Deep learning-based diagnosis of lung cancer using a nationwide respiratory cytology image set: improving accuracy and inter-observer variability.

Deep learning (DL)-based image analysis has recently seen widespread application in digital pathology. Recent studies utilizing DL in cytopathology have shown promising results, however, the development of DL models for respiratory specimens is limited. In this study, we designed a DL model to improve lung cancer diagnosis accuracy using cytological images from the respiratory tract. This retrospective, multicenter study used digital cytology images of respiratory specimens from a quality-controlled national dataset collected from over 200 institutions. The image processing involves generating extended z-stack images to reduce the phase difference of cell clusters, color normalizing, and cropping image patches to 256 × 256 pixels. The accuracy of diagnosing lung cancer in humans from image patches before and after receiving AI assistance was compared. 30,590 image patches (1,273 whole slide images [WSIs]) were divided into 27,362 (1,146 WSIs) for training, 2,928 (126 WSIs) for validation, and 1,272 (1,272 WSIs) for testing. The Densenet121 model, which showed the best performance among six convolutional neural network models, was used for analysis. The results of sensitivity, specificity, and accuracy were 95.9%, 98.2%, and 96.9% respectively, outperforming the average of three experienced pathologists. The accuracy of pathologists after receiving AI assistance improved from 82.9% to 95.9%, and the inter-rater agreement of Fleiss' Kappa value was improved from 0.553 to 0.908. In conclusion, this study demonstrated that a DL model was effective in diagnosing lung cancer in respiratory cytology. By increasing diagnostic accuracy and reducing inter-observer variability, AI has the potential to enhance the diagnostic capabilities of pathologists.

Comparing the screening property of the shortened versions of the Hypomania Checklist-32 (HCL-32): Cross-validation in Korean patients with bipolar disorder and major depressive disorder.

BACKGROUND: To date, shortened versions of the Hypomania Checklist-32 (HCL-32) were proposed to overcome the limitation of a lengthy format; however, a cross-validation study is currently needed to identify which shorter version may function optimally in a clinical sample. METHODS: In a Korean patient sample with bipolar disorder (BD) and major depressive disorder (MDD) (BD-I n = 84, BD = II n = 145, MDD n = 285), we examined the reliability and screening property of three shorter versions of the HCL (HCL-20, -16, -8) in comparison with the full HCL-32. Diagnosis was confirmed by the structured clinical interview (SCID-I). RESULTS: All three shortened HCLs demonstrated a fair screening ability (Area Under the Curve = .72~.74) to discriminate BD patients from MDD patients, which was comparable to that of the HCL-32. When sensitivity and specificity were considered, the HCL-20 showed relatively superior performance among the shortened versions. LIMITATIONS: The shorter versions were not administered in a 'stand-alone' manner. CONCLUSIONS: This is the first cross-validation study in a large clinical sample with an increased statistical power to compare the screening property of the shortened HCLs. Our results suggest that briefer versions of the HCL could be reliably and economically utilized in clinical and research settings to enhance detection of BD.

Effect of atmospheric pressure plasma on inactivation of pathogens inoculated onto bacon using two different gas compositions.

Atmospheric pressure plasma (APP) is an emerging non-thermal pasteurization method for the enhancement of food safety. In this study, the effect of APP on the inactivation of pathogens inoculated onto bacon was observed. Sliced bacon was inoculated with Listeria monocytogenes (KCTC 3596), Escherichia coli (KCTC 1682), and Salmonella Typhimurium (KCTC 1925). The samples were treated with APP at 75, 100, and 125 W of input power for 60 and 90 s. Two gases, helium (10 lpm) or a mixture of helium and oxygen, (10 lpm and 10 sccm, respectively) were used for the plasma generation. Plasma with helium could only reduce the number of inoculated pathogens by about 1-2 Log cycles. On the other hand, the helium/oxygen gas mixture was able to achieve microbial reduction of about 2-3 Log cycles. The number of total aerobic bacteria showed 1.89 and 4.58 decimal reductions after plasma treatment with helium and the helium/oxygen mixture, respectively. Microscopic observation of the bacon after plasma treatment did not find any significant changes, except that the L∗-value of the bacon surface was increased. These results clearly indicate that APP treatment is effective for the inactivation of the three pathogens used in this study, although further investigation is needed for elucidating quality changes after treatment.