Deep learning-based classification of dermatological lesions given a limited amount of labeled data

BACKGROUND

Artificial intelligence (AI) techniques are promising in early diagnosis of skin diseases. However, a precondition for their success is the access to large-scaled annotated data. Until now, obtaining this data has only been feasible with very high personnel and financial resources.

OBJECTIVES

The aim of this study was to overcome the obstacle caused by the scarcity of labeled data.

METHODS

To simulate the scenario of label shortage, we discarded a proportion of labels of the training set. The training set consisted of both labeled and unlabeled images. We then leveraged a self-supervised learning technique to pre-train the AI model on the unlabeled images. Next, we fine-tuned the pre-trained model on the labeled images.

RESULTS

When the images in the training dataset were fully labeled, the self-supervised pre-trained model achieved 95.7% of accuracy, 91.7% of precision and 90.7% of sensitivity. When only 10% of the data was labeled, the model could still yield 87.7% of accuracy, 81.7% of precision and 68.6% of sensitivity. In addition, we also empirically verified that the AI model and dermatologists are consistent in visually inspecting the skin images.

CONCLUSIONS

The experimental results demonstrate the great potential of the self-supervised learning in alleviating the scarcity of annotated data.

Ex vivo confocal laser scanning microscopy: A diagnostic technique for easy real-time evaluation of benign and malignant skin tumours

Ex vivo confocal laser scanning microscopy (ex vivo CLSM) is a novel diagnostic tool for a quick bedside evaluation of freshly excised tissue, comparable to histology. We aimed to assess the sensitivity and specificity of ex vivo CLSM in detecting malignant features, to validate its reliability in identifying various skin tumours based on a combination of confocal features and to evaluate the digital staining mode (DS). One-hundred twenty freshly excised skin samples from 91 patients were evaluated. Each lesion was screened for the presence of 23 predefined confocal criteria with ex vivo CLSM, followed by a histopathological examination. The diagnostic agreement between ex vivo CLSM and histology was 89.2%. The diagnostic accuracy of ex vivo CLSM in detecting malignancy reached a sensitivity of 98% and a specificity of 76%. Ex vivo CLSM enabled a rapid identification of the most common skin tumours, the tumour dignity and cytological features. The DS demonstrated a close resemblance to conventional histopathology.

FusionM4Net: A multi-stage multi-modal learning algorithm for multi-label skin lesion classification

Skin disease is one of the most common diseases in the world. Deep learning-based methods have achieved excellent skin lesion recognition performance, most of which are based on only dermoscopy images. In recent works that use multi-modality data (patient's meta-data, clinical images, and dermoscopy images), the methods adopt a one-stage fusion approach and only optimize the information fusion at the feature level. These methods do not use information fusion at the decision level and thus cannot fully use the data of all modalities. This work proposes a novel two-stage multi-modal learning algorithm (FusionM4Net) for multi-label skin diseases classification. At the first stage, we construct a FusionNet, which exploits and integrates the representation of clinical and dermoscopy images at the feature level, and then uses a Fusion Scheme 1 to conduct the information fusion at the decision level. At the second stage, to further incorporate the patient's meta-data, we propose a Fusion Scheme 2, which integrates the multi-label predictive information from the first stage and patient's meta-data information to train an SVM cluster. The final diagnosis is formed by the fusion of the predictions from the first and second stages. Our algorithm was evaluated on the seven-point checklist dataset, a well-established multi-modality multi-label skin disease dataset. Without using the patient's meta-data, the proposed FusionM4Net's first stage (FusionM4Net-FS) achieved an average accuracy of 75.7% for multi-classification tasks and 74.9% for diagnostic tasks, which is more accurate than other state-of-the-art methods. By further fusing the patient's meta-data at FusionM4Net's second stage (FusionM4Net-SS), the entire FusionM4Net finally boosts the average accuracy to 77.0% and the diagnostic accuracy to 78.5%, which indicates its robust and excellent classification performance on the label-imbalanced dataset. The corresponding code is available at: https://github.com/pixixiaonaogou/MLSDR.

Immunofluorescence and histopathological assessment using ex vivo confocal laser scanning microscopy in lichen planus

Ex vivo confocal laser scanning microscopy (CLSM) provides rapid, high-resolution imaging, fluorescence detection and digital haematoxylin-eosin (H&E)-like staining. We aimed to assess the performance of ex vivo CLSM in identifying histomorphology and immunoreactivity in lichen planus (LP) and comparing its accuracy with conventional histopathology and direct immunofluorescence (DIF). Thirty-three sections of 17 LP patients stained with acridine orange (AO) and FITC-labelled anti-fibrinogen antibody and 21 control samples stained with AO were examined using ex vivo CLSM. Ex vivo CLSM was in perfect agreement with conventional histopathology in identifying interface dermatitis, vacuolar degeneration and band-like infiltration. ROC analysis showed that the presence of vacuolar degeneration, interface dermatitis and band-like infiltration was useful to distinguish LP sections from controls (p < .0001). The detection rates of fibrinogen deposition using DIF and in conclusion ex vivo CLSM were 93.8% and 62.5%, respectively. ex vivo CLSM enables histopathological and immunofluorescence examination in LP with the advantage of digital H&E-like staining.

Ex vivo Confocal Laser Scanning Microscopy: A Potential New Diagnostic Imaging Tool in Onychomycosis Comparable With Gold Standard Techniques

Ex vivo confocal laser scanning microscopy (CLSM) is an innovative imaging tool that enables real-time examination of specimens and may be used in evaluating fungal infections. We aimed to assess the applicability of ex vivo CLSM in the diagnosis of onychomycosis by comparing results to those obtained by histopathology, potassium hydroxide (KOH) examination, and fungal culture. In this prospective study, 57 patients with the clinical diagnosis of distal nail fungal infection were examined and compared using all four of the above-mentioned diagnostic tools in terms of sensitivity, positive and negative predictive value. Ex vivo CLSM showed the highest sensitivity, followed by KOH examination, histopathology and fungal culture. Regarding positive and negative predictive values, ex vivo CLSM was superior and showed even higher sensitivity than the combined gold standard comprised of KOH examination, fungal culture or histopathology.

Recurrence of Pemphigus Vulgaris Under Nivolumab Therapy

For many types of cancer, immune checkpoint inhibitors have proven to be a highly effective treatment. The monoclonal anti-PD-1 antibody nivolumab stimulates the immune system and is one of the newest treatment options for non-small cell lung cancer. In doing so, immune checkpoint inhibitors can trigger many skin lesions that have not yet been completely investigated in their entirety. In this case report, pemphigus vulgaris is identified as a potential adverse event that occurs under the treatment with nivolumab. In addition to the standard methods, we examined our patient's samples with ex vivo confocal laser scanning microscopy. This is a new and innovative diagnostic method that uses vertical scanning to provide fast, high-resolution imaging of freshly excised tissue, even using fluorescently labeled antibodies.

Ex vivo confocal laser scanning microscopy: An innovative method for direct immunofluorescence of cutaneous vasculitis

Ex vivo confocal laser scanning microscopy (ex vivo CLSM) offers an innovative diagnostic approach through vertical scanning of skin samples with a resolution close to conventional histology. In addition, it enables fluorescence detection in tissues. We aimed to assess the applicability of ex vivo CLSM in the detection of vascular immune complexes in cutaneous vasculitis and to compare its diagnostic accuracy with direct immunofluorescence (DIF) microscopy. Eighty-two sections of 49 vasculitis patients with relevant DIF microscopy findings were examined using ex vivo CLSM following staining with fluorescent-labeled IgG, IgM, IgA, C3 and fibrinogen antibodies. DIF microscopy showed immunoreactivity of vessels with IgG, IgM, IgA, C3 and Fibrinogen in 2.0%, 49.9%, 12.2%, 59.2% and 44.9% of the patients, respectively. Ex vivo CLSM detected positive vessels with the same antibodies in 2.0%, 38.8%, 8.2%, 42.9% and 36.7% of the patients, respectively. The detection rate of positive superficial dermal vessels was significantly higher in DIF microscopy as compared to ex vivo CLSM (P < .05). Whereas, ex vivo CLSM identified positive deep dermal vessels more frequently compared to DIF microscopy. In conclusion, ex vivo CLSM could identify specific binding of the antibodies in vessels and showed a comparable performance to conventional DIF microscopy in diagnosing vasculitis.

Simple 3-criteria-based ex vivo confocal diagnosis of basal cell carcinoma

BACKGROUND

Fast and simple microscopic evaluation of basal cell carcinoma (BCC) together with its subtype determination would accelerate diagnostic and therapeutic procedures in dermatology including Mohs surgery.

OBJECTIVES

Assessing whether simplified 3-criteria-based ex vivo confocal microscopic (CM) examination can reliably predict BCC diagnosis and subtype. Analyzing interobserver agreement between expert and novice examiner.

METHODS

CM images of 235 skin samples from 150 patients were prospectively evaluated by 2 blinded examiners for the presence of 3 predefined BCC criteria namely presence of tumor mass, peripheral palisading and clefting.

RESULTS

Out of 235 skin samples 116 showed histological presence of BCC, confocally expert diagnosed a BCC in 110 and novice examiner in 107 samples. The overall sensitivity and specificity of detecting residual BCC was 96.6% and 98.7%, respectively. Confocally, examiners diagnosed correctly nodular BCC in 96.6%, respectively, 98.3%, superficial BCC in 96.8%, respectively, 93.5%, infiltrating BCC in 88.9%, respectively, 83.3% and other BCC subtype in 22.2%, respectively, 0% (expert and novice examiner, respectively).

CONCLUSION

Ex vivo CM allowed intraoperative examination of BCC based on only 3-criteria with high sensitivity and specificity, provided useful information on tumor subtype and showed that both experienced and non-experienced examiners may use this diagnostic approach with excellent results.

Ex vivo confocal microscopy features of cutaneous squamous cell carcinoma

BACKGROUND

Rapid microscopic evaluation of cutaneous squamous cell carcinoma (SCC), its grade of differentiation and level of invasiveness would enable better management of patients' therapy.

OBJECTIVES

Analyzing specific ex vivo confocal microscopy criteria whether they can predict diagnosis of invasive SCC vs carcinoma in situ and poorly differentiated or undifferentiated vs well and moderately differentiated SCC.

METHODS

Ex vivo confocal images of 102 SCCs in 57 patients were evaluated immediately after excision for the presence of predefined criteria based on confocal and histological knowledge.

RESULTS

In histopathological examination, 30 SCCs were in situ and 72 invasive. Of these, 29 invasive SCC tumors were well, 19 moderately, 15 poorly differentiated and 9 undifferentiated. χ² analysis demonstrated that presence of erosion/ulceration, plump bright or speckled cells in dermis, keratin pearls and peritumoral inflammatory infiltrate correlated with diagnosis of invasive SCC. Erosion/ulceration and peritumoral inflammatory infiltrate were observed more frequently in poorly differentiated or undifferentiated tumors. Plump bright or speckled cells in the dermis were observed less often in well-differentiated tumors. The presence of keratin pearls was associated with well or moderately differentiated tumors.

CONCLUSION

Ex vivo CLSM allowed rapid examination of SCC and provided useful information on invasiveness and grading of the tumor.

Immunofluorescence and confocal microscopy for ex-vivo diagnosis of melanocytic and non-melanocytic skin tumors: A pilot study

BACKGROUND

Ex-vivo confocal laser scanning microscopy (ex-vivo CLSM) offers rapid examination of freshly excised tissue. During the conventional examination immunohistochemistry enables to distinguish various cell types. The possibility of immunofluorescent techniques could enhance the accuracy of the diagnosis performed by ex-vivo CLSM.

METHODS

The tissue probes from various skin tumors were stained with FITC-labeled S-100A10, Melan-A and anti-Ber-EP4 antibodies before examination with ex-vivo CLSM in the fluorescence and reflectance modes. Results were compared to negative controls and conventional histopathology. The staining protocols were evaluated by establishing a scoring system according to the signal intensity found in ex-vivo CLSM.

RESULTS

S100 immunostaining was successful in 55.6%. Dilution of 1:200 resulted in the best possible evaluation of the tumor. The best suitable protocol was protocol B (phosphate buffered saline [PBS], without blocking agent). Melan A immunostaining was positive in 66.7%, the best dilution was 1:500 and protocol B (PBS, without blocking agent) was the most suitable. Ber-EP4 immunostaining presented a signal in 85.7%, the best dilutions were 1:200 and 1:500 and protocol A (PBS, with blocking agent) showed most optimal results.

CONCLUSION

The use of fluorescent-labeled antibodies in ex-vivo CLSM is possible and could improve intraoperative diagnostics of skin tumors.

Quality of life following aesthetic plastic surgery: a prospective study

BACKGROUND

The objective of this study was the prospective evaluation of quality of life in patients undergoing aesthetic plastic surgery procedures. We examined pre- and postoperative changes in quality of life, and performed a comparison of our data with a representative random sample.

METHOD

228 patients agreed to participate in the present study. Measurements were taken preoperatively as well as 3 and 6 months postoperatively. One hundred and thirty two patients completed the three months postoperative evaluation (T1), 82 answered the 6 months follow-up evaluation (T2). The testing instrument included a standardised self-assessment test on satisfaction and quality of life (FLZ(M)), consisting of three modules: satisfaction with general life, health and appearance. Further, a postoperative complication questionnaire was used in order to evaluate the satisfaction with the surgical outcome and to estimate postoperative complications and side effects.

RESULTS

Significantly increasing values in two aspects of quality of life were found: health and appearance. Whereas the positive influence on health is persistent, there is a diminishing influence of appearance 6 months postoperatively. Although higher values for some of the individual items of the FLZ(M) modules of the norm data were found in comparison to our study group, a general preoperative lower level of quality of life of the aesthetic surgery patients could not be confirmed. Over 84% were satisfied or very satisfied with the aesthetic result. 85% would undergo the same treatment again and 94% of the patients would further recommend their operation. More than half of the study group did not report a decrease in physical fitness or reduced social contacts in the direct postoperative period.

CONCLUSION

Our study reveals that aesthetic plastic surgery increases most aspects of quality of life, especially regarding body satisfaction and health. It is very well tolerated by the patients and is therefore a recommended option.

Blood vitamin concentrations in privately owned dogs fed non-standardized commercial diets and after intake of diets with specified vitamin concentrations

The objective was to investigate in a survey study the blood vitamin concentrations in healthy dogs fed non-specified commercial complete diets and in an intervention study to determine the effects of defined dietary vitamin intakes on blood vitamin levels and hair and skin condition. Sixty-four privately owned dogs, aged from 1 to 8 years, without history of skin or coat problems were included. All animals were fed commercial complete diets with uncertain vitamin concentrations before enrolment. The animals were assigned, according to weight and gender, to four groups with graded vitamin intakes. The blood vitamin levels and skin and coat quality of the dogs were investigated at days 0 and day 122. Coat and hair condition was not influenced by the experimental diets. The retinol concentrations were reduced at the end of the experiment compared with the baseline levels, retinyl esters were not influenced. 25-Hydroxycholecalciferol decreased in all groups, alpha-tocopherol was constant or tended to decrease. Ascorbic acid, thiamine pyrophosphate and riboflavin concentrations were not affected by treatment, flavin adenine dinucleotide and pyridoxal-5'-phosphate were partially reduced on day 122. Cobalamin, pantothenate and biotin concentrations increased with higher dietary intakes, folate levels in tendency. In conclusion, this study gives a survey of blood vitamin concentrations in healthy dogs and provides a data base for the evaluation of the vitamin status in health and disease.