Automated segmentation of epidermis in high-frequency ultrasound of pathological skin using a cascade of DeepLab v3+ networks and fuzzy connectedness

High-Frequency Ultrasound Dataset for Deep Learning-Based Image Quality Assessment

“…Besides of this, future improvement can include three-class analysis, other body parts and diseases, and a broader range of frequencies and HFUS machines commonly used in dermatological practice, like 33, 50, or 75 MHz. Additionally, we plan to introduce FIS output weights as the pre-processing step for previously described segmentation [17] and classification [4] frameworks to evaluate their influence on the obtained results. Moreover, it needs to be validated in clinical practice.…”

“…In recent years, there appeared in literature [2,4,10,[13][14][15] different solutions in computer-aided diagnosis (CAD) of skin in HFUS data, which target segmentation, detection, and classification of the affected areas. A robust skin layer segmentation in HFUS images was as first described by Gao et al [16], and developed by Sciolla et al [15], to finally gain Dice index of 0.919 in [17] for epidermis segmentation, and 0.934 for fetus body segmentation in embryonic mice HFUS volume image analysis. The skin tumor segmentation frameworks in HFUS data start from [8] to finally reach Dice of 0.86 for skin tumor segmentation in clinical dataset [18].…”

“…Since the development of HFUS image analysis coincided with the dynamic development of machine learning algorithms, especially in the area of deep learning, most of the newest approaches [4,10,17] utilize their benefits. The first applications of convolutional neural network (CNN) to HFUS image segmentation are described in [2,19].…”

“…The same architecture, followed by the Savitzky-Golay filter and Fourier Domain Filtering, is described in [20] for epidermis and hair follicle segmentation in optical coherence tomography images. A development of [2] is the framework described in [17], where the authors expanded the CNN-based approach by fuzzy connectedness analysis for robust epidermis segmentation in HFUS.…”

“…As we mentioned before, the HFUS image processing algorithms described in the literature [2,14,15,17] assume that the input dataset consists of preselected good quality image data. Among many possible applications, CNNs are for the first applied to reduce the analyzed dataset of HFUS to the informative part in [10].…”