Integrating Patient Data Into Skin Cancer Classification Using Convolutional Neural Networks: Systematic Review.

BACKGROUND: Recent years have been witnessing a substantial improvement in the accuracy of skin cancer classification using convolutional neural networks (CNNs). CNNs perform on par with or better than dermatologists with respect to the classification tasks of single images. However, in clinical practice, dermatologists also use other patient data beyond the visual aspects present in a digitized image, further increasing their diagnostic accuracy. Several pilot studies have recently investigated the effects of integrating different subtypes of patient data into CNN-based skin cancer classifiers. OBJECTIVE: This systematic review focuses on the current research investigating the impact of merging information from image features and patient data on the performance of CNN-based skin cancer image classification. This study aims to explore the potential in this field of research by evaluating the types of patient data used, the ways in which the nonimage data are encoded and merged with the image features, and the impact of the integration on the classifier performance. METHODS: Google Scholar, PubMed, MEDLINE, and ScienceDirect were screened for peer-reviewed studies published in English that dealt with the integration of patient data within a CNN-based skin cancer classification. The search terms skin cancer classification, convolutional neural network(s), deep learning, lesions, melanoma, metadata, clinical information, and patient data were combined. RESULTS: A total of 11 publications fulfilled the inclusion criteria. All of them reported an overall improvement in different skin lesion classification tasks with patient data integration. The most commonly used patient data were age, sex, and lesion location. The patient data were mostly one-hot encoded. There were differences in the complexity that the encoded patient data were processed with regarding deep learning methods before and after fusing them with the image features for a combined classifier. CONCLUSIONS: This study indicates the potential benefits of integrating patient data into CNN-based diagnostic algorithms. However, how exactly the individual patient data enhance classification performance, especially in the case of multiclass classification problems, is still unclear. Moreover, a substantial fraction of patient data used by dermatologists remains to be analyzed in the context of CNN-based skin cancer classification. Further exploratory analyses in this promising field may optimize patient data integration into CNN-based skin cancer diagnostics for patients' benefits.

Clinical characteristics and therapy response in unresectable melanoma patients stage IIIB-IIID with in-transit and satellite metastases.

INTRODUCTION: Cutaneous melanoma is notorious for the development of in-transit metastases (ITM). For unknown biological reasons, ITM remain the leading tumour manifestation without progression to distant sites in some patients. METHODS: In total, 191 patients with initially unresectable stage III ITM and satellite metastases from 16 skin cancer centres were retrospectively evaluated for their tumour characteristics, survival and therapy response. Three groups according to disease kinetics (no distant progress, slow (>6 months) and fast (<6 months) distant progression) were analysed separately. RESULTS: Median follow-up time was 30.5 (range 0.8-154.0) months from unresectable ITM. Progression to stage IV was observed in 56.5% of cases. Patients without distant metastasis were more often female, older (>70 years) and presented as stage III with lymph node or ITM at initial diagnosis in 45.7% of cases. Melanoma located on the leg had a significantly better overall survival (OS) from time of initial diagnosis compared to non-leg localised primaries (hazard ratio [HR] = 0.61, 95% confidence interval [CI] 0.40-0.91; p = 0.017), but not from diagnosis of unresectable stage III (HR = 0.67, 95% CI 0.45-1.02; p = 0.06). Forty percent of patients received local therapy for satellite and ITM. Overall response rate (ORR) to all local first-line treatments was 38%; disease control rate (DCR) was 49%. In total, 72.3% of patients received systemic therapy for unresectable stage IIIB-D. ORR for targeted therapy (n = 19) was highest with 63.2% and DCR was 84.2% compared to an ORR of 31.4% and a DCR of 54.3% in PD-1 treated patients (n = 70). Patients receiving PD-1 and intralesional talimogene laherparepvec (n = 12) had an ORR of 41.7% and a DCR of 75%. CONCLUSION: Patients with unresectable ITM and without distant progression are more often female, older, and have a primary on the leg. Response to PD-1 inhibitors in this cohort was lower than expected, but further investigation is required to elucidate the biology of ITM development and the interplay with the immune system.

Effects of Label Noise on Deep Learning-Based Skin Cancer Classification.

Recent studies have shown that deep learning is capable of classifying dermatoscopic images at least as well as dermatologists. However, many studies in skin cancer classification utilize non-biopsy-verified training images. This imperfect ground truth introduces a systematic error, but the effects on classifier performance are currently unknown. Here, we systematically examine the effects of label noise by training and evaluating convolutional neural networks (CNN) with 804 images of melanoma and nevi labeled either by dermatologists or by biopsy. The CNNs are evaluated on a test set of 384 images by means of 4-fold cross validation comparing the outputs with either the corresponding dermatological or the biopsy-verified diagnosis. With identical ground truths of training and test labels, high accuracies with 75.03% (95% CI: 74.39-75.66%) for dermatological and 73.80% (95% CI: 73.10-74.51%) for biopsy-verified labels can be achieved. However, if the CNN is trained and tested with different ground truths, accuracy drops significantly to 64.53% (95% CI: 63.12-65.94%, p < 0.01) on a non-biopsy-verified and to 64.24% (95% CI: 62.66-65.83%, p < 0.01) on a biopsy-verified test set. In conclusion, deep learning methods for skin cancer classification are highly sensitive to label noise and future work should use biopsy-verified training images to mitigate this problem.

Treatment of unresectable stage IV metastatic melanoma with aviscumine after anti-neoplastic treatment failure: a phase II, multi-centre study.

BACKGROUND: Aviscumine, a recombinant plant protein, is an immune modulator that induces ribotoxic stress at the 28S ribosomal RNA subunit. In this way cytokine release and T-cell responses are enhanced. This phase II trial was conducted to test the efficacy and safety of aviscumine in patients with systemically pre-treated metastatic melanoma stage IV. METHODS: A total of 32 patients with progressive stage IV melanoma after failure of standard therapy were enrolled onto a single-arm, multi-centre, open-label, phase II trial. All patients had an ECOG performance status of 0 or 1. Patients received 350 ng aviscumine twice weekly by subcutaneous injection until progression. The primary end points were progression-free survival (PFS) and overall survival (OS). Safety was assessed as adverse events (AEs). Tumor response was assessed every eight weeks and survival of patients was followed up to one year after the end of therapy. Thirty one patients (intent-to-treat population (ITT)) were assessed for efficacy; safety was assessed in the whole population. RESULTS: One patient achieved a partial response (PR) and 10 patients showed stable disease/no change (SD). The median progression-free survival (mPFS) was 63 days (95% CI 57-85) and median overall survival (mOS) was 335 days (95% CI 210-604). In total 210 treatment-emergent adverse events were recorded. Grade 1 or 2 AEs occurred in 72% of patients and were mostly application-site effects such as pruritus Grade 3-4 treatment-emergent drug-related adverse events occurred in 9% of patients. CONCLUSION: These results suggest that aviscumine may have a clinical impact in patients with previously treated metastatic melanoma and provide rationale for further clinical evaluation of this agent. In the light of effective new immune checkpoint blockers it might be a candidate for combinations with these agents. TRIAL REGISTRATION: ClinicalTrials.gov: NCT00658437.

Phase II trial (BREAK-2) of the BRAF inhibitor dabrafenib (GSK2118436) in patients with metastatic melanoma.

PURPOSE: Dabrafenib (GSK2118436) is a potent inhibitor of mutated BRAF kinase. Our multicenter, single-arm, phase II study assessed the safety and clinical activity of dabrafenib in BRAF(V600E/K) mutation-positive metastatic melanoma (mut(+) MM). PATIENTS AND METHODS: Histologically confirmed patients with stage IV BRAF(V600E/K) mut(+) MM received oral dabrafenib 150 mg twice daily until disease progression, death, or unacceptable adverse events (AEs). The primary end point was investigator-assessed overall response rate in BRAF(V600E) mut(+) MM patients. Secondary end points included progression-free survival (PFS) and overall survival (OS). Exploratory objectives included the comparison of BRAF mutation status between tumor-specific circulating cell-free DNA (cfDNA) and tumor tissue, and the evaluation of cfDNA as a predictor of clinical outcome. RESULTS: Seventy-six patients with BRAF(V600E) and 16 patients with BRAF(V600K) mut(+) MM were enrolled onto the study. In the BRAF(V600E) group, 45 patients (59%) had a confirmed response (95% CI, 48.2 to 70.3), including five patients (7%) with complete responses. Two patients (13%) with BRAF(V600K) mut(+) MM had a confirmed partial response (95% CI, 0 to 28.7). In the BRAF(V600E) and BRAF(V600K) groups, median PFS was 6.3 months and 4.5 months, and median OS was 13.1 months and 12.9 months, respectively. The most common AEs were arthralgia (33%), hyperkeratosis (27%), and pyrexia (24%). Overall, 25 patients (27%) experienced a serious AE and nine patients (10%) had squamous cell carcinoma. Baseline cfDNA levels predicted response rate and PFS in BRAF(V600E) mut(+) MM patients. CONCLUSION: Dabrafenib was well tolerated and clinically active in patients with BRAF(V600E/K) mut(+) MM. cfDNA may be a useful prognostic and response marker in future studies.

The price of tumor control: an analysis of rare side effects of anti-CTLA-4 therapy in metastatic melanoma from the ipilimumab network.

BACKGROUND: Ipilimumab, a cytotoxic T-lymphocyte antigen-4 (CTLA-4) blocking antibody, has been approved for the treatment of metastatic melanoma and induces adverse events (AE) in up to 64% of patients. Treatment algorithms for the management of common ipilimumab-induced AEs have lead to a reduction of morbidity, e.g. due to bowel perforations. However, the spectrum of less common AEs is expanding as ipilimumab is increasingly applied. Stringent recognition and management of AEs will reduce drug-induced morbidity and costs, and thus, positively impact the cost-benefit ratio of the drug. To facilitate timely identification and adequate management data on rare AEs were analyzed at 19 skin cancer centers. METHODS AND FINDINGS: Patient files (n = 752) were screened for rare ipilimumab-associated AEs. A total of 120 AEs, some of which were life-threatening or even fatal, were reported and summarized by organ system describing the most instructive cases in detail. Previously unreported AEs like drug rash with eosinophilia and systemic symptoms (DRESS), granulomatous inflammation of the central nervous system, and aseptic meningitis, were documented. Obstacles included patients delay in reporting symptoms and the differentiation of steroid-induced from ipilimumab-induced AEs under steroid treatment. Importantly, response rate was high in this patient population with tumor regression in 30.9% and a tumor control rate of 61.8% in stage IV melanoma patients despite the fact that some patients received only two of four recommended ipilimumab infusions. This suggests that ipilimumab-induced antitumor responses can have an early onset and that severe autoimmune reactions may reflect overtreatment. CONCLUSION: The wide spectrum of ipilimumab-induced AEs demands doctor and patient awareness to reduce morbidity and treatment costs and true ipilimumab success is dictated by both objective tumor responses and controlling severe side effects.

GABAB receptor antibodies in paraneoplastic cerebellar ataxia.

Autoantibodies to the gamma-aminobutyric acid-B (GABAB) receptor were recently described in patients with limbic encephalitis presenting with early or prominent seizures. We report on a 64-year-old man with malignant melanoma who during adjuvant therapy with interferon (IFN)-alpha developed cerebellar ataxia. Indirect immunofluorescence on brain tissue sections revealed high-titer (1:20,000) IgG1 serum autoantibodies to the cerebellar molecular and granular layer, which were confirmed to be directed against GABAB receptor in a cell-based assay. This case highlights cerebellar ataxia in the absence of seizures as a clinical manifestation of GABAB receptor autoimmunity and extends the spectrum of tumors underlying this condition to malignant melanoma. IFN-alpha therapy may have contributed to the development of autoimmunity in this patient.

Dabrafenib in patients with Val600Glu or Val600Lys BRAF-mutant melanoma metastatic to the brain (BREAK-MB): a multicentre, open-label, phase 2 trial.

BACKGROUND: Brain metastases are common in patients with metastatic melanoma and median overall survival from their diagnosis is typically 17-22 weeks. We assessed dabrafenib in patients with Val600Glu or Val600Lys BRAF-mutant melanoma metastatic to the brain. METHODS: We undertook a multicentre, open-label, phase 2 trial in 24 centres in six countries. We enrolled patients with histologically confirmed Val600Glu or Val600Lys BRAF-mutant melanoma and at least one asymptomatic brain metastasis (≥5 mm and ≤40 mm in diameter). Eligible patients were aged 18 years or older, had an Eastern Cooperative Oncology Group performance status of 0 or 1, and had adequate organ function. Patients were split into two cohorts: those in cohort A had not received previous local treatment for brain metastases and those in cohort B had progressive brain metastases after previous local treatments. Patients received 150 mg oral dabrafenib twice a day until disease progression, death, or unacceptable adverse events. The primary endpoint was the proportion of patients with Val600Glu BRAF-mutant melanoma who achieved an overall intracranial response, which was defined as a complete response or partial response assessed with a modified form of Response Evaluation Criteria in Solid Tumors (RECIST 1.1). We included patients who received at least one dose of dabrafenib in efficacy and safety analyses. This study is registered with ClinicalTrials.gov, number NCT01266967. FINDINGS: Between Feb 2, 2011, and Aug 5, 2011, we enrolled 172 patients: 89 (52%) in cohort A and 83 (48%) in cohort B. 139 (81%) had Val600Glu BRAF-mutant melanoma. 29 (39·2%, 95% CI 28·0-51·2) of 74 patients with Val600Glu BRAF-mutant melanoma in cohort A achieved an overall intracranial response, as did 20 (30·8%, 19·9-43·4) of 65 in cohort B. One (6·7%, 0·2-31·9) of 15 patients with Val600Lys BRAF-mutant melanoma achieved an overall intracranial response in cohort A, as did four (22·2%, 6·4-47·6) of 18 such patients in cohort B. Treatment-related adverse events of grade 3 or worse occurred in 38 (22%) patients. Eleven (6%) patients developed squamous-cell carcinoma (five [6%] patients in cohort A, of whom one also had keratoacanthoma; six [7%] in cohort B). Four grade 4 treatment-related adverse events occurred in cohort A: one blood amylase increase, one convulsion, one lipase increase, and one neutropenia. Two grade 4 events occurred in cohort B: one agranulocytosis and one intracranial haemorrhage. 51 (30%) patients had a serious adverse event. The three most frequent serious adverse events were pyrexia (ten [6%] patients), intracranial haemorrhage (ten [6%]; one treatment-related), and squamous-cell carcinoma (11 [6%]). INTERPRETATION: Dabrafenib has activity and an acceptable safety profile in patients with Val600Glu BRAF-mutant melanoma and brain metastases irrespective of whether they are untreated or have been previously treated but have progressed. FUNDING: GlaxoSmithKline.

Brain morphology alterations in the basal ganglia and the hypothalamus following prenatal exposure to antiepileptic drugs.

If humans are exposed prenatally to antiepileptic drugs (AEDs), cognitive impairment may be the consequence. Driven by results of experimental work showing that AEDs may induce neuronal death in the developing rodent brain, we wanted to explore whether prenatal exposure to AEDs (PAE) may result in structural changes in the human brain. For this purpose we investigated a group of healthy young adults with PAE and a group of age-matched unexposed healthy controls by magnetic resonance imaging (MRI) of the brain. Local differences in cerebral morphology associated with PAE were analysed in volumetric MRI data by use of voxelwise comparisons of grey and white matter images. Significant regional decreases of grey matter volumes were found in PAE subjects in the area of the lentiform nucleus, including both pallidum and putamen bilaterally, and the hypothalamus. No significant regional differences in white matter volumes were found. We conclude that PAE causes subtle morphological changes in grey matter of the human brain which are conform with lower cell numbers in the basal ganglia and the hypothalamus.