AI image diagnosis device and dental OCT image diagnosis device
An AI image diagnosis apparatus into which three-dimensional tooth image data captured by a dental OCT device is inputted and which analyzes the inputted three-dimensional tooth image data, comprising: a model executor that sequentially inputs two-dimensional tomographic image data constituting the diagnostic target three-dimensional tooth image data into a trained model, to thereby obtain, as an execution processing result of the trained model, lesion information data, which is data related to a part identified as a characteristic part such as a lesion in the inputted tomographic image data, for each piece of tomographic image data in the three-dimensional tooth image data, and detect lesions from the inputted three-dimensional tooth image data using the obtained lesion information data, wherein the trained model is constructed by training of three-dimensional tooth image data of multiple examinees captured by a dental OCT device in the past.
1 . An AI image diagnosis apparatus into which three-dimensional tooth image data captured by a dental Optical Coherence Tomography (OCT) device is input and which analyzes the input three-dimensional tooth image data, the apparatus comprising:
a processor configured to execute functions of:
a model executor that sequentially inputs two-dimensional tomographic image data constituting the diagnostic target three-dimensional tooth image data into a trained model to output, for each tomographic image, lesion information including at least an image representing per-pixel degree of matching to a lesion class and that reconstructs a three-dimensional volume based on images representing per-pixel degrees of matching; and
a display controller that superimposes the reconstructed three-dimensional volume on the input three-dimensional tooth image data for display, wherein lesions are detected in the three-dimensional tooth image data based on the lesion information,
wherein the trained model is constructed by machine learning using training data, and
wherein training data includes at least one of
A-plane tomographic image data which is parallel to a plane specified by both a B-axis direction orthogonal to the A-axis direction, which is an irradiation direction of OCT laser light on the tooth, and the A-axis direction,
L-plane tomographic image data which is parallel to a plane specified by both a V-axis direction, orthogonal to the A-axis direction and the B-axis direction, and the A-axis direction,
S-plane tomographic image data which is parallel to a plane specified by both the B-axis direction and the V-axis direction,
en-face image data synthesized from information on a surface of the tooth irradiated with the OCT laser light and information on the A-axis direction, and
three-dimensional image data composed of several consecutive pieces of tomographic image data.
2 . The AI image diagnosis apparatus according to claim 1 , wherein pixels with a degree of matching less than a predetermined threshold are converted to zero before the reconstructing.
3 . The AI image diagnosis apparatus according to claim 2 , wherein the predetermined threshold is adaptively determined based on statistics of the degrees of matching across tomographic positions for each lesion class within the input three-dimensional tooth image data.
4 . The AI image diagnosis apparatus according to claim 1 , wherein
the two-dimensional tomographic image data sequentially inputted to the trained model by the model executor is the tomographic image data constituting the three-dimensional tooth image data being captured by the dental OCT device.
5 . The AI image diagnosis apparatus according to claim 1 , wherein
the model executor reconstructs class-specific three-dimensional detection volumes from the images representing per-pixel degrees of matching, and
the display controller superimposes a selected class-specific volume on the input three-dimensional tooth image data.
6 . The AI image diagnosis apparatus according to claim 1 , wherein the display controller lists detected items as selectable entries and, upon user selection, highlights lesion boundaries with a color-coded overlay while displaying corresponding tomographic images.
7 . The AI image diagnosis apparatus according to claim 1 , wherein the training data further includes tomographic images containing lesion characteristic parts and tomographic images containing non-lesion characteristic parts captured by a dental OCT device.
8 . The AI image diagnosis apparatus according to claim 7 , wherein the training data further includes labels indicating at least a lesion label and a tooth type.
9 . An AI image diagnosis apparatus into which three-dimensional tooth image data captured by a dental Optical Coherence Tomography (OCT) device is input and which analyzes the input three-dimensional tooth image data, the apparatus comprising:
a processor configured to execute functions of:
a model executor that sequentially inputs two-dimensional tomographic image data into a trained model to obtain, for each tomographic image, lesion information including at least a lesion class name, a center position of a point having the highest degree of similarity, and a degree of similarity and that graphs, for each lesion class, a relationship between tomographic positions and degrees of similarity based on the lesion information, identifies, continuous regions in which the degree of similarity exceeds a predetermined threshold over consecutive tomographic positions, determines, a region name for each identified region, associates, the lesion class name, a tomographic position of the highest degree of similarity within the region, and the corresponding center position with the region name to generate a lesion detection result, and that reconstructs a three-dimensional similarity volume from images representing per-pixel degrees of matching output from the trained model; and
a display controller that superimposes the reconstructed volume on the input three-dimensional tooth image data for display, and, upon user selection of the region name, extracts and displays tomographic image data corresponding to the selected region,
wherein the trained model is constructed by machine learning using training data, and
wherein the training data includes at least one of
A-plane tomographic image data which is parallel to a plane specified by both a B-axis direction orthogonal to the A-axis direction, which is an irradiation direction of OCT laser light on the tooth, and the A-axis direction,
L-plane tomographic image data which is parallel to a plane specified by both a V-axis direction, orthogonal to the A-axis direction and the B-axis direction, and the A-axis direction,
S-plane tomographic image data which is parallel to a plane specified by both the B-axis direction and the V-axis direction,
en-face image data synthesized from information on a surface of the tooth irradiated with the OCT laser light and information on the A-axis direction, and
three-dimensional image data composed of several consecutive pieces of tomographic image data.
10 . The AI image diagnosis apparatus according to claim 9 , wherein pixels with a degree of similarity less than the predetermined threshold are converted to zero prior to the reconstructing.
11 . The AI image diagnosis apparatus according to claim 10 , wherein the predetermined threshold is adaptively determined based on statistics of the degrees of similarity across tomographic positions for each lesion class within the input three-dimensional tooth image data.
12 . The AI image diagnosis apparatus according to claim 9 , wherein
the model executor accepts an instruction of either horizontal scanning or vertical scanning by user operation, and upon acceptance of the instruction of horizontal scanning, inputs the A-plane tomographic image data to the trained model, and upon acceptance of the instruction of vertical scanning, inputs the L-plane tomographic image data to the trained model, and executes arithmetic processing of the trained model.
13 . The AI image diagnosis apparatus according to claim 9 , wherein
the model executor determines scanning direction information of the dental OCT device when imaging, which is included in the inputted three-dimensional tooth image data, and upon determination that the inputted three-dimensional tooth image data includes horizontal scanning information, inputs the A-plane tomographic image data to the trained model, and upon determination that the inputted three-dimensional tooth image data includes vertical scanning.
14 . A dental OCT image diagnosis apparatus comprising: the AI image diagnosis apparatus according to claim 9 installed in a dental OCT device.