AI-Assisted Smartphone Dermoscopy for Skin Cancer

Overseen By Khoa Nguyen

Age: 18+

Sex: Any

Trial Phase: Academic

Sponsor: OHSU Knight Cancer Institute

Disqualifiers: Fitzpatrick type 5 or 6, Pregnant, others

No Placebo Group All trial participants will receive the active study treatment (no placebo)

Approved in 1 Jurisdiction This treatment is already approved in other countries

Compensation: Varies

Number of Visits: 1

Trial Summary

Do I need to stop my current medications for this trial?

The trial protocol does not specify whether you need to stop taking your current medications. However, since the study focuses on skin lesion analysis and not on medication effects, it's likely you won't need to stop your medications. Please confirm with the study coordinator.

What data supports the idea that AI-Assisted Smartphone Dermoscopy for Skin Cancer is an effective treatment?

The available research shows that AI-Assisted Smartphone Dermoscopy can classify melanomas with accuracy similar to experienced dermatologists. One study found that the AI system had a high sensitivity and specificity for identifying melanomas, meaning it was good at correctly identifying both the presence and absence of the disease. Additionally, patients generally felt positive about using AI to help doctors diagnose and manage skin cancer, indicating that it could be a useful tool in healthcare settings. However, the AI system needs more data to improve its accuracy for other skin conditions like seborrheic keratoses. ^{1 2 3 4 5}

What safety data exists for AI-assisted smartphone dermoscopy for skin cancer?

The research indicates that AI-based smartphone apps for skin cancer risk assessment need more evaluation and better regulation. Current CE marking processes do not adequately protect the public from risks associated with these apps. There is a need for more robust evidence and regulation to ensure safety. ^{6 7 8 9 10}

Is AI-Assisted Smartphone Dermoscopy a promising treatment for skin cancer?

Yes, AI-Assisted Smartphone Dermoscopy is a promising treatment for skin cancer. It helps doctors quickly and accurately diagnose skin cancer using smartphone apps, making it easier for patients to get the care they need. This technology can speed up the process of getting a diagnosis and improve early detection of melanoma, a serious type of skin cancer. Patients also feel confident and comfortable with this technology being used in their care. ^{1 2 4 5 11}

What is the purpose of this trial?

This is a new protocol to analyze how the use of the Sklip System enables laypersons to safely triage self-selected pigmented skin lesions of concern (PSLCs) from home with the same or better accuracy than pre-specified performance goals for the detection of PSLCs that require biopsy (Melanoma and atypical melanocytic nevi with uncertain malignant, Squamous cell carcinoma, Basal cell carcinoma). The study protocol will also compare the accuracy of the Sklip System when used by a layperson (Participant) versus near-perfect Sklip System user (Study Coordinator), assess whether Sklip System improves triage of PSLCs \< 6 mm in diameter and triage of thin melanomas with \<0.8 mm Breslow depth as suspicious, as compared to the current medical provider virtual triage method that relies on store-and-forward of smartphone clinical images (SCI), and assess accuracy of layperson-performed self-skin-exams (SSEs) at-home in the identification of all suspicious PSLCs present on their body as compared to the same layperson (Participant) evaluated with a full body skin examination (FBSE) by a dermatology Provider (DP) in-person.

Research Team

Sancy A. Leachman, MD, PhD

Principal Investigator

OHSU Knight Cancer Institute

Eligibility Criteria

This trial is for adults over 21 with at least one mole, who are not urgently sick. They must have skin types 1-4, speak English, be able to use a smartphone/tablet for communication and give consent. Excluded are those with recent skin checks, darker skin types (5-6), vulnerable groups like children or prisoners, vision impaired individuals, and pregnant people.

Inclusion Criteria

Participant must be a current or new patient through self-referral or Provider-referral at the participating Study Site.

I speak English.

I am 21 or older and have at least one mole on my body.

Exclusion Criteria

Participant who have had a skin check visit with a dermatology Provider within the last 90 days will be excluded to avoid self-selection bias, unless the Participant identifies a new unexamined (not previously documented) spot of concern.

You have very dark skin (Fitzpatrick Skin Type 5 or 6).

If you are pregnant, you cannot take part in this study.

Timeline

Screening

Participants are screened for eligibility to participate in the trial

2 weeks

At-Home Examination

Participants perform self-skin exams using naked-eye criteria and take smartphone clinical images (SCI) and digital dermoscopy images (DDIs) of each pigmented skin lesion of concern (PSLC). The Sklip System is applied to each PSLC of concern.

2 weeks

At-home

In-Office Examination

Participants undergo an in-office full body skin exam (FBSE) by a dermatology provider.

4 weeks

1 visit (in-person)

Follow-up

Participants are monitored for accuracy of triage and biopsy results, including the assessment of suspicious lesions and their pathology reports.

4 months

Treatment Details

Interventions

• At-Home Dermoscopy Artificial Intelligence

Trial Overview The Sklip System's ability to help non-professionals check moles at home is being tested against traditional medical evaluations. It aims to see if laypersons can accurately identify suspicious lesions needing biopsy and compares this method to virtual triage by medical providers using photos.

Participant Groups

Group I: Self-skin exam (SSE), digital dermoscopy image (DDI), Sklip System, full body skin exam (FBSE). Experimental Treatment 6 Interventions

This is a single-arm prospective trial. Participants perform self-skin exams using naked-eye criteria and will take smartphone clinical images (SCI) of each PSL of concern (PSLC). Participants will also take digital dermoscopy images (DDIs) and apply the Sklip System (integrating the Sklip Mole Scan Algorithm (SMSA) to each PSLC of concern in up to 14 days. Within up to 28 days of completing the at-home exams, participants will undergo an in-office visit full body skin exam (FBSE).

Who Is Running the Clinical Trial?

OHSU Knight Cancer Institute

Lead Sponsor

Trials

239

Recruited

2,089,000+

Oregon Health and Science University

Collaborator

Trials

1,024

Recruited

7,420,000+

Findings from Research

A meta-analysis of 30 studies found that artificial intelligence had a slightly higher sensitivity (91%) for melanoma diagnosis compared to dermoscopy (88%), indicating AI may be slightly better at identifying true positives.

However, dermoscopy showed significantly better specificity (86%) than artificial intelligence (79%), meaning it was more effective at correctly identifying non-melanoma cases. Overall, both methods performed similarly in diagnosing melanocytic skin lesions.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Systematic review of dermoscopy and digital dermoscopy/ artificial intelligence for the diagnosis of melanoma. Rajpara, SM., Botello, AP., Townend, J., et al.[2018]

A study involving 268 patients assessed their perspectives on the use of artificial intelligence as a medical device (AIaMD) in skin cancer diagnosis, revealing a generally positive sentiment towards its integration into healthcare.

Most respondents expressed confidence in AIaMD assisting doctors with diagnoses and management plans, indicating strong patient acceptability for this technology in dermatology, which is crucial for its successful implementation.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Patient perspectives of artificial intelligence as a medical device in a skin cancer pathway. Kawsar, A., Hussain, K., Kalsi, D., et al.[2023]

The smartphone-based 'You Only Look Once' neural network model demonstrated high sensitivity (0.88) and specificity (0.87) for classifying melanomas, performing comparably to experienced dermatologists.

The algorithm also outperformed beginner dermatologists in identifying melanocytic nevi, but its sensitivity for seborrheic keratoses was lower (0.52), indicating a need for a larger dataset to improve accuracy in this category.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Accuracy of a Smartphone-Based Artificial Intelligence Application for Classification of Melanomas, Melanocytic Nevi, and Seborrheic Keratoses. Liutkus, J., Kriukas, A., Stragyte, D., et al.[2023]

The TELESPOT project developed a smartphone-based dermoscopy application that achieved high satisfaction rates among patients (89%) and general practitioners (94%) in primary healthcare settings.

In a study involving 80 patients, the application identified high-priority skin cancer cases in 7.6% of instances, demonstrating its effectiveness in facilitating rapid diagnosis and improving patient flow to specialized care.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Value of Teledermoscopy in Primary Healthcare Centers: Preliminary Results of the TELESPOT Project in Belgium. Damsin, T., Canivet, G., Jacquemin, P., et al.[2020]

The study assessed the agreement in malignancy risk evaluation of pigmented skin lesions (PSLs) between a specialist and an artificial neural network (ANN)-based digital dermoscopy system, highlighting the potential of digital tools in skin cancer assessment.

However, the ANN algorithm needs additional training and validation to ensure its reliability for clinical use, indicating that while digital dermoscopy is promising, it is not yet fully ready for widespread application in evaluating PSL malignancy risk.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Evaluation of digital dermoscopy in a pigmented lesion clinic: clinician versus computer assessment of malignancy risk. Boldrick, JC., Layton, CJ., Nguyen, J., et al.[2019]

In a study with 32 participants, those who received a correct paraphrase of their medical query from conversational assistants were more likely to follow the advice given, highlighting the importance of accurate information delivery.

Participants who received a disclaimer message were significantly more inclined to consult a physician before acting on the advice, suggesting that incorporating disclaimers can help mitigate the risks of harmful medical recommendations from conversational assistants.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Mitigating Patient and Consumer Safety Risks When Using Conversational Assistants for Medical Information: Exploratory Mixed Methods Experiment. Bickmore, TW., Ólafsson, S., O'Leary, TK.[2022]

The approval of AI/ML-based medical devices has significantly increased since 2015, particularly in radiology, with 222 devices approved in the USA and 240 in Europe, though few are classified as high-risk.

There is currently no specific regulatory pathway for these devices in either region, highlighting the need for improved transparency and rigorous regulation to ensure their safety and efficacy.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Approval of artificial intelligence and machine learning-based medical devices in the USA and Europe (2015-20): a comparative analysis. Muehlematter, UJ., Daniore, P., Vokinger, KN.[2021]

AI-based medical devices are already available in Europe, but the regulatory framework is less stringent than in the USA, allowing for a wider variety of devices to enter the market.

The current EU regulations place a significant burden on consumers to verify the effectiveness and safety of these AI devices, emphasizing the need for better oversight and understanding of the legal implications in clinical practice.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Approval and Certification of Ophthalmic AI Devices in the European Union. Grzybowski, A., Brona, P.[2023]

Smartphone apps that use artificial intelligence are becoming more common in healthcare, but there is a need for strong evidence to support their effectiveness and safety.

Current regulatory processes for CE marking these diagnostic apps are insufficient, potentially exposing users to risks associated with their use.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

AI-based smartphone apps for risk assessment of skin cancer need more evaluation and better regulation. Matin, RN., Dinnes, J.[2022]

An AI solution trained on Pillcam® images demonstrated high sensitivity (97.4%) and specificity (98.8%) for detecting angiectasias, even when applied to images from a different system (MiroCam®), indicating its potential effectiveness across different capsule endoscopy platforms.

The study suggests that AI can significantly reduce processing time for image analysis, with MiroCam® images processed faster (20.7 ms) than Pillcam® images (24.6 ms), highlighting the efficiency of AI in enhancing capsule endoscopy interpretation.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

A multisystem-compatible deep learning-based algorithm for detection and characterization of angiectasias in small-bowel capsule endoscopy. A proof-of-concept study. Houdeville, C., Souchaud, M., Leenhardt, R., et al.[2022]

The novel decision forest classifier demonstrated a high sensitivity of 97.4% for detecting melanoma in a study of 173 dermoscopic images, suggesting it could be a valuable tool for early melanoma detection.

While the classifier's specificity was lower at 44.2%, it outperformed 30 dermatology clinicians in sensitivity, indicating it may assist in clinical decision-making regarding biopsies for skin lesions.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Computer-aided classification of melanocytic lesions using dermoscopic images. Ferris, LK., Harkes, JA., Gilbert, B., et al.[2017]

References

1. England pubmed.ncbi.nlm.nih.gov

Systematic review of dermoscopy and digital dermoscopy/ artificial intelligence for the diagnosis of melanoma. [2018]

2. Switzerland pubmed.ncbi.nlm.nih.gov

Patient perspectives of artificial intelligence as a medical device in a skin cancer pathway. [2023]

3. Switzerland pubmed.ncbi.nlm.nih.gov

Accuracy of a Smartphone-Based Artificial Intelligence Application for Classification of Melanomas, Melanocytic Nevi, and Seborrheic Keratoses. [2023]

4. Switzerland pubmed.ncbi.nlm.nih.gov

Value of Teledermoscopy in Primary Healthcare Centers: Preliminary Results of the TELESPOT Project in Belgium. [2020]

5. United States pubmed.ncbi.nlm.nih.gov

Evaluation of digital dermoscopy in a pigmented lesion clinic: clinician versus computer assessment of malignancy risk. [2019]

6. Canada pubmed.ncbi.nlm.nih.gov

Mitigating Patient and Consumer Safety Risks When Using Conversational Assistants for Medical Information: Exploratory Mixed Methods Experiment. [2022]

7. England pubmed.ncbi.nlm.nih.gov

Approval of artificial intelligence and machine learning-based medical devices in the USA and Europe (2015-20): a comparative analysis. [2021]

8. England pubmed.ncbi.nlm.nih.gov

Approval and Certification of Ophthalmic AI Devices in the European Union. [2023]

9. England pubmed.ncbi.nlm.nih.gov

AI-based smartphone apps for risk assessment of skin cancer need more evaluation and better regulation. [2022]

10. Netherlands pubmed.ncbi.nlm.nih.gov

A multisystem-compatible deep learning-based algorithm for detection and characterization of angiectasias in small-bowel capsule endoscopy. A proof-of-concept study. [2022]

11. United States pubmed.ncbi.nlm.nih.gov

Computer-aided classification of melanocytic lesions using dermoscopic images. [2017]