**3.2 Unsupervised learning for dermatology**

Unsupervised learning means that the algorithm is only given input data without corresponding output values. This type of algorithm is more of an exploration and does not have the correct output value [83]. Therefore, unsupervised methods are often suitable for situations where the statistics vary widely. For example, in a study

*Integrated Optical Coherence Tomography and Deep Learning for Evaluating of the Injectable… DOI: http://dx.doi.org/10.5772/intechopen.106006*

**Figure 10.**

*Deep CNN layout. Reprinted with permission from reference [80].*

by Kharazmi et al. [84] detecting basal cell carcinoma (BCC), unsupervised learning was used to classify vascular features in BCC dermoscopy for automatic cancer detection. The results show that the framework outperforms the state-of-the-art norm supervised learning results by 7% in Area Under Curve (AUC), with a sensitivity of 98.1% for BBC detection. This approach eliminates the need for supervised learning to use large and dense datasets at multiple scales to find appropriate images to build blocks of visual content [85]. So the advantage over supervised learning is that the algorithm is free to act in order to learn more about the data and present interesting findings. It is popular in clustering applications or the behavior of discovering groups in data and associations or predicting the behavior of rules that describe data.
