Melanoma (Skin Cancer)
Conditions
Keywords
melanoma, machine learning, artificial intelligence, diagnostic accuracy, clinical desicion supportive tool, Neoplasms by Histologic Type, Neoplasms, Nevi and Melanomas, Skin Neoplasms, Neoplasms by Site, Skin Diseases
Brief summary
In Sweden, approximately 9,000 Swedes are affected by melanoma annually, and each year, 500 individuals die from metastatic melanoma. The prognosis for melanoma primarily depends on the thickness of the tumor at diagnosis. Melanomas that only grow in the epidermis and have not yet grown into the dermis are called melanoma in situ or pre-melanoma. These melanomas lack the potential to spread in the body. Melanomas that grow into the dermis, on the other hand, are called invasive or malignant melanomas. Invasive melanomas have the potential to spread in the body. To improve melanoma diagnostics, a dermatoscope is used. A dermatoscope is a type of magnifying glass equipped with a strong light. Using a dermatoscope makes the structures in the epidermis and dermis clearer. Although most melanomas are relatively easy to detect, it is often difficult to determine whether melanomas are invasive or in situ based on the dermatoscopic image. Despite the fact that all suspected melanomas (regardless of melanoma depth) should be operated on, it is important to form an opinion on whether the melanoma is invasive or in situ. This decision is important because it: 1. Provides guidance on how quickly healthcare must prioritize a patient for surgery. 2. Provides guidance on the margin the surgeon chooses for the first operation. 3. Affects the information we give the patient even before the first operation. In recent years, several applications of machine learning have shown great potential in research contexts within dermatology and venereology. However, these tools have been evaluated to a very limited extent in clinical trials, which is naturally a prerequisite before they can be safely implemented in routine healthcare.
Interventions
DEVICEMedical Device Dermalyser
Dermalyser is an Artificial Intelligence (AI) application (app) that allows medical professionals to take pictures of cutaneous lesions with the help of a smartphone camera. A dermatoscope is connected to the smartphone camera and is used to take the digital image of cutaneous lesions with suspicion of melanoma. Based on image processing algorithms, the app does a detailed analysis of the captured cutaneous lesion. In this clinical investigation, the objective is to test the device performance in a prospective setting in patients with a suspicion of primary cutaneous melanoma, to validate the added AI component. The intended purpose of the device is not to replace the physician's assessment, but rather to assist physicians in their assessment. Consequently, the final device should be regarded as a second opinion to augment clinical decision-making. The ultimate aim is to develop a tool that may augment clinical decision-making.
Sponsors
Vastra Gotaland Region
Study design
Observational model
OTHER
Time perspective
PROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
1. Patient needs to be ≥18 years of age at inclusion 2. The patient has a suspected primary melanoma where surgery is planned 3. Willingness and ability to provide informed consent
Exclusion criteria
1. The suspected lesion size is too small or too large to fit the white circle in the screen even after zooming in and out at its maximum. The lesion should not be \< 2 mm or \> 20 mm in diameter. 2. As per judgement by the investigator, to exclude when there are factors that may affect the quality of the photo such as when: 1. Lesion with any form of modifications such as tattoos, pen markings, and/or pigments covering or surrounding skin lesions. 2. Lesions located on areas such as genitalia, acral skin (foot sole or palm of the hand), mucosal surfaces (mouth, eye, nose, etc.), psoriasis or similar skin conditions. 3. Lesions masked by an excessive amount of hair, generally found in armpits, scalp, beards or other hairy parts of the body. 4. Lesions that are scarred, non-intact, injured, previously biopsied or undergone surgery. in areas that are not suitable for dermoscopy imaging. 3. Individuals with skin type V and VI according to the Fitzpatrick scale (darker brown or black coloured skin) 4. Patients that do not perform surgery or die before the planned surgery 5. Missing or uninterpretable diagnostic data from the Department of Pathology.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Area under the curve (AUC) for discrimination between invansive and in situ melanoma for the in distribution data set. | From enrollment to the end of the inclusion when the images of the lesion have been obtained. This will most often be achieved on the same day as that the patient is included in the study. | Please note that preoperatively there will still be uncertainty if an included lesion will be histopathologically verified as a melanoma. After histopathological analysis all lesions will be separated into an in distrubution data set, i.e. lesions that were confirmed as melanomas and an out-of distribution set, i.e., lesions that proved to have an alternative diagnosis. Please note that the primary outcome measure will be limited the in distrubution data set, i.e., lesions that turned out to be a melanoma. Image analysis with the help of AI tool Dermalyser in a prospective setting. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Sensitivity and Specificity | From enrollment to the end of the inclusion when the images of the lesion have been obtained. This will most often be achieved on the same day as that the patient is included in the study. | Please note that preoperatively there will still be uncertainty if an included lesion will be histopathologically verified as a melanoma. After histopathological analysis all lesions will be separated into an in distrubution data set, i.e. lesions that were confirmed as melanomas and an out-of distribution set, i.e., lesions that proved to have an alternative diagnosis. Please note that the sensitivity and specificity measure will be limited the in distrubution data set, i.e., lesions that turned out to be a melanoma. Image analysis with the help of AI tool Dermalyser in a prospective setting. |
Other
| Measure | Time frame | Description |
|---|---|---|
| Algorithmic and physicians assessment of the out of distribution set | From enrollment to the end of the inclusion when the images of the lesion have been obtained. This will most often be achieved on the same day as that the patient is included in the study. | Please note that preoperatively there will still be uncertainty if an included lesion will be histopathologically verified as a melanoma. After histopathological analysis all lesions will be separated into an in distrubution data set, i.e. lesions that were confirmed as melanomas and an out-of distribution set, i.e., lesions that proved to have an alternative diagnosis. Please note that this pre-specified outcome will be limited the out of distrubution data set, i.e., lesions that turned out not to be a melanoma. Image analysis with the help of AI tool Dermalyser in a prospective setting. |
Countries
Sweden
Contacts
Primary ContactSam Polesie, MD, PhD
Backup ContactFilippos Giannopoulos, MD
Outcome results
None listed