Safety and Effectiveness Results for DERM-ASSESS II Prospective, Multi-Centre Study1,2
Overall study performance summary:
Study reported no adverse events as well as no significant difference between DermaSensor’s and dermatologists’ sensitivity and specificity.
DermaSensor’s sensitivity for melanoma was 95% while the study dermatologists’ was 90%. For non-melanoma skin cancer (NMSC), i.e. BCC and SCC, DermaSensor’s performance was 94% and the dermatologists’ was 98%.2
DermaSensor and Dermatologist Sensitivity For Skin Cancers
- There were no adverse events, confirming the safety of this light-based, non-invasive device.1,2
- There was no statistical difference between the sensitivity of DermaSensor (92%) and dermatologists’ (96%) nor between their specificity (32% and 37%, respectively) for lesions biopsied by the dermatologists per their standard of care.2,3
- DermaSensor’s specificity was 45% for benign lesions assessed by the study dermatologists as being suggestive of skin cancer to non-specialist healthcare professionals.2,4
- DermaSensor’s sensitivity results for melanoma, SCC and BCC were 95%, 93%, and 94%, respectively. Dermatologists’ sensitivity results were 90%, 96% and 100%, respectively.2,3
- As reported in Nature in 2019, “... most skin lesions are diagnosed by primary care doctors, and problems with inaccuracy have been underscored; if AI can be reliably shown to simulate experienced dermatologists, that would represent a significant advance.”5
- In a randomized, prospective study of DermaSensor utility with 57 GPs, these physicians made over 5,000 assessments of skin lesions. The study results showed that physicians’ correctly referred or biopsied cancerous lesions 13 percent more when the DermaSensor output was available to them, compared to their evaluation with no device output.6
- DermaSensor increased physicians’ cancer detection sensitivity from 81% to 94%, and this improvement was statistically significant (p = .0009).6 There was no statistically significant change in the GP’s specificity, or false positive rate, for benign lesions (p = .3558).6
DermaSensor Sensitivity Summary Table
| Pathology | Study 0027 | DERM-ASSES II2 | NZ IIS9 |
|---|---|---|---|
| All skin cancers | 94% | 92% | 98% |
| Melanoma | 100% | 95% | N/A8 |
| SCC | 93% | 93% | 100% |
| BCC | 94% | 94% | 100% |
DermaSensor Specificity Summary Table
| Pathology | Skin specialist biopsied lesions of concerns (DA-II)2 | GP lesions of concern | Patient lesions of concern2 | |
|---|---|---|---|---|
| DA-II2,4 | NZ IIS9 | |||
| All benign lesion | 32% | 45% | 47% | 48% |
| Benign nevi | 50% | 75% | 60% | 43% |
| Seborrheic Keratosis | 38% | 20% | 52% | 64% |
| AK | 24% | 50% | 27% | 43% |
| Benign Other | 45% | 61% | N/A8 | 33% |
DermaSensor NPV/PPV Performance Across Spectral Score Ranges 1-10
DermaSensor NPV/PPV Performance
| Performance Metric | NZ IIS9 | Patient-Select2 |
|---|---|---|
| NPV | 99.5% | 98.6% |
| PPV | 18.2% | 10.6% |
| 1-4 PPV | 7.1% | 8.6% |
| 5-7 PPV | 27.9% | 12.5% |
| 8-10 PPV | 58.6% | 66.7% |
Device Sensitivity and Specificity for Detecting Malignant Lesions9
| Performance Metric | Result | Exact 95% CI |
|---|---|---|
| Specificity | 46.5% | 41.8% to 51.2% |
| Specificity Excluding AKs | 52.6% | 47.2% to 58.0% |
| Sensitivity | 98.2% | 90.3% to 100.0% |
Device NPV and PPV with and without AKs for Detecting Malignant Lesions9
| Performance Metric | Result | Exact 95% CI |
|---|---|---|
| NPV | 99.5% | 97.4% to 100.0% |
| NPV Excluding AKs | 99.5% | 97.0% to 100.0% |
| PPV | 18.2% | 14.0% to 23.0% |
| PPV Excluding AKs | 24.9% | 19.3% to 31.2% |
Indications for Use
The DermaSensor™ device is indicated for use as an objective tool to assist qualified healthcare professionals in evaluating skin lesions suggestive of melanoma, basal cell carcinoma, and/or squamous cell carcinoma. The DermaSensor device is intended to assist the user in deciding whether skin lesions require further clinical care and is not intended to be used for direct diagnosis of skin cancer. DermaSensor is only for use by qualified healthcare professionals appropriately trained in
the assessment of skin lesions for cancer.
Risks
False-positive and false-negative results may lead to unnecessary care or to a malignant skin lesion not being optimally managed, respectively. However, it is important to note that biopsy is used to confirm pathology and that elastic scattering spectroscopy is to be used as an adjunctive tool to visual inspection and history-taking.
Spectral Score Groupings 1-5 and 6-109
| Spectral Scores Groupings | PPV | Frequency of ‘Investigate Further’ Lesions |
|---|---|---|
| 1-5 | 8.5% | 76.8% |
| 6-10 | 42.4% | 23.2% |
Spectral Score Groupings 1-4, 5-7 and 8-109
| Spectral Scores Groupings | PPV | Frequency of ‘Investigate Further’ Lesions |
|---|---|---|
| 1-4 | 7.1% | 64.2% |
| 5-7 | 27.9% | 28.7% |
| 8-10 | 58.6% | 7.1% |
1Benvenuto-Andrade C, Manolakos D, Cognetta AB. Safety and Effectiveness of Elastic Scattering Spectroscopy and Machine Learning in the Evaluation of Skin Lesions. Poster Presentation, World Congress of Teledermatology, Nov 2020.
2Data on file, DermaSensor Inc.
3For lesions biopsied in DERM-ASSESS II, dermatologist performance (i.e. dermatologist sensitivity and specificity) is based on the study dermatologists’ in-person binary assessment of biopsied lesions as being malignant or benign, prior to receiving pathology results.
4For unbiopsied lesions the dermatologists’ clinical determination of the lesion as benign was used as the reference standard; however, for the dermatologists’ clinical assessment there was no reference standard since no biopsies were performed and accordingly no specificity is reported for their evaluations.
5Topol E. High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine. 2019;25:44-66.
6Tepedino K, Tablada A, Barnes E, Da Silva, T. Clinical Utility of a Handheld Elastic Scattering Spectroscopy Tool and Machine Learning on the Diagnosis and Management of Skin Cancer by Primary Care Physicians. Poster Presentation, SDPA Fall Conference, Nov 4-7, 2021.
7Rodriguez-Diaz E, et al. Optical Spectroscopy as a Method for Skin Cancer Risk Assessment. Photochem Photobiol. 2019;95(6):1441-1445.
8N/A: Where the sample size was less than 10 lesions, results for sensitivity and specificity were excluded due to small sample size.
9Salmon P and Bonning M. Use of Elastic-scattering Spectroscopy and Machine Learning When Assessing Skin Lesions Suggestive of Skin Cancer, Poster Presentation, SDPA Fall Conference, Nov 4-7, 2021.