Search

Despite recent progress in artificial intelligence and computer vision, the lack of data for acral melanoma, a particular type of skin disease under nails, makes it difficult to develop its automatic visual diagnosis system. This paper introduces a large data set of dermoscopic images for acral melanoma, which is annotated by senior dermatologists. It contains 6,066 images of two categories: subungual hematomas and other acral melanoma symptoms of malignant tendency (i.e., just bleeding under nails or more critical disorder that requires treatment). We hope this benchmark data set will encourage further research on acral melanoma recognition and will continue to maintain this data set to better serve it. We address the classification task using various computer vision algorithms from conventional techniques to cutting edge deep learning. Currently, we achieve 0.928 accuracy.

An accurate diagnosis of tinea unguium is necessary for the selection of antimycotics and successful treatment. To rapidly and accurately identify the aetiological agents causing tinea unguium, we improved upon the conventional boiling method for DNA extraction and developed a novel real-time PCR detection system that includes two assays. The two assays, based on the amplification of ribosomal internal transcribed spacer regions and 28S rDNA, were designed to detect pan-dermatophyte and Trichophyton rubrum, respectively. The analytical sensitivities of both assays permitted the detection of ten copies of plasmid DNA template. The analytical specificity of the detection system was confirmed using 11 dermatophyte strains and 25 non-dermatophyte strains. In total, 165 nail specimens were examined by microscopy, culture, conventional PCR, and the novel real-time PCR method. Real-time PCR gave positive results in 47.3 % of the specimens (78), a rate exceeding those obtained using microscopy (72, 43.6 %), conventional PCR (69, 41.8 %), and culture (49, 29.7 %). All conventional PCR-positive specimens were detected by real-time PCR, and real-time PCR detected nine specimens that were missed by conventional PCR. The results from latent class analysis, and further calculations, showed that real-time PCR was the most sensitive method, but the diagnostic specificity of the four approaches was equivalent. In particular, molecular approaches may be more effective than microscopy and culture when the clinical symptoms of tinea unguium are not evident.

A 64-year-old man presented with widespread, erythematous-to-violaceous papules that were distributed symmetrically on his trunk, arms, and legs. Skin biopsy revealed palisading granulomatous inflammation that surrounded degenerating collagen within the dermis.