IADR Abstract Archives

A Novel AI- Based Method as a Prognosticator of the Biological Behavior of Oral Lesions

Objectives: To evaluate the efficacy of AI based analysis of the intensity of staining in oral hyalinizing lesions to predict their biologic behaviour.
To evaluate the prognosticator value of hyalinization in oral mucosal lesions
Methods: In this hospital based, retrospective study, 50 histopathologically diagnosed premalignant disorders (n-25) and oral squamous cell carcinoma (n=25). Control: 25 normal tissues. The paraffin- embedded tissue block were retrieved from the archives. Two sections 5µm thickness each were made and mounted on the glass slide. The slides were deparaffinised, rehydrated and subjected to Herovici's polychrome differential staining technique.
The stained slides were viewed under 40x magnification in the microscope and images of 6 fields were captured using a 2MP camera attached the microscope. The images were classified, transferred and stored in the computer for image analysis.
Image analysis: The hyalinization present in these histopathologic slides were then analysed for the histological content of type I and III collagen both manually and using an image processing software (ImageJ).
The collagen content was then corelated with the histopathologic grade and compared with the grading given manually.
Results: Compared to standard histologic evaluation, Differential staining using herovici’s modified staining procedure and evaluation using computer assisted image analysis was capable of identifying and quantifying even the more mildly stained areas that are usually missed by the human eye. This makes it a more reliable method of analysis as opposed to the traditional method that is prone to error.
Conclusions: Classification based on the features viewed in the pathologic slide for the purpose of treatment and prognosis is the ultimate goal of histopathologic examination. Differential staining is an inexpensive and highly reproducible method of evaluating the composition of the stroma by means of stain intensity, and pattern, which can in turn be correlated to the aggressiveness of the lesion. Using artificial intelligence to carry out this analysis makes the process more objective and renders it bias free.

2021 South East Asian Division Meeting (Hong Kong)
Hong Kong
2021
104
Oral Medicine & Pathology Research
  • S, Samyukta  ( Chettinad dental college and research institute , Chennai , India )
  • NONE
    Poster Session
    Regenerative dentistry and craniofacial biology III
    Thursday, 12/09/2021 , 12:00PM - 01:00PM