Kidney Lesion Malignancy Classification Using 3D Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for classifying malignancy risks of renal lesions, particularly cystic and solid renal masses, are ambiguous and inconsistent, leading to suboptimal patient outcomes due to user variability and lack of incorporation of sophisticated imaging features, which can result in misclassification and inappropriate management.
Innovation Solution
A computer-implemented method and system using a deep learning approach with multiple neural networks for segmenting kidney regions, detecting lesions, and classifying malignancy risk, incorporating imaging and histopathologic data to enhance accuracy and reduce false positives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional imaging criteria (Bosniak classification) are used to classify renal lesions, then the classification process is simple and quick, but the classification accuracy is low and user variability is high
Solution Approach 1:
The patent applies segmentation by dividing the renal mass into multiple tissue types (cystic components, solid components, enhancing components, non-enhancing components) and analyzing each segment separately. This allows for precise characterization of complex renal lesions by evaluating specific features of each tissue type rather than relying on overall impression, thereby improving classification accuracy while maintaining manageable system complexity through structured analysis.
Solution Approach 2:
The patent transitions from traditional 2D imaging assessment to 3D volumetric analysis of renal lesions. By calculating volumes of different tissue components and using three-dimensional rendering, the system captures spatial relationships and morphological features that are not visible in conventional 2D images, significantly improving classification precision without excessive complexity increase.
2Measurement precision
If sophisticated imaging features are incorporated to improve classification accuracy, then the malignancy risk assessment becomes more accurate, but the analysis time and computational resources increase
Solution Approach 1:
The patent performs preliminary segmentation and characterization of renal lesions into distinct tissue types before conducting the actual malignancy risk assessment. By pre-identifying and quantifying cystic vs. solid components, enhancing vs. non-enhancing areas, and calculating volume ratios in advance, the system prepares structured data that speeds up the final classification decision, reducing analysis time while maintaining high accuracy.
Solution Approach 2:
The patent transforms complex imaging data into simplified quantitative parameters such as volume ratios (cystic volume/total volume, enhancing volume/total volume), density measurements, and texture features. These derived parameters capture essential diagnostic information in a compact form that can be processed quickly by classification algorithms, improving accuracy without proportionally increasing analysis time.
3Reliability
If manual examination by physicians is used, then flexibility in interpretation is maintained, but consistency and reliability of classification are poor
Solution Approach 1:
The patent implements feedback mechanisms where the automated classification system provides structured results including confidence scores and detailed feature breakdowns to physicians. This allows physicians to review and adjust classifications when necessary while benefiting from the consistency of automated analysis for routine cases, achieving high reliability without excessive automation complexity.
Solution Approach 2:
The patent introduces an intermediate structured reporting layer between the imaging data and the final classification decision. The system generates standardized reports with quantified features (volume ratios, enhancement patterns, texture metrics) that serve as an objective basis for physician decision-making, improving consistency while maintaining the flexibility of human judgment through the intermediary structured format.
Data Source
Figure 1~2
Figure 3~4
AI summary
The invention relates to a computer-implemented method for classifying a malignancy risk of a kidney, in particular a human kidney. The method comprises the steps of providing imaging data of an anatomy of a subject patient, wherein the imaging data comprises at least partially a representation of a kidney of the subject patient, using a first neural network to segment at least one region of the kidney representation which is based on the imaging data, using a second neural network to detect one or more suspected lesions of the segmented kidney representation and classifying the detected suspected lesion with a malignancy risk using a third neural network, wherein the third neural network is a deep profiler.