Machine-Learned Prognosis Determination Using Interpolated Treatment Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Medical image diagnosis often requires specialist interpretation, making it difficult to determine an appropriate treatment method and prognosis without expert input.
Innovation Solution
A prognosis determination device and method that utilizes a discriminator trained through machine learning on actual and interpolated treatment data, including medical images and biological parameters, to predict treatment outcomes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If specialist interpretation is used for medical image diagnosis, then diagnostic accuracy is improved, but device complexity and operational requirements increase
Solution Approach 1:
The patent creates a virtual copy of specialist diagnostic capability through machine learning models trained on actual treatment data and typical models. The discriminator learns to replicate specialist decision-making by processing medical images and biological parameters, producing prognosis predictions that mirror expert judgment without requiring actual specialists to be present in the system
Solution Approach 2:
The patent replaces the mechanical system of specialist interpretation with an automated information processing system. The machine learning-based discriminator substitutes human expert analysis with algorithmic processing of medical images and biological parameters, transforming the diagnostic process from human-centric to system-centric while maintaining diagnostic accuracy
2Reliability
If actual treatment data only is used for training, then data authenticity is improved, but training data quantity is insufficient
Solution Approach 1:
The patent merges actual treatment data with typical models to create a hybrid training dataset. The typical models provide additional structured information about treatment outcomes and disease progression, complementing the real-world variability captured in actual treatment data. This combination increases the total training data quantity while maintaining reliability through the grounding in authentic clinical cases
Solution Approach 2:
The patent performs preliminary data preparation by creating typical models from existing actual treatment data before the main training process. These typical models serve as pre-processed templates that capture common treatment patterns and outcomes, which are then integrated with additional actual treatment data to form a comprehensive training dataset, ensuring sufficient quantity while maintaining authenticity
Data Source
AI summary
A prognosis determination device including an acquisition unit that acquires a medical image and a biological parameter that are related to a disease of a subject; and an output unit that inputs, to a discriminator, information on the medical image, information on the biological parameter, and information on treatment to be performed, and causes the discriminator to output information on a prognosis for a case where the treatment is performed, in which the discriminator has been generated through a machine learning process based on actual treatment data and interpolation treatment data, the actual treatment data is data on the treatment that has been actually performed, the actual treatment data including a pretreatment medical image as well as a pretreatment biological parameter of a patient, information on the treatment performed on the patient, and information on a prognosis for the patient after the treatment, and the interpolation treatment data is information for interpolating a posttreatment prognosis for the patient, the interpolation treatment data including information generated from the actual treatment data and a typical model obtained when the treatment is performed.