New Trends in Melanoma Detection Using Neural Networks: A Systematic Review

Abstract

Due to its increasing incidence, skin cancer, and especially melanoma, is a serious health disease today. The high mortality rate associated with melanoma makes it necessary to detect the early stages to be treated urgently and properly. This is the reason why many researchers in this domain wanted to obtain accurate computer-aided diagnosis systems to assist in the early detection and diagnosis of such diseases. The paper presents a systematic review of recent advances in an area of increased interest for cancer prediction, with a focus on a comparative perspective of melanoma detection using artificial intelligence, especially neural network-based systems. Such structures can be considered intelligent support systems for dermatologists. Theoretical and applied contributions were investigated in the new development trends of multiple neural network architecture, based on decision fusion. The most representative articles covering the area of melanoma detection based on neural networks, published in journals and impact conferences, were investigated between 2015 and 2021, focusing on the interval 2018-2021 as new trends. Additionally presented are the main databases and trends in their use in teaching neural networks to detect melanomas. Finally, a research agenda was highlighted to advance the field towards the new trends.

Keywords: deep learning; image classifiers; image processing; image segmentation; machine learning; melanoma detection; neural networks; review; skin lesion; statistic performances.

Figure 1

Artifacts in Me images collected from the ISIC 2016 dataset [14]: (a–c)—presence of hair, (d)—presence of blood vessels, (e,f)—presence of oil drops.

Figure 2

Methods workflow for Me detection: (a) classical method, (b) NN approach.

Figure 3

Searches for important terms in the Web of Science, Scopus, and PubMed DBs between 2015 and 2021 with the AND connector: (a) CNN AND Me, (b) DL AND Me, (c) ML AND Me, and (d) AI AND Me.

Figure 3

Searches for important terms in the Web of Science, Scopus, and PubMed DBs between 2015 and 2021 with the AND connector: (a) CNN AND Me, (b) DL AND Me, (c) ML AND Me, and (d) AI AND Me.

Figure 4

PRISMA flow diagram of our research.

Figure 5

Frequently DSs used in Me detection between 2018 and 2020.

Figure 6

The four most used DSs for Me detection in 2021 (percentage).

Figure 7

Frequently NNs used in Me detection between 2018 and 2020.

Figure 8

The most used NNs for Me detection in 2021 (percentage).

Figure 9

AlexNet basic architecture.

Figure 10

Inception module used in GoogLeNet.

Figure 11

GoogleNet architecture’s simplified block diagram.

Figure 12

Inception v3 basic architecture.

Figure 13

VGG 16 network architecture [98].

Figure 14

VGG 19 network architecture [98].

Figure 15

Residual block.

Figure 16

ResNet-152 basic architecture.

Figure 17

YOLO v3 architecture [101].

Figure 18

Xception network architecture.

Figure 19

EfficientNet architecture [107].

Figure 20

Five-layer DenseNet architecture [108].

Figure 21

U-Net architecture [110].

Figure 22

GAN standard network architecture.

Figure 23

Percent of research papers per year with the highest impact for the new trends in Me detection by NN.

Figure 24

The schematic architecture of the proposed system for hair removal from skin lesion images from [117].

Figure 25

The architecture of the proposed system for skin lesion classification [1].

Figure 26

Multi-network system architecture based on decision fusion for Me detection [5].

Figure 27

Ensemble strategy of the group decision [52].

Figure 28

The architecture of the Me classification system proposed in [6], based on several NNs connected on two levels of classification.

Figure 29

The schematic architecture of the skin lesion classification system based on CNN for the segmentation, feature extraction, and intelligent classification [59].

Figure 30

The architecture of the SL classification system proposed in [43].