His study makes use of astudy utilizes a U-Net model, which was previously
His study makes use of astudy uses a U-Net model, which was previously developed for sophisticated yses of analyses of organ lesions of biomedical IQP-0528 In Vivo science [36].science is suitable for the organ lesions in the field within the field of biomedical U-Net [36]. U-Net is appropriate for the problem of precisely detecting the shape of the analysis target by simultaneously challenge of precisely detecting the shape on the analysis target by simultaneously learn- learning ing the the worldwide and local info ofimage. The structure of U-NetU-Net was modified to international and local information and facts of the the image. The structure of was modified develop EfficientNet, a neural network that extracts image information. to develop EfficientNet, a neural network that extracts image data. U-Net infers the probability that an arbitrary pixel is definitely an expansion joint device and U-Net infers appropriate answer for every arbitrary pixel right answer masking. For that reason, the learns the the probability that an pixel in the is an expansion joint device and learns the correct error forfor every single pixel from image patchanswer masking. As a result, the Icosabutate supplier prediction answer all pixels M with the the right expressed as BCE is as follows: prediction error for all pixels M from the image patch expressed as BCE is as follows: M L patch L pixel i=1 (-yi log(1i – (1)- yi ) log(1 – Pi )),1,two,three, 1, 2, three, . . .(five) , N (5) == = = (- log – P – log(1 – )) , = i = … , The final cost function J is expressed by summing the imply error for N image patches The final cost function J is expressed by summing the imply error for N image patches and theand regularization term: term: L2 the L2 regularization= 1( N + j | | ) J= L + | w |two N j=1 patch (6)(6)exactly where = 10 . The gradient descent approach updates model parameters in the direction of minimizing where = 10-4 . The J as follows: the price functiongradient descent method updates model parameters inside the direction of minimizing the cost function J as follows: (7) w w – g – (7) where where = 10-4 , g = J . = 10 , = . w We compared the overall performance among U-Net’s masking image image plus the correct We compared the overall performance in between U-Net’s masking as well as the appropriate masking image inside the test the test dataset. Table 7 the pixel-level classification overall performance masking image in dataset. Table 7 shows shows the pixel-level classification overall performance for the test the test pixelthe pixel precision of the expansion joint device was 96.61 ,was rate for set; the set; precision on the expansion joint device was 96.61 , recall price recall 94.38 , and94.38 , and f1-score The f1-score ofThe expansion joint expansion joint device pixel was f1-score was 95.49 . was 95.49 . the f1-score on the device pixel detection was within 5 . In the post-processing process, by correcting the by correctingpredicted of the detection was inside five . In the post-processing method, error of your the error masking image of U-Net, the minimum gap point was detected, plus the distance was measured.Appl. Syst. Innov. 2021, 4,14 ofAppl. Syst. Innov. 2021, 4, x FOR PEER REVIEWpredicted masking image of U-Net, the minimum gap point was detected, and the distance was measured.Table 7. Gap identification accuracy by form of expansion Table 7. Gap identification accuracy by sort of expansion joint (2017019).14 ofAccuracyby Form of Expansion Joint Accuracy by Sort of Expansion Joint Optimistic (pixels expansion joints) Positive (pixels ofof expansion joints) Adverse (other pixels)Unfavorable (other pixels)Pr.