Skin Lesion Segmentation Ensemble with Diverse Training Strategies

Abstract

This paper presents a novel strategy to perform skin lesion segmentation from dermoscopic images. We design an effective segmentation pipeline, and explore several pre-training methods to initialize the features extractor, highlighting how different procedures lead the Convolutional Neural Network (CNN) to focus on different features. An encoder-decoder segmentation CNN is employed to take advantage of each pre-trained features extractor. Experimental results reveal how multiple initialization strategies can be exploited, by means of an ensemble method, to obtain state-of-the-art skin lesion segmentation accuracy.

One DAG to Rule Them All

“…Pixels named with "X" were inside the mask in the previous iteration while pixels named with "Y" are currently inside. [1], ..., [2], [3], [4], [3,4], [5], ..., [3,4,5,7]]…”

“…D EEP Learning, and (Convolutional) Neural Networks (CNN) in general, whose growth in popularity begun in the early 2010s, have marked a shift of Computer Vision, permeating most of the academic research fields of the last decade. Thanks to their ability of learning a hierarchical representation of raw input data without relying on handcrafted features, CNNs have rapidly become a methodology of choice for analyzing medical images [1], [2], [3], [4], [5], perceiving and elaborating an interpretation of dynamic scenes [6], [7], [8], [9], [10], handwriting analysis and speech recognition [11], [12], surveillance, traffic monitoring and autonomous driving [13], [14], [15], [16], people tracking [17], [18], skeletonization [19], image synthesis [20] and so on. This became possible thanks to the increase of processing capabilities, aided by the fast development of Graphics Processing Units (GPUs), and thanks to the collection of massive amounts of datasets [16], [17], [21], [22], required during the training of the models.…”

Optimized Block-Based Algorithms to Label Connected Components on GPUs

“…Originally introduced by Rosenfeld and Pfaltz in 1966 [1], CCL has been in use for more than 50 years in multiple image processing and computer vision pipelines, including Object Tracking [2], Video Surveillance [3], Image Segmentation [4], [5], [6], Medical Imaging Applications [7], [8], [9], [10], Document Restoration [11], [12], Graph Analysis [13], [14], and Environmental Applications [15].…”

Evaluation of the Classification Accuracy of the Kidney Biopsy Direct Immunofluorescence through Convolutional Neural Networks

“…Therefore, many efforts have been devoted to finding valid regularization techniques 4-6 . Convolutional Neural Networks have been widely proved to outperform other strategies in countless computer vision tasks such as semantic segmentation, object detection, object classification and others [7][8][9] . This kind of architecture is able to process images in an extremely effective and computationally efficient manner.…”