Implementation of transfer learning for the segmentation of human mesenchymal stem cells—A validation study

Adnan, Niha; Umer, Fahad; Malik, Shahzaib

doi:10.1016/j.tice.2023.102149

Cited by 4 publications

(4 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The pixel-wise segmentation of a feature in an image is considered to be a superior method for localising of the region of interest. [5][6][7] It has the advantage of separating the background in the image and each pixel of the feature J Pak Med Assoc (Suppl. 4)…”

Section: Discussionmentioning

confidence: 99%

“…To improve its performance, further training on a dataset from various centres need to be included. 6 The number of images used for training was relatively smaller and to mitigate this issue, image augmentation techniques were applied to inflate the total number of images from 300 to 1294. This still does not overcome limited generalisability of this model and a wide variety of images exposed from different machines are still required.…”

Section: S-7mentioning

confidence: 99%

“…5 Deep Learning (DL) models are dominating the field of computer vision especially in the medical and dental fields, aiding clinicians in diagnoses and treatment planning. 6 DL is a subset of AI that consist of neural networks that allow these models to train independently and learn various features with limited need for manual human input. 7 Various diagnostic DL models exist that aid dentists in teeth segmentation and numbering as well pathology detection on different types of radiographs used in clinical practice.…”

Section: Introductionmentioning

confidence: 99%

“…7 A more advanced technique that allows precise delineation of the region of interest is segmentation. [5][6][7] Therefore, for better localisation of implants via segmentation is better suited that identifies each pixel associated with the implant fixture for a more robust diagnostic process. 7 There is lack of literature on DL models trained for implant detection on PA radiographs and this study aims to detect implants by performing pixel-wise segmentation.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

An Artificial Intelligence model for implant segmentation on periapical radiographs

Niha Adnan,

Muhammad Hanif,

Khurram Khan

et al. 2024

J Pak Med Assoc

View full text Add to dashboard Cite

Objective: To segment dental implants on PA radiographs using a Deep Learning (DL) algorithm. To compare theperformance of the algorithm relative to ground truth determined by the human annotator.Methodology: Three hundred PA radiographs were retrieved from the radiographic database and consequentlyannotated to label implants as well as teeth on the LabelMe annotation software. The dataset was augmented toincrease the number of images in the training data and a total of 1294 images were used to train, validate and testthe DL algorithm. An untrained U-net was downloaded and trained on the annotated dataset to allow detection ofimplants using polygons on PA radiographs.Results: A total of one hundred and thirty unseen images were run through the trained U-net to determine itsability to segment implants on PA radiographs. The performance metrics are as follows: accuracy of 93.8%, precisionof 90%, recall of 83%, F-1 score of 86%, Intersection over Union of 86.4% and loss = 21%.Conclusion: The trained DL algorithm segmented implants on PA radiographs with high performance similar tothat of the humans who labelled the images forming the ground truth.Keywords: Deep Learning, Dental Implants, Algorithms, dentistry, Neural Networks, Intraoral Radiography

show abstract

Section: Discussionmentioning

confidence: 99%

Section: S-7mentioning

confidence: 99%