User:Toggle78/Dermatoscopy

Lead
Dermatoscopy also known as dermoscopy or epiluminescence microscopy, is the examination of skin lesions with a dermatoscope. It is a tool smiilar to a camera to allow for inspection of skin lesions unobstructed by skin surface reflections. The dermatoscope consists of a magnifier, a light source (polarized or non-polarised), a transparent plate and sometimes a liquid medium between the instrument and the skin. Dermatoscope is often handheld, although there are stationary cameras allowing the capture of whole body images in a single shot. When the images or video clips are digitally captured or processed, the instrument can be referred to as a digital epiluminescence dermatoscope. The image is then analyzed automatically and given a score indicating how dangerous it is. This technique is useful to dermatologists and skin cancer practitioners in distinguishing benign from malignant (cancerous) lesions, especially in the diagnosis of melanoma.

Types of dermatoscopy
There are two main types of dermatoscopes, hand held portable and stationary mounted type.

A hand held dermatoscope is composed of a transilluminating light source and a magnifying optic (usually a 10-fold magnification). There are three main modes of dermoscopy:


 * Nonpolarized light, contact [1]
 * Polarized light, contact [2]
 * Polarized light, noncontact [3]

Polarized light allows for visualization of deeper skin structures, while non-polarized light provide information about the superficial skin. Most modern dermatoscopes allow the user to toggle between the two modes, which provide complementary information.Others may also allow the user to have different zoom levels and color overlay.

A stationary type allows a full body image to be captured in one snap. It is then transfered into image analysis algorithms that generates a three dimensional model of the person. Lesions on the person are marked and analyzed using Artificial intelligence.

Artificial Intelligence in Dermatoscopy
Artificial intelligence is used to automatically distinguish benign from malignant (cancerous) lesions. Modern software technology allows the usage of databases to aid in this process. Patients will consent their lesion pictures to be stored in a database which acts as an archive and allow Artificial intelligence programs to compare newly taken ones. The program then compares key features of a new image with known features of benign and malignant lesions. Often times a score is given to a specific lesion, indicating how dangerous and likely it is to be a malignant lesion. It is then flagged for further examination through a dermatologist. This speeds up the diagnosis process.

One limit is that since not many patients get their lesions documented, the sample size is miniscule compared to what an AI needs.

Proposed solutions include generating synthetic images of skin lesions to improve the algorithm. Then, the AI needs to differentiate whether the sample came from the synthetic samples or from real data sets. It needs to minimize the probability that it will predict its outputs as fake while also maximizing its probability to correctly distinguish between real and fake samples.