MyMedR

Displaying all 5 publications

Abstract:

Sort:

A deep convolutional neural network for the automatic segmentation of glioblastoma brain tumor: Joint spatial pyramid module and attention mechanism network

Liu H, Huang J, Li Q, Guan X, Tseng M

Artif Intell Med, 2024 Feb;148:102776.
PMID: 38325925 DOI: 10.1016/j.artmed.2024.102776

This study proposes a deep convolutional neural network for the automatic segmentation of glioblastoma brain tumors, aiming sat replacing the manual segmentation method that is both time-consuming and labor-intensive. There are many challenges for automatic segmentation to finely segment sub-regions from multi-sequence magnetic resonance images because of the complexity and variability of glioblastomas, such as the loss of boundary information, misclassified regions, and subregion size. To overcome these challenges, this study introduces a spatial pyramid module and attention mechanism to the automatic segmentation algorithm, which focuses on multi-scale spatial details and context information. The proposed method has been tested in the public benchmarks BraTS 2018, BraTS 2019, BraTS 2020 and BraTS 2021 datasets. The Dice score on the enhanced tumor, whole tumor, and tumor core were respectively 79.90 %, 89.63 %, and 85.89 % on the BraTS 2018 dataset, respectively 77.14 %, 89.58 %, and 83.33 % on the BraTS 2019 dataset, and respectively 77.80 %, 90.04 %, and 83.18 % on the BraTS 2020 dataset, and respectively 83.48 %, 90.70 %, and 88.94 % on the BraTS 2021 dataset offering performance on par with that of state-of-the-art methods with only 1.90 M parameters. In addition, our approach significantly reduced the requirements for experimental equipment, and the average time taken to segment one case was only 1.48 s; these two benefits rendered the proposed network intensely competitive for clinical practice.
Fulltext Sales of anti-cancer medicines; China, Indonesia, Kazakhstan, Malaysia, Philippines and Thailand

Ferrario A, Stephens P, Guan X, Ross-Degnan D, Wagner A

Bull World Health Organ, 2020 Jul 01;98(7):467-474.
PMID: 32742032 DOI: 10.2471/BLT.19.243998

OBJECTIVE: To assess sales of anti-cancer medicines in the 2017 World Health Organization's WHO Model list of essential medicines in China, Indonesia, Kazakhstan, Malaysia, Philippines and Thailand from 2007 (2008 for Kazakhstan and Malaysia) to 2017.
METHODS: We extracted sales volume data for 39 anti-cancer medicines from the IQVIA database. We divided the total quantity sold by the reference defined daily dose to estimate the total number of defined daily doses sold, per country per year, for three types of anti-cancer therapies (traditional chemotherapy, targeted therapy and endocrine therapy). We adjusted these data by the number of new cancer cases in each country for each year.
FINDINGS: We observed an increase in sales across all types of anti-cancer therapies in all countries. The largest number of defined daily doses of traditional chemotherapy per new cancer case was sold in Thailand; however, the largest relative increase per new cancer case occurred in Indonesia (9.48-fold). The largest absolute and relative increases in sales of defined daily doses of targeted therapies per new cancer case occurred in Kazakhstan. Malaysia sold the largest number of adjusted defined daily doses of endocrine therapies in 2017, while China and Indonesia more than doubled their adjusted sales volumes between 2007 and 2017.
CONCLUSION: The use of sales data can fill an important knowledge gap in the use of anti-cancer medicines, particularly during periods of insurance coverage expansion. Combined with other data, sales volume data can help to monitor efforts to improve equitable access to essential medicines.
Fulltext Dissolved N2O concentrations in oil palm plantation drainage in a peat swamp of Malaysia

Nishina K, Melling L, Toyoda S, Itoh M, Terajima K, Waili JWB, et al.

Sci Total Environ, 2023 May 10;872:162062.
PMID: 36804973 DOI: 10.1016/j.scitotenv.2023.162062

Oil palm plantations in Southeast Asia are the largest supplier of palm oil products and have been rapidly expanding in the last three decades even in peat-swamp areas. Oil palm plantations on peat ecosystems have a unique water management system that lowers the water table and, thus, may yield indirect N2O emissions from the peat drainage system. We conducted two seasons of spatial monitoring for the dissolved N2O concentrations in the drainage and adjacent rivers of palm oil plantations on peat swamps in Sarawak, Malaysia, to evaluate the magnitude of indirect N2O emissions from this ecosystem. In both the dry and wet seasons, the mean and median dissolved N2O concentrations exhibited over-saturation in the drainage water, i.e., the oil palm plantation drainage may be a source of N2O to the atmosphere. In the wet season, the spatial distribution of dissolved N2O showed bimodal peaks in both the unsaturated and over-saturated concentrations. The bulk δ15N of dissolved N2O was higher than the source of inorganic N in the oil palm plantation (i.e., N fertilizer and soil organic nitrogen) during both seasons. An isotopocule analysis of the dissolved N2O suggested that denitrification was a major source of N2O, followed by N2O reduction processes that occurred in the drainage water. The δ15N and site preference mapping analysis in dissolved N2O revealed that a significant proportion of the N2O produced in peat and drainage is reduced to N2 before being released into the atmosphere.
Identifying dominant environmental predictors of freshwater wetland methane fluxes across diurnal to seasonal time scales

Knox SH, Bansal S, McNicol G, Schafer K, Sturtevant C, Ueyama M, et al.

Glob Chang Biol, 2021 08;27(15):3582-3604.
PMID: 33914985 DOI: 10.1111/gcb.15661

While wetlands are the largest natural source of methane (CH4 ) to the atmosphere, they represent a large source of uncertainty in the global CH4 budget due to the complex biogeochemical controls on CH4 dynamics. Here we present, to our knowledge, the first multi-site synthesis of how predictors of CH4 fluxes (FCH4) in freshwater wetlands vary across wetland types at diel, multiday (synoptic), and seasonal time scales. We used several statistical approaches (correlation analysis, generalized additive modeling, mutual information, and random forests) in a wavelet-based multi-resolution framework to assess the importance of environmental predictors, nonlinearities and lags on FCH4 across 23 eddy covariance sites. Seasonally, soil and air temperature were dominant predictors of FCH4 at sites with smaller seasonal variation in water table depth (WTD). In contrast, WTD was the dominant predictor for wetlands with smaller variations in temperature (e.g., seasonal tropical/subtropical wetlands). Changes in seasonal FCH4 lagged fluctuations in WTD by ~17 ± 11 days, and lagged air and soil temperature by median values of 8 ± 16 and 5 ± 15 days, respectively. Temperature and WTD were also dominant predictors at the multiday scale. Atmospheric pressure (PA) was another important multiday scale predictor for peat-dominated sites, with drops in PA coinciding with synchronous releases of CH4 . At the diel scale, synchronous relationships with latent heat flux and vapor pressure deficit suggest that physical processes controlling evaporation and boundary layer mixing exert similar controls on CH4 volatilization, and suggest the influence of pressurized ventilation in aerenchymatous vegetation. In addition, 1- to 4-h lagged relationships with ecosystem photosynthesis indicate recent carbon substrates, such as root exudates, may also control FCH4. By addressing issues of scale, asynchrony, and nonlinearity, this work improves understanding of the predictors and timing of wetland FCH4 that can inform future studies and models, and help constrain wetland CH4 emissions.
Integrated image-based deep learning and language models for primary diabetes care

Li J, Guan Z, Wang J, Cheung CY, Zheng Y, Lim LL, et al.

Nat Med, 2024 Jul 19.
PMID: 39030266 DOI: 10.1038/s41591-024-03139-8

Primary diabetes care and diabetic retinopathy (DR) screening persist as major public health challenges due to a shortage of trained primary care physicians (PCPs), particularly in low-resource settings. Here, to bridge the gaps, we developed an integrated image-language system (DeepDR-LLM), combining a large language model (LLM module) and image-based deep learning (DeepDR-Transformer), to provide individualized diabetes management recommendations to PCPs. In a retrospective evaluation, the LLM module demonstrated comparable performance to PCPs and endocrinology residents when tested in English and outperformed PCPs and had comparable performance to endocrinology residents in Chinese. For identifying referable DR, the average PCP's accuracy was 81.0% unassisted and 92.3% assisted by DeepDR-Transformer. Furthermore, we performed a single-center real-world prospective study, deploying DeepDR-LLM. We compared diabetes management adherence of patients under the unassisted PCP arm (n = 397) with those under the PCP+DeepDR-LLM arm (n = 372). Patients with newly diagnosed diabetes in the PCP+DeepDR-LLM arm showed better self-management behaviors throughout follow-up (P