Classifier Using Union Labels for Cross-Modality Dental Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional technologies fail to effectively utilize information from images captured by one imaging modality when viewing images from a different modality, and similarly, they struggle to leverage information from measurements by one type of sensor when presented with data from another type of sensor.
Innovation Solution
The use of union labels, which are calculated by performing a union operation on labels from different imaging modalities or sensor measurements, allowing a classifier to exploit knowledge from one type of image or sensor measurement to classify images or measurements from another type, even if the modalities or sensors are different.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional classification methods are used separately for each imaging modality, then the classification process is simple and straightforward, but the classifier cannot exploit knowledge from one imaging modality to improve classification of another modality
Solution Approach 1:
The patent merges multiple imaging modalities and their corresponding labels into a unified training framework. By combining labels from different modalities (e.g., fluorescent biomarker images and white light images) into joint training sets, the classifier learns to exploit knowledge across modalities, enabling improved classification performance without requiring complex separate processing systems for each modality.
2Measurement precision
If human experts label all images manually, then the labeling accuracy is high, but the time and cost required for labeling increases significantly
Solution Approach 1:
The patent applies partial human expert labeling combined with computer vision algorithms. Instead of requiring human experts to label all images, the system uses computer vision to generate initial labels and applies human expert labeling only to a subset of images (e.g., 10-50% of the training set). This partial action approach maintains high labeling accuracy while significantly reducing the time and cost associated with complete manual labeling.
3Measurement precision
If high-radiation or expensive imaging methods are used for all diagnostic cases, then the diagnostic accuracy is improved, but the cost and patient exposure to radiation increases
Solution Approach 1:
The patent uses a trained classifier as an intermediary to bridge between lower-cost imaging modalities and higher-cost or higher-radiation modalities. The classifier, trained on joint data from multiple modalities including high-radiation CT scans, can infer diagnostic information from lower-radiation MRI or ultrasound images, thereby reducing patient radiation exposure and cost while maintaining diagnostic accuracy through the intermediary classification system.
Data Source
AI summary
A human expert may initially label a white light image of teeth, and computer vision may initially label a filtered fluorescent image of the same teeth. Each label may indicate presence or absence of dental plaque at a pixel. The images may be registered. For each pixel of the registered images, a union label may be calculated, which is the union of the expert label and computer vision label. The union labels may be applied to the white light image. This process may be repeated to create a training set of union-labeled white light images. A classifier may be trained on this training set. Once trained, the classifier may classify a previously unseen white light image, by predicting union labels for that image. Alternatively, the items that are initially labeled may comprise images captured by two different imaging modalities, or may comprise different types of sensor measurements.