Diagnostic and Prognostic Deep Learning Applications for Histological Assessment of Cutaneous Melanoma

Abstract

Melanoma is among the most devastating human malignancies. Accurate diagnosis and prognosis are essential to offer optimal treatment. Histopathology is the gold standard for establishing melanoma diagnosis and prognostic features. However, discrepancies often exist between pathologists, and analysis is costly and time-consuming. Deep-learning algorithms are deployed to improve melanoma diagnosis and prognostication from histological images of melanoma. In recent years, the development of these machine-learning tools has accelerated, and machine learning is poised to become a clinical tool to aid melanoma histology. Nevertheless, a review of the advances in machine learning in melanoma histology was lacking. We performed a comprehensive literature search to provide a complete overview of the recent advances in machine learning in the assessment of melanoma based on hematoxylin eosin digital pathology images. In our work, we review 37 recent publications, compare the methods and performance of the reviewed studies, and highlight the variety of promising machine-learning applications in melanoma histology.

Keywords: deep learning; dermatopathology; melanoma.

Figure 1

Overview of journal article identification and data-extraction methods.

Figure 2

(A) Comparison of the number of publications with diagnostic or prognostic applications. (B) Number of papers published each year. As our literature search was conducted on 21 March 2022, years for this figure are defined by 21st March of the labelled year through 21st March of the following year. (C) Total size of dataset used for each study.

Figure 3

Schematic overview of image-processing workflow for deep-learning applications with whole-slide images. Example depicts application in diagnosis of nevi and melanoma. (A) Upload of digital whole slide image (B) Tiling of the image (C) Training of deep learning model using features extracted from individual tiles (D) Classification of individual tiles (E) Final image classification based on tile majority.