Central Skin: Tell me about your background and education?
My name is Thiago Pradi, I am 25 years old and I'm a software developer from Brazil. I graduated with a degree in Computer Science from FURB, the regional university of Blumenau. I'm currently living in Jaraguá do Sul, a small town from the south of Brazil.
Central Skin: When did you first start to get an interest in product development for dermatology?
I started to became interested in the field after taking some classes from the teacher that mentored my research (Prof. Aurélio Hoppe). He was giving a lecture on Geometric Algorithms, and he tried to illustrate the principles to the class with some real world examples. So, he showed some research in the health informatics field, including one for automated melanoma detection. So, I became really interested in the field, and it was the beginning of my research.
Central Skin: How common is skin cancer in Brazil?
Surprisingly, skin cancer is pretty common in Brazil. Skin cancer is the most common type of cancer in Brazil and corresponds to 25% of all malignant tumors registered in the country. Skin cancer is most prevelant mainly in the South/Southeast of Brazil, because of the higher UV indices and because of the local population of Germans/Italian descendants, with lighter skin types (which are more likely to have skin cancer). Also, the town where my college is located has one of the highest rates of melanoma in the world!!
Central Skin: What were some of the challenges in creating an automatic technology for recognizing skin cancer?
There are a lot of challenges in detecting skin cancer. Mainly, there are a lot of different skin cancers, with different characteristics and visual forms. My research was focused only on Melanoma. The first challenge was the specific knowledge, because I'm a computer science student, and I didn't had any dermatological information. I studied a bit about, and had help from a Dermatologist (Dr. Roberto Coswig Fiss). In the programming part, we still faced a lot of challenges like:
- Different illumination on the images
- Low contrast between the lesion and the skin surrounding the lesion
- Presence of hair and other noises in the image
- Different skin colors
Also, the hardest part of the algorithm is to translate the ABCD rule into a Computer Vision algorithm, that needs to be fast enough to run into a mobile device (like a iPhone)
My application is called iMelanoma. It consist of a iPhone application that the user can take a picture of the lesion, run the algorithm and see the results. It was tested with a educational database of skin lesions, with 90% success rate. Nevertheless, my application didn't get tested yet in a clinical trial, so, it's not suitable for clinical use right now.
It was pretty good. My application was featured in a few different medias( like the Jornal of the Brazilian Society of Dermatology). Also, I did some interviews for local TV Channels. Everyone is excited with this kind of technology, including the doctors.
Comments
Leave a Comment