 9. Patient satisfaction

Patient satisfaction survey. Patients were asked to evaluate the scanning process in terms of ease of following instructions and overall satisfaction with the process on a scale from 5 to 1, strongly agree to strongly disagree, Figure 9. In addition, patients were asked to indicate whether they would prefer to come back for the analysis at a later time or date (48.44%), to wait for the analysis (32.81%); (18.75%) were indifferent.

IV. Discussion

The Melanoscan model presents an innovative approach to evidence-based dermatological care. We report a high degree of patient satisfaction with a process that requires a greater proportion of technician to clinician time and presents the opportunity for patients to become more involved in their own health care. It is also interesting that over 60 percent of our patient group opted to pay for a scan regardless of need criteria, thus demonstrating patient understanding of the importance of documentation for quality health care. The scanned population was high risk: almost all (97%) scored as such by a validated skin cancer risk survey [16, 17] and 48.44 percent had a personal history of skin cancer.

Error Reductions. Tracking clinically significant lesions eliminates the inaccuracies associated with recall on the part of either the physician or patient [18]. Systematic refinements in this paradigm will lead to fewer unnecessary biopsies and overlooked cancers.

Patient Acceptance. Patient satisfaction with scanning is high. The potential sources of enthusiasm are as follows: the individualized selection process, the comprehensive machine environment, the secondary level of examination by dermoscopy, and finally, the highly prepared physician. The scanning and analysis processes can occur independently and apart from the dermatology visit; almost half of our patients would prefer to make a separate appointment with the physician despite the possibility of a second co-pay. The separation of services evolved in response to the increase in demand for patient imaging from outside dermatologists.

Image Classifiers. Finally, studies have demonstrated the application of lesion classifiers on dermoscopy and clinical images [19]. The accuracy of such classifiers should improve with increasing camera resolution and characterization of lesion change features. Facilitation of comprehensive cutaneous imagery sets the stage for computer vision image analysis-based strategies for the identification of melanoma. The characteristics of personal imagery can be used for risk assessment strategies such as global spot counting.

Three Image Positions. The reproducibility of image positioning requires audiovisual guides and able, compliant subjects. People without physical distortion due to age or disease can reproduce the study positioning with high fidelity based on a multimedia training module. Selection of patient positions is subject to data feedback. It consists of three positions carefully designed to promote reproducibility and maximal exposure.

Foot Compass. The bidirectional footprints, camera angles and horizontal lines representing camera altitudes are reproduced to allow technicians to create small sets of comparison images in the clinics. Our term for these is freehand images. These images are viewed with the Melanoscan analysis software for comparison to prior images taken from the same perspectives. The Melanoscan is not required for such clinical images.

Figure 10
Figure 10. CD interface

Privacy. Privacy is reinforced on various levels. The risks and benefits of the use of occluding clothing are described to the patient. Their use is at the discretion of the patient. At each imaging session, disposable undergarments are made available. Patients’ requests to use their own undergarments are also honored. Most patients have confidence in the privacy measures and remove all of their clothing and jewelry. Images for off site usage are printed on CDs with its own viewing software (Figure 10). On these CD images, a privacy cover is used for women’s breasts and the genitals in both sexes. Images are stored and backed up in a 128 bit encrypted proprietary format readable only by the Melanoscan analysis software.

Comparison to the “two-step method.” The current Melanoscan approach parallels the “two-step method” evaluated by the University of Barcelona [20]. In that method, standard hand-held camera views are obtained and compared to the patients in a side-by-side (montage) technique. In the Barcelona technique, dermoscopy images are captured when change is detected and annotated to the baseline images. The time required was 30-40 minutes of an experienced dermatologist. Our system differs from the Barcelona method in many details described in the methods section.

Clinical Flow. An unanticipated benefit of the dermatologist’s report review is improved clinical flow. This occurs when the physician identifies a lesion on the report that may require a surgical procedure. Preparations for the anticipated procedure may be made before the physician enters the room for the physical examination. Reported lesions are marked at the time of dermoscopic image capture with washable ink, validating the location of an ensuing procedure. Anticipated procedures are either performed or aborted at the will of the dermatologist and the consenting patient. Notification of the clinical support staff of anticipated procedures at the time of report review reduces the need for secondary procedure visits.

Skin Occlusion. Aside for variations in patient privacy preferences, skin occlusion is a function of the patient condition. In this study, we did not analyze the effect of age on outcomes, but undergoing scanning requires that the subject stand at attention and raise the arms above the head. Morbid obesity, pendulous breasts, extreme body hair, deep wrinkles and surgical defects also occlude skin. Patients with rashes, especially generalized conditions should be treated before scanning to improve the effectiveness of the studies.

Monitoring Technicians. Technician performance is key as well. Our technicians are supervised by a dermatologist by the case review method. The technician’s spot-picking behaviors are developed to complement the requirements of the skin disease in question. The diagnosis feedback loop is constant because the technician delivers the reports in person, in order that any key patient information gleaned in the study is optimally transferred. This function could be performed by a resident in training. Technicians in this role improve with extended training. Good eye function is a prerequisite for top performance.

Other Diseases. One important testimony to the function of the Melanoscan is that the technicians’ activity in tracking skin changes has been extended to application in the vast majority of skin diseases in our clinic, making melanoma screening one of many applications for time-lapse comparison studies. Experience improves performance. It is especially natural to use the Melanoscan to track the results of phototherapy in generalized diseases such as cutaneous T cell lymphoma, high-risk non-melanoma skin cancer, and psoriasis. The integration of the scanning apparatus with the phototherapy booth makes this a natural application, especially in the case of PUVA (Psoralen and Ultraviolet A) therapy that puts patients at increased risk for melanoma [21].

Medical Records. The Melanoscan imagery and analysis activity is integrated with the electronic medical record system of the office. Each report is reviewed by the dermatologist and the technician together, for quality assurance, diagnosis and clinical planning. Reports are communicated with stakeholders: referring and consulting physicians, patients, caregivers, and insurance companies. As with melanoma screening, subsequent care is facilitated by the ability to elucidate disease characteristics that were imaged previously and undergo transformation, benign or malignant. Reference to the baseline or follow-up studies of comprehensive cutaneous evaluation serves for three major purposes: 1) case presentation prior to the patient encounter by the clinical support staff; 2) anatomic mapping of key annotations such as clinical notes, operative reports, dermatoscopy images and pathology reports; 3) communications of consultations such as pathology requisitions and referrals to consulting surgeons are easily and accurately generated.

Quality Measures. The most important quality control for the entire system is the whole body skin examination in the clinic. Occluded areas are not effectively imaged and require examination. Rashes must be sorted out and final decisions made. Careful examination of the patient by an experienced dermatologist is the key to quality assurance. Missed lesions are tracked and reported. This is reinforced by the pathology requisition system embedded in the Melanoscan analysis software.

Resource Utilization. Possible disadvantages of this clinical approach: application of our strategy without clinical specificity may lead to excess resource utilization. However, because there is more work in the image analysis than in the capture, perhaps baseline images should be performed liberally with follow-up screening stratified by skin cancer risk assessment. The time demand of establishing baseline image information may be longer at the beginning, but should produce much needed evidence-based treatment methodologies for clinical evaluation and management events in dermatology.

V. Conclusion

Serial whole body photo comparison focuses the dermatologist on key changes. This method reduces clinician time spent accumulating evidence in support of informed diagnosis to the patient. Not only does the system place a premium on the dermatologist’s time, but early detection of melanoma results from this evidence-based analysis. The system employs technicians using serial automatic whole body imaging analysis to define a set of skin lesions to undergo further analysis by dermoscopy, surgery, and pathology. The primary test, the clinical image change detection by flicker analysis, improves the sensitivity and specificity of dermatology diagnosis as judged by pathology outcomes. The limited sample number in the study speaks to the pilot nature of this study. The demonstrable advantages of the system are enhanced physician accuracy, high patient acceptance, early melanoma detection, and reduced biopsy number.


1. Criscione VC, Weinstock MA. Melanoma thickness trends in the United States, 1988-2006. J Invest Dermatol 2010 Mar;130(30):793-797. [PubMed]

2. Tsao H, Atkins MB, Sober AJ. Management of cutaneous melanoma. N Engl J Med 2004;351:998-1012. [PubMed]

3. Tsao H, Rogers GS, Sober AJ. An estimate of the annual direct cost of treating cutaneous melanoma. J Am Acad Dermatol 1998;38:669-680. [PubMed]

4. Goodson AG, Grossman D. Strategies for early melanoma detection: approaches to the patient with nevi. J Am Acad Dermatol 2009;50:719-35. [PubMed]

5. Whited JD, Hall RP, Simel DL, Foy ME, Stechnuchak KM, Drugge RJ, Grichnik JM, Myers SA, Horner RD. Reliability and accuracy of dermatologists’ clinic-based and digital image consultations. J Am Acad Dermatol. 1999 Nov 41(5 Pt 1): 693-702. [PubMed]

6. Drugge RJ, Nguyen C, Drugge ED, Gliga L, Broderick PA, McClain SA, Brown CC. Melanoma screening with serial whole body photographic change detection using melanoscan technology. Dermatol Online J. 2009 Jun 15; 15(6): 1-8. [PubMed]

7. Kelly JW, Yeatman JM, Regalia C, Mason G,Henham AP. A high incidence of melanoma found in patients with multiple dysplastic naevi by photographic surveillance. Med J Aust 1997; 167:191-4. [PubMed]

8. Banky JP, Kelly JW, English DR, Yeatman JM, Dowling JP. Incidence of new and changed nevi and melanomas detected using baseline images and dermoscopy in patients at high risk for melanoma. Arch Dermatol 2005; 141:998-1006. [PubMed]

9. Feit NE, Dusza SW, Marghoob AA. Melanomas detected with the aid of total cutaneous photography. Br J Dermatol 2004; 150:706-14. [PubMed]

10. Wang SQ, Kopf AW, Koenig K, Polsky D, Nudel K, Bart RS. Detection of melanomas in patients followed up with total cutaneous examinations, total cutaneous photography, and dermoscopy. J Am Acad Dermatol 2004;50:15-20. [PubMed]

11. Fuller SR, Bowen GM, Tanner B, Florell SR, Grossman D. Digital dermoscopic monitoring of atypical nevi in patients at risk for melanoma. Dermatol Surg 2007;33:1198-206. [PubMed]

12. Schiffner R, Schniffner-Rohe J, Landthaler M, Stolz W. Long-term dermoscopic follow-up of melanocytic naevi: clinical outcome and patient compliance. Br J Dermatol 2003;149:79-86. [PubMed]

13. Lucas CR, Sanders LL, Murray JC, Myers SA, Hall RO Grichnik JM. Early melanoma detection: nonuniform dermoscopic features and growth. J Am Acad Dermatol. 2003;48(5):663-71. [PubMed]

14. Grichnik JM. Total Body Photography: A valid approach for improving patient care.

15. Papier A, Chalmers RJ, Byrnes JA, Goldsmith JA, Goldsmith LA. Framework for improved communication: the Dermatology Lexicon Project. J Am Acad Dermatol 2004 Apr;50(4):630-4. [PubMed]

16. Drugge R, Melanoma Risk Self-Assessment Test. SID Poster, Stamford, CT; 2006.

17. Drugge RJ, Naylor M, Smith R, Drugge ED, Nguyen C, Gliga L, Brown CC. Use of a Quantitative, Questionnarie-based Melanoma Risk Assessment Tool: analysis of Results and implications for Melanoma Risk Screening [In preparation].

18. Ke M, Moul D, Camouse M, Avram M, Carranza D, Soriano T, Lask G. Where is it? The utility of biopsy-site photography. Dermatol Surg. 2009 Dec 21. [Epub ahead of print] [PubMed]

19. Chang Y, Stanley RJ, Moss RH, Van Stoecker W. A systematic heuristic approach for feature selection for melanoma discrimination using clinical images. Skin res Technol 2005 Aug;11(3);165-178. [PubMed]

20. Malvehy J, Puig S. Follow-up of Melanocytic skin Lesions with digital total-body photography and digital dermoscopy: a two-step method. Clinics in Dermatology 2002;20:297-304. [PubMed]

21. Stern RS; PUVA Follow up Study. The risk of melanoma in association with long-term exposure to PUVA. J Am Acad Dermatol 2001 May;44(5):755-61. [PubMed]

© 2010 Dermatology Online Journal