Requirements for implementation of artificial intelligence in the practice of gastrointestinal pathology

Tremendous advances in artificial intelligence(AI)in medical image analysis have been achieved in recent years.The integration of AI is expected to cause a revolution in various areas of medicine,including gastrointestinal(GI)pathology.Currently,deep learning algorithms have shown promising benefits in areas of diagnostic histopathology,such as tumor identification,classification,prognosis prediction,and biomarker/genetic alteration prediction.While AI cannot substitute pathologists,carefully constructed AI applications may increase workforce productivity and diagnostic accuracy in pathology practice.Regardless of these promising advances,unlike the areas of radiology or cardiology imaging,no histopathology-based AI application has been approved by a regulatory authority or for public reimbursement.Thus,implying that there are still some obstacles to be overcome before AI applications can be safely and effectively implemented in real-life pathology practice.The challenges have been identified at different stages of the development process,such as needs identification,data curation,model development,validation,regulation,modification of daily workflow,and cost-effectiveness balance.The aim of this review is to present challenges in the process of AI development,validation,and regulation that should be overcome for its implementation in real-life GI pathology practice.

Permanent Rare-Earth Magnets—The Need to Protect Them against Corrosion

The RE-Fe-B permanent magnets have a complex microstructure and they are susceptible to corrosion process. In this paper, the commercial nickel coatings adhesion was investigated. The microstructure of the RE-Fe-B without coating was analyzed by scanning electronic microscopy and electrochemical techniques. The interface magnet/coating was studied by scanning electron microscopy and the nickel-plated Nd-Fe-B commercial magnets were tested in a salt spray chamber. The ferromagnetic and RE-rich phases were observed. After the anodic polarization curve, a strong intergranular corrosion was observed and the RE-rich phase was preferentially attacked. The interface magnet/Ni coating presented inter-granular corrosion that can affect the nickel coating adherence. This attack had probably occurred during the electrodeposition process. Not all the samples suffered localized corrosion during the salt spray tests and the Ni triple-layer coating presented a few corrosion points. RE-Fe-B alloy magnets need to be protected with appropriate coatings to each environment to which they will be exposed and the protective coating must not be damaged.

Cosmological constraints on the Undulant Universe

We use the redshift Hubble parameter H（z） data derived from relative galaxy ages, distant type Ia supernovae （SNe Ia）, the Baryonic Acoustic Oscillation （BAO） peak, and the Cosmic Microwave Background （CMB） shift parameter data, to constrain cosmological parameters in the Undulant Universe. We marginalize the like- lihood functions over h by integrating the probability density 19 ∝ e-x^2/2. By using a Markov Chain Monte Carlo （MCMC） technique, we obtain the best fitting results and give the confidence regions in the b - Ωm0 plane. Then we compare their constraints. Our results show that the H（z） data play a similar role with the SNe Ia data in cosmological study. By presenting the independent and joint constraints, we find that the BAO and CMB data play very important roles in breaking the degeneracy compared with the H（z） and SNe Ia data alone. Combined with the BAO or CMB data, one can remarkably improve the constraints. The SNe Ia data sets constrain Ωm0 much tighter than the H（z） data sets, but the H（z） data sets constrain b much tighter than the SNe Ia data sets. All these results show that the Undulant Universe approaches the ACDM model. We expect more H（z） data to constrain cosmological parameters in the future.

Impact of blurs on machine-learning aided digital pathology image analysis

BACKGROUND Digital pathology image(DPI)analysis has been developed by machine learning(ML)techniques.However,little attention has been paid to the reproducibility of ML-based histological classification in heterochronously obtained DPIs of the same hematoxylin and eosin(HE)slide.AIM To elucidate the frequency and preventable causes of discordant classification results of DPI analysis using ML for the heterochronously obtained DPIs.METHODS We created paired DPIs by scanning 298 HE stained slides containing 584 tissues twice with a virtual slide scanner.The paired DPIs were analyzed by our MLaided classification model.We defined non-flipped and flipped groups as the paired DPIs with concordant and discordant classification results,respectively.We compared differences in color and blur between the non-flipped and flipped groups by L1-norm and a blur index,respectively.RESULTS We observed discordant classification results in 23.1%of the paired DPIs obtained by two independent scans of the same microscope slide.We detected no significant difference in the L1-norm of each color channel between the two groups;however,the flipped group showed a significantly higher blur index than the non-flipped group.CONCLUSION Our results suggest that differences in the blur-not the color-of the paired DPIs may cause discordant classification results.An ML-aided classification model for DPI should be tested for this potential cause of the reduced reproducibility of the model.In a future study,a slide scanner and/or a preprocessing method of minimizing DPI blur should be developed.

Identification of Newly Diagnosed Diabetes and Prediabetes Using Fasting Plasma Glucose and Urinary Glucose in a Chinese Population： A Multicenter Cross-Sectional Study

Background： Although fasting plasma glucose （FPG） has been highly recommended as the sole test for diabetes screening, the efficacy of FPG alone for diabetes screening is potentially limited due to its low sensitivity. The aim of this study was to improve the efficacy of FPG for diabetes screening using urinary glucose （UG）. Methods： This study was initiated on November 12, 2015, and ended on June 28, 2016. A representative sample of individuals aged between 18 and 65 years, with no history of diabetes, from 6 cities in Jiangsu Province participated in this study. A 75-g oral glucose tolerance test was used to diagnose diabetes. All urine samples were collected within 2 h of oral glucose loading to measure UG. Partial correlation analyses were used to evaluate the associations between UG and other glycemic variables, including FPG, 2-h plasma glucose （2h-PG）, and glycated hemoglobin A 1 c, after adjustment for age. Tile perfbnnance of UG was evaluated using a receiver operating characteristic （ROC） curve analysis. Results： Of the 7485 individuals included, 8% were newly diagnosed with diabetes and 48.7% had prediabetes. The areas under the ROC curves for UG were 0.75 for estimation of 2h-PG ≥7.8 mmol/L and 0,90 for 2h-PG ≥11.1 mmol/L, respectively. The sensitivity and specificity of UG were 52.3% and 87.8%, respectively, for 2h-PG ≥7.8 mmol/L （cutoff point ≥130 mg）, and 83.5% and 87.5%, respectively, for 2h-PG ≥11.1 mmol/L （cutoff point ≥ 178.5 mg）. The combination of FPG and UG demonstrated a significantly higher sensitivity than that of FPG alone for the identification of diabetes （[483/597] 80.9% vs. [335/597] 56.1%, x2 = 85.0, P 〈 0.001） and glucose abnormalities （[2643/4242] 62.3% vs. [2365/4242] 55.8%, x2 = 37.7 P 〈 0.001）. Conclusions： The combination of UG and FPG substantially improves the efficacy of using FPG alone for diabetes screening; this combination might be a practical screening tool and is worth being recommended in the future.