Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

Pathology Pathology

Synthetic Sample Selection via Reinforcement Learning

ArXiv Preprint

Synthesizing realistic medical images provides a feasible solution to the shortage of training data in deep learning based medical image recognition systems. However, the quality control of synthetic images for data augmentation purposes is under-investigated, and some of the generated images are not realistic and may contain misleading features that distort data distribution when mixed with real images. Thus, the effectiveness of those synthetic images in medical image recognition systems cannot be guaranteed when they are being added randomly without quality assurance. In this work, we propose a reinforcement learning (RL) based synthetic sample selection method that learns to choose synthetic images containing reliable and informative features. A transformer based controller is trained via proximal policy optimization (PPO) using the validation classification accuracy as the reward. The selected images are mixed with the original training data for improved training of image recognition systems. To validate our method, we take the pathology image recognition as an example and conduct extensive experiments on two histopathology image datasets. In experiments on a cervical dataset and a lymph node dataset, the image classification performance is improved by 8.1% and 2.3%, respectively, when utilizing high-quality synthetic images selected by our RL framework. Our proposed synthetic sample selection method is general and has great potential to boost the performance of various medical image recognition systems given limited annotation.

Jiarong Ye, Yuan Xue, L. Rodney Long, Sameer Antani, Zhiyun Xue, Keith Cheng, Xiaolei Huang


General General

Cascade marker removal algorithm for thyroid ultrasound images.

In Medical & biological engineering & computing ; h5-index 32.0

During thyroid ultrasound diagnosis, radiologists add markers such as pluses or crosses near a nodule's edge to indicate the location of a nodule. For computer-aided detection, deep learning models achieve classification, segmentation, and detection by learning the thyroid's texture in ultrasound images. Experiments show that manual markers are strong prior knowledge for data-driven deep learning models, which interferes with the judgment mechanism of computer-aided detection systems. Aiming at this problem, this paper proposes cascade marker removal algorithm for thyroid ultrasound images to eliminate the interference of manual markers. The algorithm consists of three parts. First, in order to highlight marked features, the algorithm extracts salient features in thyroid ultrasound images through feature extraction module. Secondly, mask correction module eliminates the interference of other features besides markers' features. Finally, the marker removal module removes markers without destroying the semantic information in thyroid ultrasound images. Experiments show that our algorithm enables classification, segmentation, and object detection models to focus on the learning of pathological tissue features. At the same time, compared with mainstream image inpainting algorithms, our algorithm shows better performance on thyroid ultrasound images. In summary, our algorithm is of great significance for improving the stability and performance of computer-aided detection systems. Graphical Abstract During thyroid ultrasound diagnosis, doctors add markers such as pluses or crosses near nodule's edge to indicate the location of nodule. Manual markers are strong prior knowledge for data-driven deep learning models, which interferes the judgment mechanism of computer-aided diagnosis system based on deep learning. Markers make models overfit the specific labeling forms easily, and performs poorly on unmarked thyroid ultrasound images. Aiming at this problem, this paper proposes a cascade marker removal algorithm to eliminate the interference of manual markers. Our algorithm make deep learning models pay attention on nodules' features of thyroid ultrasound images, which make computer-aided diagnosis system performs good in both marked imaging and unmarked imaging.

Ying Xiang, Zhang Yulin, Yu Mei, Wei Xi, Zhu Jialin, Gao Jie, Liu Zhiqiang, Shen Hongqian, Zhang Ruixuan, Li Xuewei, Yu Ruiguo


Computer-assisted, Deep learning, Diagnosis, Diagnostic imaging, Thyroid neoplasms, Ultrasonography

General General

DELPHI: accurate deep ensemble model for protein interaction sites prediction.

In Bioinformatics (Oxford, England)

MOTIVATION : Proteins usually perform their functions by interacting with other proteins, which is why accurately predicting protein-protein interaction (PPI) binding sites is a fundamental problem. Experimental methods are slow and expensive. Therefore, great efforts are being made towards increasing the performance of computational methods.

RESULTS : We propose DELPHI (DEep Learning Prediction of Highly probable protein Interaction sites), a new sequence-based deep learning suite for PPI binding sites prediction. DELPHI has an ensemble structure which combines a CNN and a RNN component with fine tuning technique. Three novel features, HSP, position information, and ProtVec are used in addition to nine existing ones. We comprehensively compare DELPHI to nine state-of-the-art programs on five datasets, and DELPHI outperforms the competing methods in all metrics even though its training dataset shares the least similarities with the testing datasets. In the most important metrics, AUPRC and MCC, it surpasses the second best programs by as much as 18.5% and 27.7%, resp. We also demonstrated that the improvement is essentially due to using the ensemble model and, especially, the three new features. Using DELPHI it is shown that there is a strong correlation with protein-binding residues (PBRs) and sites with strong evolutionary conservation. In addition DELPHI's predicted PBR sites closely match known data from Pfam. DELPHI is available as open sourced standalone software and web server.

AVAILABILITY : The DELPHI web server can be found at, with all datasets and results in this study. The trained models, the DELPHI standalone source code, and the feature computation pipeline are freely available at

SUPPLEMENTARY INFORMATION : Supplementary data are available at Bioinformatics online.

Li Yiwei, Golding G Brian, Ilie Lucian


General General

Machine learning reveals multiple classes of diamond nanoparticles.

In Nanoscale horizons

Generating samples of nanoparticles with specific properties that allow for structural diversity, rather than requiring structural precision, is a more sustainable prospect for industry, where samples need to be both targeted to specific applications and cost effective. This can be better enabled by defining classes of nanoparticles and characterising the properties of the class as a whole. In this study, we use machine learning to predict the different classes of diamond nanoparticles based entirely on the structural features and explore the populations of these classes in terms of the size, shape, speciation and charge transfer properties. We identify 9 different types of diamond nanoparticles based on their similarity in 17 dimensions and, contrary to conventional wisdom, find that the fraction of sp2 or sp3 hybridized atoms are not strong determinants, and that the classes are only weakly related to size. Each class has been describe in such way as to enable rapid assignment using microanalysis techniques.

Parker Amanda J, Barnard Amanda S


Radiology Radiology

Continuous Learning AI in Radiology: Implementation Principles and Early Applications.

In Radiology ; h5-index 91.0

Artificial intelligence (AI) is becoming increasingly present in radiology and health care. This expansion is driven by the principal AI strengths: automation, accuracy, and objectivity. However, as radiology AI matures to become fully integrated into the daily radiology routine, it needs to go beyond replicating static models, toward discovering new knowledge from the data and environments around it. Continuous learning AI presents the next substantial step in this direction and brings a new set of opportunities and challenges. Herein, the authors discuss the main concepts and requirements for implementing continuous AI in radiology and illustrate them with examples from emerging applications.

Pianykh Oleg S, Langs Georg, Dewey Marc, Enzmann Dieter R, Herold Christian J, Schoenberg Stefan O, Brink James A


Radiology Radiology

What is the point of innovation in patient dose monitoring?

In Annals of the ICRP

The Medical Futurist says that radiology is one of the fastest growing and developing areas of medicine, and therefore this might be the speciality in which we can expect to see the largest steps in development. So why do they think that, and does it apply to dose monitoring? The move from retrospective dose evaluation to a proactive dose management approach represents a serious area of research. Indeed, artificial intelligence and machine learning are consistently being integrated into best-in-class dose management software solutions. The development of clinical analytics and dashboards are already supporting operators in their decision-making, and these optimisations - if taken beyond a single machine, a single department, or a single health network - have the potential to drive real and lasting change. The question is for whom exactly are these innovations being developed? How can the patient know that their scan has been performed to the absolute best that the technology can deliver? Do they know or even care how much their lifetime risk for developing cancer has changed post examination? Do they want a personalised size-specific dose estimate or perhaps an individual organ dose assessment to share on Instagram? Let's get real about the clinical utility and regulatory application of dose monitoring, and shine a light on the shared responsibility in applying the technology and the associated innovations.

Hislop-Jambrich J


CT, Dose; Innovation; Applications