Background: Federated learning is a decentralized approach to machine learning; it is a training strategy that overcomes medical data privacy regulations and generalizes deep learning algorithms. Federated learning mitigates many systemic privacy risks by sharing only the model and parameters for training, without the need to export existing medical data sets. In this study, we performed ultrasound image analysis using federated learning to predict whether thyroid nodules were benign or malignant. Objective: The goal of this study was to evaluate whether the performance of federated learning was comparable with that of conventional deep learning. Methods: A total of 8457 (5375 malignant, 3082 benign) ultrasound images were collected from 6 institutions and used for federated learning and conventional deep learning. Five deep learning networks (VGG19, ResNet50, ResNext50, SE-ResNet50, and SE-ResNext50) were used. Using stratified random sampling, we selected 20% (1075 malignant, 616 benign) of the total images for internal validation. For external validation, we used 100 ultrasound images (50 malignant, 50 benign) from another institution. Results: For internal validation, the area under the receiver operating characteristic (AUROC) curve for federated learning was between 78.88% and 87.56%, and the AUROC for conventional deep learning was between 82.61% and 91.57%. For external validation, the AUROC for federated learning was between 75.20% and 86.72%, and the AUROC curve for conventional deep learning was between 73.04% and 91.04%. Conclusions: We demonstrated that the performance of federated learning using decentralized data was comparable to that of conventional deep learning using pooled data. Federated learning might be potentially useful for analyzing medical images while protecting patients’ personal information.

We investigate the potential of mobile smartphone-based multispectral imaging for the quantitative diagnosis and management of skin lesions. Recently, various mobile devices such as a smartphone have emerged as healthcare tools. They have been applied for the early diagnosis of nonmalignant and malignant skin diseases. Particularly, when they are combined with an advanced optical imaging technique such as multispectral imaging and analysis, it would be beneficial for the early diagnosis of such skin diseases and for further quantitative prognosis monitoring after treatment at home. Thus, we demonstrate here the development of a smartphone-based multispectral imaging system with high portability and its potential for mobile skin diagnosis. The results suggest that smartphone-based multispectral imaging and analysis has great potential as a healthcare tool for quantitative mobile skin diagnosis.

We demonstrate a noncontact single-beam acoustic trapping method for the quantification of the mechanical properties of a single suspended cell with label-free. Experimentally results show that the single-beam acoustic trapping force results in morphological deformation of a trapped cell. While a cancer cell was trapped in an acoustic beam focus, the morphological changes of the immobilized cell were monitored using bright-field imaging. The cell deformability was then compared with that of a trapped polystyrene microbead as a function of the applied acoustic pressure for a better understanding of the relationship between the pressure and degree of cell deformation. Cell deformation was found to become more pronounced as higher pressure levels were applied. Furthermore, to determine if this acoustic trapping method can be exploited in quantifying the cell mechanics in a suspension and in a non-contact manner, the deformability levels of breast cancer cells with different degrees of invasiveness due to acoustic trapping were compared. It was found that highly-invasive breast cancer cells exhibited greater deformability than weakly-invasive breast cancer cells. These results clearly demonstrate that the single-beam acoustic trapping technique is a promising tool for non-contact quantitative assessments of the mechanical properties of single cells in suspensions with label-free.

Extracting manufactured features such as buildings, roads, and water from aerial images is critical for urban planning, traffic management, and industrial development. Recently, convolutional neural networks (CNNs) have become a popular strategy to capture contextual features automatically. In order to train CNNs, a large training data are required, but it is not straightforward to use free-accessible data sets due to imperfect labeling. To address this issue, we make a large scale of data sets using RGB aerial images and convert them to digital maps with location information such as roads, buildings, and water from the metropolitan area of Seoul in South Korea. The numbers of training and test data are 72,400 and 9600, respectively. Based on our self-made data sets, we design a multiobject segmentation system and propose an algorithm that utilizes pyramid pooling layers (PPLs) to improve U-Net. Test results indicate that U-Net with PPLs, called UNetPPL, learn fine-grained classification maps and outperforms other algorithms of fully convolutional network and U-Net, achieving the mean intersection of union (mIOU) of 79.52 and the pixel accuracy of 87.61% for four types of objects (i.e., building, road, water, and background).

Change detection is an important task in the field of remote sensing. Various change detection methods based on convolutional neural networks (CNNs) have recently been proposed for remote sensing using satellite or aerial images. However, existing methods allow only the partial use of content information in images during change detection because they adopt simple feature similarity measurements or pixel-level loss functions to construct their network architectures. Therefore, when these methods are applied to complex urban areas, their performance in terms of change detection tends to be limited. In this article, a novel CNN-based change detection approach, referred to as a local similarity Siamese network (LSS-Net), with a cosine similarity measurement, was proposed for better urban land change detection in remote sensing images. To use content information on two sequential images, a new change attention map-based content loss function was developed in this study. In addition, to enhance the performance of the LSS-Net in terms of change detection, a suitable feature similarity measurement method, incorporated into a local similarity attention module, was determined through systemic experiments. To verify the change detection performance of the LSS-Net, it was compared with other state-of-the-art methods. The experimental results show that the proposed method outperforms the state-of-the-art methods in terms of the F1 score (0.9630, 0.9377, and 0.7751) and kappa (0.9581, 0.9351, and 0.7646) on the three test datasets, thus suggesting its potential for various remote sensing applications.

Acoustic holography has been gaining attention for various applications such as non-contact particle manipulation, non-invasive neuromodulation, and medical imaging. However, only a few studies on how to generate acoustic holograms have been conducted, and even conventional acoustic hologram algorithms show limited performance in the fast and accurate generation of acoustic holograms, thus hindering the development of novel applications. We here propose a deep learning-based framework to achieve fast and accurate acoustic hologram generation. The framework has an autoencoder-like architecture; thus the unsupervised training is realized without any ground truth. For the framework, we demonstrate a newly developed hologram generator network, the Holographic Ultrasound generation Network (HU-Net), which is suitable for unsupervised learning of hologram generation, and a novel loss function that is devised for energy-efficient holograms. Furthermore, for considering various hologram devices (i.e., ultrasound transducers), we propose a physical constraint layer. Simulation and experimental studies were carried out for two different hologram devices such as a 3D printed lens, attached to a single element transducer, and a 2D ultrasound array. The proposed framework was compared with the iterative angular spectrum approach (IASA) and the state-of-the-art iterative optimization method, Diff-PAT. In the simulation study, our framework showed a few hundred times faster generation speed, along with comparable or even better reconstruction quality, than those of IASA and Diff-PAT. In the experimental study, the framework was validated with 3D-printed lenses fabricated based on different methods, and the physical effect of the lenses on the reconstruction quality was discussed. The outcomes of the proposed framework in various cases (i.e., hologram generator networks, loss functions, hologram devices) suggest that our framework may become a very useful alternative tool for other existing acoustic hologram applications and it can expand novel medical applications.