Dermoscopic images allow the detailed examination of subsurface characteristics of the skin, which led to creating several substantial databases of diverse skin lesions. However, the dermoscope is not an easily accessible tool in some regions. A less expensive alternative could be acquiring medium resolution clinical macroscopic images of skin lesions. However, the limited volume of macroscopic images available, especially mobile-acquired, hinders developing a clinical mobile-based deep learning approach. In this work, we present a technique to efficiently utilize the sizable number of dermoscopic images to improve the segmentation capacity of macroscopic skin lesion images. A Cycle-Consistent Adversarial Network is used to translate the image between the two distinct domains created by the different image acquisition devices. A visual inspection was performed on several databases for qualitative evaluation of the results, based on the disappearance and appearance of intrinsic dermoscopic and macroscopic features. Moreover, the Fréchet Inception Distance was used as a quantitative metric. The quantitative segmentation results are demonstrated on the available macroscopic segmentation databases, SMARTSKINS and Dermofit Image Library, yielding test set thresholded Jaccard Index of 85.13% and 74.30%. These results establish a new state-of-the-art performance in the SMARTSKINS database.

With the increasing adoption of teledermatology, there is a need to improve the automatic organization of medical records, being dermatological image modality a key filter in this process. Although there has been considerable effort in the classification of medical imaging modalities, this has not been in the field of dermatology. Moreover, as various devices are used in teledermatological consultations, image acquisition conditions may differ. In this work, two models (VGG-16 and MobileNetV2) were used to classify dermatological images from the Portuguese National Health System according to their modality. Afterwards, four incremental learning strategies were applied to these models, namely naive, elastic weight consolidation, averaged gradient episodic memory, and experience replay, enabling their adaptation to new conditions while preserving previously acquired knowledge. The evaluation considered catastrophic forgetting, accuracy, and computational cost. The MobileNetV2 trained with the experience replay strategy, with 500 images in memory, achieved a global accuracy of 86.04% with only 0.0344 of forgetting, which is 6.98% less than the second-best strategy. Regarding efficiency, this strategy took 56 s per epoch longer than the baseline and required, on average, 4554 megabytes of RAM during training. Promising results were achieved, proving the effectiveness of the proposed approach.

Teledermatology has developed rapidly in recent years and is nowadays an essential tool for early diagnosis. In this work, we aim to improve existing Teledermatology processes for skin lesion diagnosis by developing a deep learning approach for risk prioritization with a dataset of retrospective data from referral requests of the Portuguese National Health System. Given the high complexity of this task, we propose a new prioritization pipeline guided and inspired by domain knowledge. We explored automatic lesion segmentation and tested different learning schemes, namely hierarchical classification and curriculum learning approaches, optionally including additional patient metadata. The final priority level prediction can then be obtained by combining predicted diagnosis and a baseline priority level accounting for explicit expert knowledge. In both the differential diagnosis and prioritization branches, lesion segmentation with 30% tolerance for contextual information was shown to improve classification when compared with a flat baseline model trained on original images; furthermore, the addition of patient information was not beneficial for most experiments. Curriculum learning delivered better results than a flat or hierarchical approach. The combination of diagnosis information and a knowledge map, created in collaboration with dermatologists, together with the priority achieved interesting results (best macro F1 of 43.93% for a validated test set), paving the way for new data-centric and knowledge-driven approaches.

With the ever-increasing occurrence of skin cancer, timely and accurate skin cancer detection has become clinically more imperative. A clinical mobile-based deep learning approach is a possible solution for this challenge. Nevertheless, there is a major impediment in the development of such a model: the scarce availability of labelled data acquired with mobile devices, namely macroscopic images. In this work, we present two experiments to assemble a robust deep learning model for macroscopic skin lesion segmentation and to capitalize on the sizable dermoscopic databases. In the first experiment two groups of deep learning models, U-Net based and DeepLab based, were created and tested exclusively in the available macroscopic images. In the second experiment, the possibility of transferring knowledge between the domains was tested. To accomplish this, the selected model was retrained in the dermoscopic images and, subsequently, fine-tuned with the macroscopic images. The best model implemented in the first experiment was a DeepLab based model with a MobileNetV2 as feature extractor with a width multiplier of 0.35 and optimized with the soft Dice loss. This model comprehended 0.4 million parameters and obtained a thresholded Jaccard coefficient of 72.97% and 78.51% in the Dermofit and SMARTSKINS databases, respectively. In the second experiment, with the usage of transfer learning, the performance of this model was significantly improved in the first database to 75.46% and slightly decreased to 78.04% in the second.

The ability to incrementally learn to categorize objects is a key feature for a personalized system in real-world applications. The major constraint for such scenario relies on the catastrophic forgetting problem, which negatively impacts the performance of the models on previously learned representations. In this work, we developed an equipment classification model to be deployed on edge devices by applying regularization and memory-based class-incremental strategies, such that it can detect new classes while preserving its ability to detect previously known classes, mitigating the forgetting phenomenon. The strategies were tested on three datasets: CIFAR100 to validate the implementation, Stanford Dogs to ensure the reliability of the results as it is a more representative dataset, and SINATRA, which is the work's industrial dataset for equipment recognition. Experimental results on these datasets show that the Experience Replay strategy performed better. For the SINATRA dataset, average accuracy values of 95.57% and of 100% were achieved for Águas e Energias do Porto and Plastaze subsets, respectively. The outcomes of this work proved that by retaining only a limited number of exemplars from old classes, it is possible to update a pre-existing system to classify new devices in a shorter period and avoid catastrophic forgetting.