Enhanced MobileNet for skin cancer image classification with fused spatial channel attention mechanism

Abstract

Skin Cancer, which leads to a large number of deaths annually, has been extensively considered as the most lethal tumor around the world. Accurate detection of skin cancer in its early stage can significantly raise the survival rate of patients and reduce the burden on public health. Currently, the diagnosis of skin cancer relies heavily on human visual system for screening and dermoscopy. However, manual inspection is laborious, time-consuming, and error-prone. In consequence, the development of an automatic machine vision algorithm for skin cancer classification turns into imperative. Various machine learning techniques have been presented for the last few years. Although these methods have yielded promising outcome in skin cancer detection and recognition, there is still a certain gap in machine learning algorithms and clinical applications. To enhance the performance of classification, this study proposes a novel deep learning model for discriminating clinical skin cancer images. The proposed model incorporates a convolutional neural network for extracting local receptive field information and a novel attention mechanism for revealing the global associations within an image. Experimental results of the proposed approach demonstrate its superiority over the state-of-the-art algorithms on the publicly available dataset International Skin Imaging Collaboration 2019 (ISIC-2019) in terms of Precision, Recall, F1-score. From the experimental results, it can be observed that the proposed approach is a potentially valuable instrument for skin cancer classification.

Keywords: Classification; Deep learning; Machine vision; Medical image analysis; Skin cancer.

Fig. 1

The accuracy curves of the proposed approach on the ISIC-2019 dataset.

Fig. 2

The loss curves of the proposed approach on the ISIC-2019 dataset.

Fig. 3

The confusion matrix of the proposed approach on the ISIC-2019 dataset. (Eight different types of skin cancer images).

Fig. 4

The ROC curve of the proposed approach on the ISIC-2019 dataset (Eight different types of skin cancer images).

Fig. 5

The influence of various attention mechanisms on the proposed approach for classifying the images in ISIC-2019 dataset.

Fig. 6

Performance comparison between the state-of-the-art attention mechanisms and the proposed approach on the ISIC-2019 dataset.

Fig. 7

Example of misclassified image and its corresponding heat map in the ISIC 2019 dataset using the proposed approach.

Fig. 8

Image samples in the ISIC 2019 dataset.

Fig. 9

The underlying structure of the model that was constructed is its core architecture. Bnect is derived from the term bottleneck module, while FSCA refers to the suggested fused spatial-channel attention mechanism. Additionally, GAP indicates the global average pooling unit.

Fig. 10

The multi-scale feature module that is employed to extract the multi-scale feature from an input image. The operation of batch normalization is denoted as BN.

Fig. 11

The depiction of the proposed attention mechanism that is utilized in the suggested method.