An Image Recognition Framework for Oral Cancer Cells.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Oral squamous cell carcinoma (OSCC) presents a significant mortality risk, with late-stage diagnoses common due to limitations in current screening accuracy.
• Early detection of OSCC is crucial for improving treatment outcomes and reducing recurrence rates.
• Integrating image recognition technology can enhance diagnostic accuracy and reduce inter-observer variability in cancer detection.

Purpose of the Study:

• To evaluate the precision and robustness of a deep learning-based approach for automated identification of OSCC extent in digitized oral images.
• To develop and validate a novel method utilizing convolutional neural network (CNN) variants for improved oral cancer cell detection.

Main Methods:

• A deep learning model employing multiple CNN variants (VGG16, VGG19, InceptionV3, InceptionResNetV2, Xception) was developed.
• Image segmentation using multiscale morphology and morphological edge detection was performed for feature extraction (cell area, perimeter).
• The CNN classifier was trained on ImageNet dataset images and independently validated on oral cancer cell images.

Main Results:

• All five CNN variants demonstrated high performance with training and validation losses below 6%.
• The developed method effectively segmented oral cell images and extracted relevant multidimensional features.
• The deep learning approach showed significant potential for accurate OSCC diagnosis.

Conclusions:

• The proposed deep learning-based method, utilizing CNNs and morphological image analysis, is a promising tool for the early and accurate diagnosis of OSCC.
• This technology can assist clinicians in improving diagnostic precision, potentially leading to better patient prognoses.
• Further validation and integration into clinical workflows could enhance OSCC screening and management.