Synthesis and preclinical evaluation of a novel molecular probe [18F]AlF-NOTA-PEG2-Asp2-PDL1P for PET imaging of PD-L1 positive tumor.

Immunotherapy has brought great benefits to cancer patients, but only some patients benefit from it. Noninvasive, real-time and dynamic monitoring of the effectiveness of immunotherapy through PET imaging may provide assistance for the treatment plan of immunotherapy. In this study, we designed and synthesized a new targeted PD-L1 peptide NOTA-PEG2-Asp2-PDL1P, which was labeled with nuclide 18F to obtain a new imaging agent [18F]AlF-NOTA-PEG2-Asp2-PDL1P. The total radiochemical yield of [18F]AlF-NOTA-PEG2-Asp2-PDL1P was 13.7 % (Uncorrected radiochemical yield, n > 5). [18F]AlF-NOTA-PEG2-Asp2-PDL1P achieved high radiochemical purity (>95 %) with a molar activity more than 51.2 GBq/µmol. [18F]AlF-NOTA-PEG2-Asp2-PDL1P exhibited good hydrophilicity and had good stability both in vivo and in vitro, it can specifically targets B16F10 tumor with PD-L1 expression, and had a relatively high retention in tumor, a relatively fast clearance in vivo and a higher tumor-to-non-target ratio, all of which could make [18F]AlF-NOTA-PEG2-Asp2-PDL1P a potential tracer for PD-L1 prediction before clinical immunotherapy.

Synthesis and preclinical evaluation of a novel probe [18F]AlF-NOTA-IPB-GPC3P for PET imaging of GPC3 positive tumor.

Glypican-3 (GPC3) is markedly overexpressed in hepatocellular carcinoma (HCC) and not expressed in normal liver tissues. In this study, a novel peptide PET imaging agent ([18F]AlF-NOTA-IPB-GPC3P) was developed to target GPC3 expressed in tumors. The overall radiochemical yield of [18F]AlF-NOTA-IPB-GPC3P was 10-15 %, and its lipophilicity, expressed as the logD value at a pH of 7.4, was -1.18 ± 0.06 (n = 3). Compared to the previously reported tracer [18F]AlF-GP2633, [18F]AlF-NOTA-IPB-GPC3P exhibited higher cellular uptake (15.13 vs 5.96) and internalized rate (80.63 % vs 35.93 %) in Huh7 cells at 120 min. Micro-PET/CT and biodistribution studies further demonstrated that [18F]AlF-NOTA-IPB-GPC3P exhibited significantly increased tumor uptake and prolonged tumor residence in Huh7 tumors compared to [18F]AlF-GP2633 (4.66 ± 0.22 % ID/g vs 0.72 ± 0.09 % ID/g at 60 min, p < 0.001; 5.05 ± 0.23 % ID/g vs 0.35 ± 0.08 % ID/g at 120 min, p < 0.001, respectively). Furthermore, the tumor-to-organ ratios of [18F]AlF-NOTA-IPB-GPC3P surpassed those of [18F]AlF-GP2633. Our results support the utilization of [18F]AlF-NOTA-IPB-GPC3P as a PET imaging agent targeting the GPC3 receptor for tumor detection.

Deep Learning Algorithm Enables Cerebral Venous Thrombosis Detection With Routine Brain Magnetic Resonance Imaging.

BACKGROUND: Cerebral venous thrombosis (CVT) is a rare cerebrovascular disease. Routine brain magnetic resonance imaging is commonly used to diagnose CVT. This study aimed to develop and evaluate a novel deep learning (DL) algorithm for detecting CVT using routine brain magnetic resonance imaging. METHODS: Routine brain magnetic resonance imaging, including T1-weighted, T2-weighted, and fluid-attenuated inversion recovery images of patients suspected of CVT from April 2014 through December 2019 who were enrolled from a CVT registry, were collected. The images were divided into 2 data sets: a development set and a test set. Different DL algorithms were constructed in the development set using 5-fold cross-validation. Four radiologists with various levels of expertise independently read the images and performed diagnosis within the test set. The diagnostic performance on per-patient and per-segment diagnosis levels of the DL algorithms and radiologist's assessment were evaluated and compared. RESULTS: A total of 392 patients, including 294 patients with CVT (37±14 years, 151 women) and 98 patients without CVT (42±15 years, 65 women), were enrolled. Of these, 100 patients (50 CVT and 50 non-CVT) were randomly assigned to the test set, and the other 292 patients comprised the development set. In the test set, the optimal DL algorithm (multisequence multitask deep learning algorithm) achieved an area under the curve of 0.96, with a sensitivity of 96% (48/50) and a specificity of 88% (44/50) on per-patient diagnosis level, as well as a sensitivity of 88% (129/146) and a specificity of 80% (521/654) on per-segment diagnosis level. Compared with 4 radiologists, multisequence multitask deep learning algorithm showed higher sensitivity both on per-patient (all P<0.05) and per-segment diagnosis levels (all P<0.001). CONCLUSIONS: The CVT-detected DL algorithm herein improved diagnostic performance of routine brain magnetic resonance imaging, with high sensitivity and specificity, which provides a promising approach for detecting CVT.

Pharmacokinetic analysis of 6-O-[18F]FEE for PET imaging of EGFR mutation.

6-O-[18F]Fluoroethylerlotinib (6-O-[18F]FEE), with a suitable half-life for commercial distribution, may be a good replacement for [11C]erlotinib to identify epidermal growth factor receptor (EGFR) positive tumors with activating mutations to tyrosine kinase inhibitors therapy. In this study, we explored the fully automated synthesis of 6-O-[18F]FEE and investigated its pharmacokinetics in tumor-bearing mice. 6-O-[18F]FEE with high specific activity (28-100 GBq/µmol) and radiochemistry purity (over 99 %) was obtained by two-step reaction and Radio-HPLC separation in PET-MF-2 V-IT-1 automated synthesizer. PET imaging of 6-O-[18F]FEE in HCC827, A431, and U87 tumor-bearing mice with different EGFR expression and mutation was performed. Uptake and blocking of PET imaging indicated that the probe specifically targeted exon 19 deleted EGFR (the quantitative analysis of tumor-to-mouse ratio for HCC827, HCC827 blocking, U87, A431 was 2.58 ± 0.24, 1.20 ± 0.15, 1.18 ± 0.19, and 1.05 ± 0.13 respectively). Dynamic imaging was used to study the pharmacokinetics of the probe in tumor-bearing mice. Logan plot graphical analysis demonstrated late linearity and a high fitting correlation coefficient (0.998), supporting reversible kinetics. According to the Akaike Information Criterion (AIC) rule, the 2-compartment reversible model was more consistent with the metabolic properties of 6-O-[18F]FEE. The automated radiosynthesis and pharmacokinetic analysis will promote clinically transformation of 6-O-[18F]FEE.

Radiosynthesis, preclinical evaluation and pilot clinical PET imaging study of a 18F-labeled tracer targeting fibroblast activation protein.

Fibroblast activation protein (FAP) is a promising molecular target for imaging in various types of cancers. Several 18F-labeled FAP inhibitor (FAPI) tracers have been evaluated in clinical study. However, these tracers display high physiological uptake in gallbladder and bile duct system. To overcome the limitation, we herein designed a novel radiotracer named 18F-FAPTG. 18F-FAPTG was produced with a non-decay-corrected radiochemical yield of 24.0 ± 6.0% and 22.0 ± 7.0% for manual and automatic synthesis, respectively. 18F-FAPTG exhibited high hydrophilicity and stability in vitro. The studies of cellular uptake, internalization, efflux properties and competitive binding to FAP of 18F-FAPTG indicated that the tracer showed high specificity, rapid internalization and low cellular efflux in FAP-positive cells. Biodistribution studies and microPET in mice bearing FAP-positive xenografts demonstrated extremely low uptake in the majority of other organs and main excretion of 18F-FAPTG through the urinary system. Furthermore, compared to 18F-FAPI-42, 18F-FAPTG showed significantly lower uptake in gallbladder, higher tumor uptake and longer tumor retention. In the pilot clinical study, 18F-FAPTG PET/CT demonstrated favorable tumor-to-background ratios in most organs and clearly displayed the malignant lesions. Our findings indicated that 18F-FAPTG had an advantage over 18F-FAPI-42 in PET imaging for cancers located in gallbladder the bile duct system. Thus, 18F-FAPTG could be an alternative to the currently available FAPI tracers.

CNN-LSTM Neural Network for Identification of Pre-Cooked Pasta Products in Different Physical States Using Infrared Spectroscopy.

Infrared (IR) spectroscopy is nondestructive, fast, and straightforward. Recently, a growing number of pasta companies have been using IR spectroscopy combined with chemometrics to quickly determine sample parameters. However, fewer models have used deep learning models to classify cooked wheat food products and even fewer have used deep learning models to classify Italian pasta. To solve these problems, an improved CNN-LSTM neural network is proposed to identify pasta in different physical states (frozen vs. thawed) using IR spectroscopy. A one-dimensional convolutional neural network (1D-CNN) and long short-term memory (LSTM) were constructed to extract the local abstraction and sequence position information from the spectra, respectively. The results showed that the accuracy of the CNN-LSTM model reached 100% after using principal component analysis (PCA) on the Italian pasta spectral data in the thawed state and 99.44% after using PCA on the Italian pasta spectral data in the frozen form, verifying that the method has high analytical accuracy and generalization. Therefore, the CNN-LSTM neural network combined with IR spectroscopy helps to identify different pasta products.

Displacement sensing in a multimode SNAP microcavity by an artificial neural network.

Benefiting from the coupling between the Surface Nanoscale Axial Photonics (SNAP) microcavity and the waveguide, i.e., influenced by their abrupt field overlap, multiple axial modes in the transmission spectrum form a functional relationship with the coupling position, thus enabling displacement sensing. However, this functional relationship is complex and nonlinear, which is difficult to be fitted using analytical methods. We introduce a back-propagation neural network (BPNN) to model this functional relationship. The numerical results show that the multimode sensing scheme has great potential for practical large-range, high-precision displacement sensing platforms compared with the single-mode sensing based on the whispering gallery mode (WGM) resonators.

Synthesis and biological evaluation of Al[18F]-NOTA-IPB-PDL1P as a molecular probe for PET imaging of PD-L1 positive tumors.

PD-L1 is widely expressed in a variety of tumors, including NSCLC, melanoma, renal cell carcinoma, gastric cancer, hepatocellular as well as cutaneous and various leukemias, multiple myeloma and so on. Herein, we designed a novel peptide imaging agent (Al[18F]-NOTA-IPB-PDL1P) that specifically targets PD-L1 expressed in tumors. The overall radiochemical yield of Al[18F]-NOTA-IPB-PDL1P from 18F- was 10-15% (corrected radiochemical yield) within 20 min and the radiochemical purity of Al[18F]-NOTA-IPB-PDL1P was > 95% with a molar activity of 44.4-64.8 GBq/µmol. The lipophilicity logP value of Al[18F]-NOTA-IPB-PDL1P at pH 7.4 was -1.768 ±â€¯0.007 (n = 3). In the cellular uptake experiment, both HCT116 and PC3 cells dispalyed high uptake to Al[18F]-NOTA-IPB-PDL1P. The results of biodistribution showed that the uptake of Al[18F]-NOTA-IPB-PDL1P was high in kidneys, gall bladder and lung, and low in muscle and brain. In vivo micro PET studies, both HCT116 and PC3 tumors displayed high uptake for Al[18F]-NOTA-IPB-PDL1P, the tumor/muscle (T/M) radio was 2.93 and 3.57 respectively at 120 min. All the results indicate that Al[18F]-NOTA-IPB-PDL1P may have potential to be a PET imaging agent of tumors with high PD-L1 expression.

Haze optical-model-based nighttime image dehazing by modifying attenuation and atmospheric light.

Scene imaging is often affected by artificial light sources within a hazy environment at night, causing degraded images with low brightness, color distortion, and glow. These problems render the traditional atmospheric scattering optical model obsolete and incompatible. To address this issue, we established an optical imaging model suitable for nighttime dehazing, and an illumination component is incorporated into the attenuation term. We also introduced the near-light source coefficient to redefine the glow. Based on this model, we propose a new nighttime dehazing method. First, the rough atmospheric light is estimated using its low-frequency characteristics. Then, the glow is calculated by the near-light source coefficient. Finally, we remove the haze and illumination to get a clear image. Extensive experiments prove that our method exhibits a better color recovery effect, which effectively improves the visibility and detail. Furthermore, we believe our method outperforms other methods, both qualitatively and quantitatively.

Efficacy and safety of lenvatinib monotreatment and lenvatinib-based combination therapy for patients with unresectable hepatocellular carcinoma: a retrospective, real-world study in China.

BACKGROUND: Lenvatinib and lenvatinib-based combination treatments are widely used in patients with unresectable hepatocellular carcinoma (uHCC) in clinical practice, but their curative effect and safety need further study in the real world. METHODS: This was a retrospective study involving patients with uHCC receiving lenvatinib monotherapy and lenvatinib-based combination treatment between Nov, 2018 and Sep, 2020 in Nanfang Hospital. Efficacy was evaluated with objective response rate (ORR), disease control rate (DCR), progression-free survival (PFS), time to tumor progression (TTP), and overall survival (OS). Treatment-related adverse events (TRAEs) were recorded and graded. Efficacy and safety of monotherapy and combination therapy were compared. Stratified analysis was performed according to systemic line of treatment and medication regimen for combination therapy. RESULTS: For lenvatinib monotherapy (n = 39), OS and PFS were 80 weeks and 24.3 weeks, respectively. For combination treatment (n = 72), median OS and PFS were 99 weeks and 45.6 weeks, respectively. OS, PFS, and TTP for patients in the combination treatment cohort were significantly longer compared to those of patients in the monotreatment cohort (OS: P = 0.04, PFS: P = 0.003; TTP, P = 0.005). The incidence of TRAEs could be controlled both in the monotherapy cohort and the combination treatment cohort. In the monotherapy cohort, OS and PFS were significantly decreased in the second-line treatment group compared with the first-line treatment group, while no differences were observed in the combination cohort. The efficacy of triple therapy (lenvatinib plus PD-1 antibody plus TACE or HAIF) was similar to lenvatinib plus PD-1 antibody or lenvatinib plus TACE or HAIF. CONCLUSIONS: Our real-world study showed that lenvatinib monotherapy and lenvatinib-based combination therapy were well tolerated, with encouraging efficacies in patients with uHCC. Lenvatinib-based combination therapy showed a better curative effect compared with lenvatinib single-agent therapy. In patients who have failed first-line TKI treatment, lenvatinib-based combination therapy may be a better choice than lenvatinib single-agent therapy. Lenvatinib-based triple therapy may not have an advantage over dual therapy.

Simulation and Optimization of SNAP-Taper Coupling System in Displacement Sensing.

Sensing applications based on whispering gallery mode (WGM) microcavities have attracted extensive attention recently, especially in displacement sensing applications. However, the traditional displacement sensing scheme based on shift in a single resonance wavelength, has a lot of drawbacks. Herein, a novel displacement sensing scheme based on the surface nanoscale axial photonics (SNAP) is proposed to achieve a wide range and high-resolution displacement sensor through analyzing the transmittance of multiple axial modes. By analyzing the surface plot of the resonance spectrum with different coupling positions, the ideal coupling parameters and ERV for displacement sensing are obtained. In the following, displacement sensing with high sensitivity and a wide range is theoretically realized through adjusting the sensitivity threshold and the number of modes. Finally, we present our views on the current challenges and the future development of the displacement sensing based on an SNAP resonator. We believe that a comprehensive understanding on this sensing scheme would significantly contribute to the advancement of the SNAP resonator for a broad range of applications.

Radiosynthesis and biological evaluation of an fluorine-18 labeled galactose derivative [18F]FPGal for imaging the hepatic asialoglycoprotein receptor.

The asialoglycoprotein receptor (ASGPR) is abundantly expressed on the surface of hepatocytes where it recognizes and endocytoses glycoproteins with galactosyl and N-acetylgalactosamine groups. Given its hepatic distribution, the asialoglycoprotein receptor can be targeted by positron imaging agents to study liver function using PET imaging. In this study, the positron imaging agent [18F]FPGal was designed to specifically target hepatic asialoglycoprotein receptor and its effectiveness was assessed in in vitro and in vivo models. The radiosynthesis of [18F]FPGal required 50 min with total radiochemical yields of [18F]FPGal from [18F]fluoride as 10% (corrected radiochemical yield). The Kd of [18F]FPGal to ASGPR in HepG2 cells was 1.99 ± 0.05 mM. Uptake values of 0.55% were observed within 30 min of incubation with HepG2 cells, which could be blocked by 200 mM d(+)-galactose (<0.1%). In vivo biodistribution analysis showed that the liver accumulation of [18F]FPGal at 30 min was 4.47 ± 0.96% ID/g in normal mice compared to 1.33 ± 0.07% ID/g in hepatic fibrotic mice (P < 0.01). Reduced uptake in the hepatic fibrosis mouse models was confirmed through PET/CT images at 30 min. Compared to normal mice, the standard uptake value (SUV) in the hepatic fibrosis mice was significantly lower when assessed through dynamic data collection for 1 h. Therefore, [18F]FPGal is a feasible PET probe that provide insight into ASGPR related liver disease.

Radiosynthesis and preclinical evaluation of [18F]FEM as a potential novel PET probe for tumor imaging.

In the 21st century, the incidence and mortality of cancer, one of the most challenging diseases in the world, have rapidly increased. The purpose of this study was to develop 2-(2-[18F]fluoroethoxy)ethyl 4-methylbenzenesulfonate ([18F]FEM) as a positron emission tomography (PET) agent for tumor imaging. In this study, [18F]FEM was synthesized with a good radiochemical yield (45.4 ± 5.8%), high specific radioactivity (over 25 GBq/µmol), and commendable radiochemical purity (over 99%). The octanol/water partition coefficient of [18F]FEM was 1.44 ± 0.04. The probe demonstrated good stability in vitro (phosphate-buffered saline (PBS) and mouse serum (MS)), and binding specificity to five different tumor cell lines (A549, PC-3, HCC827, U87, and MDA-MB-231). PET imaging of tumor-bearing mice showed that [18F]FEM specifically accumulated at the tumor site of the five different tumor cell lines. The average tumor-to-muscle (T/M) ratio was over 2, and the maximum T/M values reached about 3.5. The biodistribution and dynamic PET imaging showed that most probes were metabolized by the liver, whereas a small part was metabolized by the kidney. Moreover, dynamic brain images and quantitative data showed [18F]FEM can quickly cross the blood brain barrier (BBB) and quickly fade out, thereby suggesting it may be a promising candidate probe for the imaging of brain tumors. The presented results demonstrated that [18F]FEM is a promising probe for tumor PET imaging.

MiR-155 Enhances Insulin Sensitivity by Coordinated Regulation of Multiple Genes in Mice.

miR-155 plays critical roles in numerous physiological and pathological processes, however, its function in the regulation of blood glucose homeostasis and insulin sensitivity and underlying mechanisms remain unknown. Here, we reveal that miR-155 levels are downregulated in serum from type 2 diabetes (T2D) patients, suggesting that miR-155 might be involved in blood glucose control and diabetes. Gain-of-function and loss-of-function studies in mice demonstrate that miR-155 has no effects on the pancreatic ß-cell proliferation and function. Global transgenic overexpression of miR-155 in mice leads to hypoglycaemia, improved glucose tolerance and insulin sensitivity. Conversely, miR-155 deficiency in mice causes hyperglycemia, impaired glucose tolerance and insulin resistance. In addition, consistent with a positive regulatory role of miR-155 in glucose metabolism, miR-155 positively modulates glucose uptake in all cell types examined, while mice overexpressing miR-155 transgene show enhanced glycolysis, and insulin-stimulated AKT and IRS-1 phosphorylation in liver, adipose tissue or skeletal muscle. Furthermore, we reveal these aforementioned phenomena occur, at least partially, through miR-155-mediated repression of important negative regulators (i.e. C/EBPß, HDAC4 and SOCS1) of insulin signaling. Taken together, these findings demonstrate, for the first time, that miR-155 is a positive regulator of insulin sensitivity with potential applications for diabetes treatment.

In that vein: inflated wing veins contribute to butterfly hearing.

Insects have evolved a diversity of hearing organs specialized to detect sounds critical for survival. We report on a unique structure on butterfly wings that enhances hearing. The Satyrini are a diverse group of butterflies occurring throughout the world. One of their distinguishing features is a conspicuous swelling of their forewing vein, but the functional significance of this structure is unknown. Here, we show that wing vein inflations function in hearing. Using the common wood nymph, Cercyonis pegala, as a model, we show that (i) these butterflies have ears on their forewings that are most sensitive to low frequency sounds (less than 5 kHz); (ii) inflated wing veins are directly connected to the ears; and (iii) when vein inflations are ablated, sensitivity to low frequency sounds is impaired. We propose that inflated veins contribute to low frequency hearing by impedance matching.

Radiosynthesis and biological evaluation of 18F-labeled 4-anilinoquinazoline derivative (18F-FEA-Erlotinib) as a potential EGFR PET agent.

Epidermal growth factor receptor (EGFR) has gained significant attention as a therapeutic target. Several EGFR targeting drugs (Gefitinib and Erlotinib) have been approved by US Food and Drug Administration (FDA) and have received high approval in clinical treatment. Nevertheless, the curative effect of these medicines varied in many solid tumors because of the different levels of expression and mutations of EGFR. Therefore, several PET radiotracers have been developed for the selective treatment of responsive patients who undergo PET/CT imaging for tyrosine kinase inhibitor (TKI) therapy. In this study, a novel fluorine-18 labeled 4-anilinoquinazoline based PET tracer, 1N-(3-(1-(2-18F-fluoroethyl)-1H-1,2,3-triazol-4-yl)phenyl)-6,7-bis(2-methoxyethoxy)quinazolin-4-amine (18F-FEA-Erlotinib), was synthesized and biological evaluation was performed in vitro and in vivo. 18F-FEA-Erlotinib was achieved within 50min with over 88% radiochemical yield (decay corrected RCY), an average specific activity over 50GBq/µmol, and over 99% radiochemical purity. In vitro stability study showed no decomposition of 18F-FEA-Erlotinib after incubated in PBS and FBS for 2h. Cellular uptake and efflux experiment results indicated the specific binding of 18F-FEA-Erlotinib to HCC827 cell line with EGFR exon 19 deletions. In vivo, Biodistribution studies revealed that 18F-FEA-Erlotinib exhibited rapid blood clearance both through hepatobiliary and renal excretion. The tumor uptake of 18F-FEA-Erlotinib in HepG2, HCC827, and A431 tumor xenografts, with different EGFR expression and mutations, was visualized in PET images. Our results demonstrate the feasibility of using 18F-FEA-Erlotinib as a PET tracer for screening EGFR TKIs sensitive patients.

Efficient nitrogen removal by microalgal-bacterial granular sludge-marimo coupling process.

Current biological wastewater treatment processes usually have a drawback of insufficient nitrogen (N) removal, contributing to the ubiquitous eutrophication of aquatic ecosystems globally. To address such a challenging situation, this study explored an innovative microalgal-bacterial granular sludge-marimo (MBGS-MA) coupling process. The process removed 83.4 % of N with the effluent N concentration of 4.0 mg/L. With the growth of MBGS, there was a shift towards genes associated with nitrification and denitrification, and away from ammonia assimilation genes, revealing internal mechanism of the shift of N removal pathway. Contrarily, MA could use gaseous N2 with the N fixing genes in MA enriched, and the genes abundance related to assimilatory nitrate reduction were also raised under the mutualistic interactions between Proteobacteria and Cyanobacteria, which was beneficial to achieve efficient N removal. These findings may open a new horizon for developing innovative hybrid microalgal-bacterial processes aimed at high-efficiency N removal from wastewater.

Predicting the risk of mortality in ICU patients based on dynamic graph attention network of patient similarity.

Predicting the risk of mortality of hospitalized patients in the ICU is essential for timely identification of high-risk patients and formulate and adjustment of treatment strategies when patients are hospitalized. Traditional machine learning methods usually ignore the similarity between patients and make it difficult to uncover the hidden relationships between patients, resulting in poor accuracy of prediction models. In this paper, we propose a new model named PS-DGAT to solve the above problem. First, we construct a patient-weighted similarity network by calculating the similarity of patient clinical data to represent the similarity relationship between patients; second, we fill in the missing features and reconstruct the patient similarity network based on the data of neighboring patients in the network; finally, from the reconstructed patient similarity network after feature completion, we use the dynamic attention mechanism to extract and learn the structural features of the nodes to obtain a vector representation of each patient node in the low-dimensional embedding The vector representation of each patient node in the low-dimensional embedding space is used to achieve patient mortality risk prediction. The experimental results show that the accuracy is improved by about 1.8% compared with the basic GAT and about 8% compared with the traditional machine learning methods.

A CNN-LSTM-att hybrid model for classification and evaluation of growth status under drought and heat stress in chinese fir (Cunninghamia lanceolata).

BACKGROUND: Cunninghamia lanceolata (Chinese fir), is one of the most important timber trees in China. With the global warming, to develop new resistant varieties to drought or heat stress has become an essential task for breeders of Chinese fir. However, classification and evaluation of growth status of Chinese fir under drought or heat stress are still labor-intensive and time-consuming. RESULTS: In this study, we proposed a CNN-LSTM-att hybrid model for classification of growth status of Chinese fir seedlings under drought and heat stress, respectively. Two RGB image datasets of Chinese fir seedling under drought and heat stress were generated for the first time, and utilized in this study. By comparing four base CNN models with LSTM, the Resnet50-LSTM was identified as the best model in classification of growth status, and LSTM would dramatically improve the classification performance. Moreover, attention mechanism further enhanced performance of Resnet50-LSTM, which was verified by Grad-CAM. By applying the established Resnet50-LSTM-att model, the accuracy rate and recall rate of classification was up to 96.91% and 96.79% for dataset of heat stress, and 96.05% and 95.88% for dataset of drought, respectively. Accordingly, the R2 value and RMSE value for evaluation on growth status under heat stress were 0.957 and 0.067, respectively. And, the R2 value and RMSE value for evaluation on growth status under drought were 0.944 and 0.076, respectively. CONCLUSION: In summary, our proposed model provides an important tool for stress phenotyping in Chinese fir, which will be a great help for selection and breeding new resistant varieties in future.