Deep-learning-based, computer-aided classifier developed with a small dataset of clinical images surpasses board-certified dermatologists in skin tumour diagnosis

BACKGROUND

Application of deep-learning technology to skin cancer classification can potentially improve the sensitivity and specificity of skin cancer screening, but the number of training images required for such a system is thought to be extremely large.

OBJECTIVES

To determine whether deep-learning technology could be used to develop an efficient skin cancer classification system with a relatively small dataset of clinical images.

METHODS

A deep convolutional neural network (DCNN) was trained using a dataset of 4867 clinical images obtained from 1842 patients diagnosed with skin tumours at the University of Tsukuba Hospital from 2003 to 2016. The images consisted of 14 diagnoses, including both malignant and benign conditions. Its performance was tested against 13 board-certified dermatologists and nine dermatology trainees.

RESULTS

The overall classification accuracy of the trained DCNN was 76·5%. The DCNN achieved 96·3% sensitivity (correctly classified malignant as malignant) and 89·5% specificity (correctly classified benign as benign). Although the accuracy of malignant or benign classification by the board-certified dermatologists was statistically higher than that of the dermatology trainees (85·3% ± 3·7% and 74·4% ± 6·8%, P < 0·01), the DCNN achieved even greater accuracy, as high as 92·4% ± 2·1% (P < 0·001).

CONCLUSIONS

We have developed an efficient skin tumour classifier using a DCNN trained on a relatively small dataset. The DCNN classified images of skin tumours more accurately than board-certified dermatologists. Collectively, the current system may have capabilities for screening purposes in general medical practice, particularly because it requires only a single clinical image for classification.

New clues to the nature of immunoglobulin G4-related disease: a retrospective Japanese multicenter study of baseline clinical features of 334 cases

Background

The aim was to further characterize immunoglobulin G4-related disease (IgG4-RD) by a large-scale multicenter study of its clinical and laboratory features conducted by multidisciplinary physicians of IgG4-RD in Japan.

Methods

Various specialists retrospectively evaluated IgG4-RD patients diagnosed between 1996 and 2015 in five hospitals by analyzing their baseline clinical features, laboratory, imaging, and pathological test findings, and treatment.

Results

Of the 334 patients listed, 205 were male and median age at diagnosis was 65 years. The mean number of organs involved was 3.2 at diagnosis. The most frequently affected organs were the salivary glands, followed by the lacrimal glands, lymph nodes, pancreas, retroperitoneum/periaorta, kidneys, and lungs. The mean serum level of IgG4 was 755 mg/dl, and more than 95% of patients had elevated serum IgG4 levels. The median serum level of C-reactive protein (CRP) was 0.1 mg/dl and the level was less than 1 mg/dl in 90% of patients. A total of 34.7% of patients had low serum levels of C3. Serum levels of C3 and non-IgG4 IgG, calculated as the total IgG minus IgG4, showed an inverse correlation in patients with kidney lesions, while serum IgG4 levels were not correlated with serum C3 levels. Corticosteroid was administered in 78.0% of patients, and was effective in all.

Conclusions

The serum CRP level is generally low and the serum IgG4 level is elevated in most Japanese IgG4-RD patients, in contrast to western patients. These original findings suggest that these two parameters in IgG4-RD differ in some interesting ways from those hitherto reported in western populations. Additional studies, especially international comparative ones, are needed to elucidate the extent and significance of these differences between populations. Attention will also have to be paid to whether the existence of such differences requires consideration when devising international classification criteria.

Electronic supplementary material

The online version of this article (doi:10.1186/s13075-017-1467-x) contains supplementary material, which is available to authorized users.