DermoExpert: Skin lesion classification using a hybrid convolutional neural network through segmentation, transfer learning, and augmentation

Background and ObjectiveAlthough automated Skin Lesion Classification (SLC) is a crucial integral step in computeraided diagnosis, it remains challenging due to inconsistency in textures, colors, indistinguishable boundaries, and shapes.MethodsThis article proposes an automated dermoscopic SLC framework named Dermoscopic Expert (DermoExpert). The DermoExpert consists of preprocessing and hybrid Convolutional Neural Network (hybrid-CNN), leveraging a transfer learning strategy. The proposed hybrid-CNN classifier has three different feature extractor modules taking the same input images, which are fused to achieve better-depth feature maps of the corresponding lesion. Those unique and fused feature maps are classified using different fully connected layers, which are then ensembled to predict the lesion class. We apply lesion segmentation, augmentation, and class rebalancing in the proposed preprocessing. We have also employed geometry- and intensity-based augmentations and class rebalancing by penalizing the majority class’s loss and combining additional images to the minority classes to enhance lesion recognition outcomes. Moreover, we leverage the knowledge from a pre-trained model to build a generic classifier, although small datasets are being used. In the end, we design and implement a web application by deploying the weights of our DermoExpert for automatic lesion recognition.ResultsWe evaluate our DermoExpert on the ISIC-2016, ISIC-2017, and ISIC-2018 datasets, where the DermoExpert has achieved the area under the receiver operating characteristic curve (AUC) of 0.96, 0.95, and 0.97, respectively. The experimental results defeat the recent state-of-the-art by the margins of 10.0 % and 2.0 % respectively for the ISIC-2016 and ISIC-2017 datasets in terms of AUC. The DermoExpert also outperforms by a border of 3.0 % for the ISIC-2018 dataset concerning a balanced accuracy.ConclusionSince our framework can provide better-classification outcomes on three different test datasets, it can lead to better-recognition of melanoma to assist dermatologists. Our source code and segmented masks for the ISIC-2018 dataset will be publicly available for further improvements.

Functional impacts of the ubiquitin–proteasome system on DNA damage recognition in global genome nucleotide excision repair

The ubiquitin–proteasome system (UPS) plays crucial roles in regulation of various biological processes, including DNA repair. In mammalian global genome nucleotide excision repair (GG-NER), activation of the DDB2-associated ubiquitin ligase upon UV-induced DNA damage is necessary for efficient recognition of lesions. To date, however, the precise roles of UPS in GG-NER remain incompletely understood. Here, we show that the proteasome subunit PSMD14 and the UPS shuttle factor RAD23B can be recruited to sites with UV-induced photolesions even in the absence of XPC, suggesting that proteolysis occurs at DNA damage sites. Unexpectedly, sustained inhibition of proteasome activity results in aggregation of PSMD14 (presumably with other proteasome components) at the periphery of nucleoli, by which DDB2 is immobilized and sequestered from its lesion recognition functions. Although depletion of PSMD14 alleviates such DDB2 immobilization induced by proteasome inhibitors, recruitment of DDB2 to DNA damage sites is then severely compromised in the absence of PSMD14. Because all of these proteasome dysfunctions selectively impair removal of cyclobutane pyrimidine dimers, but not (6–4) photoproducts, our results indicate that the functional integrity of the proteasome is essential for the DDB2-mediated lesion recognition sub-pathway, but not for GG-NER initiated through direct lesion recognition by XPC.

The trajectory of intrahelical lesion recognition and extrusion by the human 8-oxoguanine DNA glycosylase

Efficient search for DNA damage embedded in vast expanses of the DNA genome presents one of the greatest challenges to DNA repair enzymes. We report here crystal structures of human 8-oxoguanine (oxoG) DNA glycosylase, hOGG1, that interact with the DNA containing the damaged base oxoG and the normal base G while they are nested in the DNA helical stack. The structures reveal that hOGG1 engages the DNA using different protein-DNA contacts from those observed in the previously determined lesion recognition complex and other hOGG1-DNA complexes. By applying molecular dynamics simulations, we have determined the pathways taken by the lesion and normal bases when extruded from the DNA helix and their associated free energy profiles. These results reveal how the human oxoG DNA glycosylase hOGG1 locates the lesions inside the DNA helix and facilitates their extrusion for repair.