Researchers have developed a novel approach for classifying oral lesions, combining deep learning with spectral analysis and demographic data. The method utilizes a fine-tuned ConvNeXtV2 network for image embeddings, which are then transformed into the hyperspectral domain. From these reconstructed hyperspectral cubes, spectral and textural descriptors are extracted and integrated with demographic information. An incremental heuristic meta learner (IHML) was created to merge calibrated base classifiers, improving diagnostic robustness on small, heterogeneous medical datasets by decoupling evidence extraction from decision fusion. This model achieved a macro F1 score of 66.23% and an overall accuracy of 64.56% on an unseen test set, demonstrating the effectiveness of RGB to hyperspectral reconstruction and ensemble meta-learning for oral lesion screening. AI
IMPACT This research demonstrates a novel approach to medical image analysis, potentially improving early detection of diseases in low-resource settings.
RANK_REASON The cluster contains a research paper detailing a novel AI model for medical image classification. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX
- Connected Papers
- ConvNeXtV2
- DagsHub
- Gotit.pub
- Hugging Face
- IHML
- Litmaps
- ScienceCast
- scite Smart Citations
- Soujanya Hazra
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
