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

Radiology Radiology

Deep Learning-Based Computed Tomography Image Standardization to Improve Generalizability of Deep Learning-Based Hepatic Segmentation.

In Korean journal of radiology

OBJECTIVE : We aimed to investigate whether image standardization using deep learning-based computed tomography (CT) image conversion would improve the performance of deep learning-based automated hepatic segmentation across various reconstruction methods.

MATERIALS AND METHODS : We collected contrast-enhanced dual-energy CT of the abdomen that was obtained using various reconstruction methods, including filtered back projection, iterative reconstruction, optimum contrast, and monoenergetic images with 40, 60, and 80 keV. A deep learning based image conversion algorithm was developed to standardize the CT images using 142 CT examinations (128 for training and 14 for tuning). A separate set of 43 CT examinations from 42 patients (mean age, 10.1 years) was used as the test data. A commercial software program (MEDIP PRO v2.0.0.0, MEDICALIP Co. Ltd.) based on 2D U-NET was used to create liver segmentation masks with liver volume. The original 80 keV images were used as the ground truth. We used the paired t-test to compare the segmentation performance in the Dice similarity coefficient (DSC) and difference ratio of the liver volume relative to the ground truth volume before and after image standardization. The concordance correlation coefficient (CCC) was used to assess the agreement between the segmented liver volume and ground-truth volume.

RESULTS : The original CT images showed variable and poor segmentation performances. The standardized images achieved significantly higher DSCs for liver segmentation than the original images (DSC [original, 5.40%-91.27%] vs. [standardized, 93.16%-96.74%], all P < 0.001). The difference ratio of liver volume also decreased significantly after image conversion (original, 9.84%-91.37% vs. standardized, 1.99%-4.41%). In all protocols, CCCs improved after image conversion (original, -0.006-0.964 vs. standardized, 0.990-0.998).

CONCLUSION : Deep learning-based CT image standardization can improve the performance of automated hepatic segmentation using CT images reconstructed using various methods. Deep learning-based CT image conversion may have the potential to improve the generalizability of the segmentation network.

Lee Seul Bi, Hong Youngtaek, Cho Yeon Jin, Jeong Dawun, Lee Jina, Yoon Soon Ho, Lee Seunghyun, Choi Young Hun, Cheon Jung-Eun

2023-Mar-07

Artificial intelligence, Automated segmentation, Image conversion, Quality control, Reproducibility

General General

Classification of household microplastics using a multi-model approach based on Raman spectroscopy.

In Chemosphere

The extensive use of plastics leads to the release and diffusion of microplastics. Household plastic products occupy a large part and are closely related to daily life. Due to the small size and complex composition of microplastics, it is challenging to identify and quantify microplastics. Therefore,a multi-model machine learning approach was developed for classification of household microplastics based on Raman spectroscopy. In this study, Raman spectroscopy and machine learning algorithm are combined to realize the accurate identification of seven standard microplastic samples, real microplastics samples and real microplastic samples post-exposure to environmental stresses. Four single-model machine learning methods were used in this study, including Support vector machine (SVM), K-nearest neighbor (KNN), Linear discriminant analysis (LDA), and Multi-layer perceptron (MLP) model. The principal components analysis (PCA) was utilized before SVM, KNN and LDA. The classification effect of four models on standard plastic samples is over 88%, and reliefF algorithm was used to distinguish HDPE and LDPE samples. A multi-model is proposed based on four single models including PCA-LDA, PCA-KNN and MLP. The recognition accuracy of multi-model for standard microplastic samples, real microplastic samples and microplastic samples post-exposure to environmental stresses is over 98%. Our study demonstrates that the multi-model coupled with Raman spectroscopy is a valuable tool for microplastic classification.

Feng Zikang, Zheng Lina, Liu Jia

2023-Mar-10

Classification, Environment stress, Household microplastics, Multi-model approach, Raman spectroscopy

General General

New proposal of viral genome representation applied in the classification of SARS-CoV-2 with deep learning.

In BMC bioinformatics

BACKGROUND : In December 2019, the first case of COVID-19 was described in Wuhan, China, and by July 2022, there were already 540 million confirmed cases. Due to the rapid spread of the virus, the scientific community has made efforts to develop techniques for the viral classification of SARS-CoV-2.

RESULTS : In this context, we developed a new proposal for gene sequence representation with Genomic Signal Processing techniques for the work presented in this paper. First, we applied the mapping approach to samples of six viral species of the Coronaviridae family, which belongs SARS-CoV-2 Virus. We then used the sequence downsized obtained by the method proposed in a deep learning architecture for viral classification, achieving an accuracy of 98.35%, 99.08%, and 99.69% for the 64, 128, and 256 sizes of the viral signatures, respectively, and obtaining 99.95% precision for the vectors with size 256.

CONCLUSIONS : The classification results obtained, in comparison to the results produced using other state-of-the-art representation techniques, demonstrate that the proposed mapping can provide a satisfactory performance result with low computational memory and processing time costs.

de Souza LuĂ­sa C, Azevedo Karolayne S, de Souza Jackson G, Barbosa Raquel de M, Fernandes Marcelo A C

2023-Mar-11

CGR DFT, COVID-19, Deep learning, GSP, SARS-CoV-2

General General

A novel use of an artificially intelligent Chatbot and a live, synchronous virtual question-and answer session for fellowship recruitment.

In BMC medical education

INTRODUCTION : Academic departments universally communicate information about their programs using static websites. In addition to websites, some programs have even ventured out into social media (SM). These bidirectional forms of SM interaction show great promise; even hosting a live Question and Answer (Q&A) session has the potential for program branding. Artificial Intelligence (AI) usage in the form of a chatbot has expanded on websites and in SM. The potential use of chatbots, for the purposes of trainee recruitment, is novel and underutilized. With this pilot study, we aimed to answer the question; can the use of an Artificially Intelligent Chatbot and a Virtual Question-and-Answer Session aid in recruitment in a Post-COVID-19 era?

METHODS : We held three structured Question-and-Answer Sessions over a period of 2 weeks. This preliminary study was performed after completion of the three Q&A sessions, in March-May, 2021. All 258 applicants to the pain fellowship program were invited via email to participate in the survey after attending one of the Q&A sessions. A 16-item survey assessing participants' perception of the chatbot was administered.

RESULTS : Forty-eight pain fellowship applicants completed the survey, for an average response rate of 18.6%. In all, 35 (73%) of survey respondents had used the website chatbot, and 84% indicated that it had found them the information they were seeking.

CONCLUSION : We employed an artificially intelligent chatbot on the department website to engage in a bidirectional exchange with users to adapt to changes brought on by the pandemic. SM engagement via chatbot and Q&A sessions can leave a favorable impression and improve the perception of a program.

Yi Peter K, Ray Neil D, Segall Noa

2023-Mar-11

Artificial intelligence, Graduate medical education, Innovation and technology, Recruitment, Social media

General General

Bandit-supported care planning for older people with complex health and care needs

ArXiv Preprint

Long-term care service for old people is in great demand in most of the aging societies. The number of nursing homes residents is increasing while the number of care providers is limited. Due to the care worker shortage, care to vulnerable older residents cannot be fully tailored to the unique needs and preference of each individual. This may bring negative impacts on health outcomes and quality of life among institutionalized older people. To improve care quality through personalized care planning and delivery with limited care workforce, we propose a new care planning model assisted by artificial intelligence. We apply bandit algorithms which optimize the clinical decision for care planning by adapting to the sequential feedback from the past decisions. We evaluate the proposed model on empirical data acquired from the Systems for Person-centered Elder Care (SPEC) study, a ICT-enhanced care management program.

Gi-Soo Kim, Young Suh Hong, Tae Hoon Lee, Myunghee Cho Paik, Hongsoo Kim

2023-03-13

General General

Comparison of photocatalysis and photolysis of 2,2,4,4-tetrabromodiphenyl ether (BDE-47): Operational parameters, kinetic studies, and data validation using three modern machine learning models.

In Chemosphere

Polybrominated diphenyl ethers (PBDEs) are halogenated organic compounds that are among the major pollutants of water, and there is an urgent need for their removal. This work compared the application of two techniques, i.e., photocatalytic reaction (PCR) and photolysis (PL), for 2,2,4,4- tetrabromodiphenyl ether (BDE-47) degradation. Although a limited degradation of BDE-47 was observed by photolysis (LED/N2), photocatalytic oxidation by using TiO2/LED/N2 proved to be effective in the degradation of BDE-47. The use of a photocatalyst enhanced the extent of BDE-47 degradation by around 10% at optimum conditions in anaerobic systems. Experimental results were systematically validated through modeling with three new and powerful Machine Learning (ML) approaches, including Gradient Boosted Decision Tree (GBDT), Artificial Neural Network (ANN), and Symbolic Regression (SBR). Four statistical criteria (Coefficient of Determination (R2), Root Mean Square Error (RMSE), Average Relative Error (ARER), and Absolute Error (ABER)) were calculated for model validation. Among the applied models, the developed GBDT was the desirable model for predicting the remaining concentration (Ce) of BDE-47 for both processes. Total Organic Carbon (TOC) and Chemical Oxygen Demand (COD) results confirmed that BDE-47 mineralization required additional time than its degradation in both PCR and PL systems. The kinetic study demonstrated that BDE-47 degradation for both processes followed the pseudo-first-order form of the Langmuir-Hinshelwood (L-H) model. More importantly, the calculated electrical energy consumption of photolysis was shown to be ten percent higher than that for photocatalysis, possibly due to the higher irradiation time required in direct photolysis, which in turn increases electricity consumption. This study is useful in proposing a feasible and promising treatment process for the degradation of BDE-47.

Motamedi Mahsa, Yerushalmi Laleh, Haghighat Fariborz, Chen Zhi, Zhuang Yanbin

2023-Mar-10

Direct photolysis, Machine learning, PBDEs, Photocatalysis