System and method for an automated parsing pipeline for anatomical localization and condition classification
An automated parsing pipeline system and method for anatomical localization and condition classification is disclosed. The system comprises an input even source, a memory unit and processor including a volumetric image processor, a voxel parsing engine, localization layer and a detection module. The volumetric image processor is configured to receive volumetric image from the input source and parse the received volumetric image. The voxel parsing engine is configured to assign each voxel a distant anatomical structure. The localization layer is configured to crop a defined anatomical structure with surroundings. The detection module is configured to classify conditions for each defined anatomical structure within the cropped image. The disclosed system and method provide accurate localization of a tooth and detects several common conditions in each tooth.
1. An automated parsing pipeline system for anatomical localization and condition classification, said system comprising:
a processor;
a non-transitory storage element coupled to the processor;
encoded instructions stored in the nor-transitory storage element, wherein the encoded instructions when implemented by the processor, configure the automated parsing pipeline system to:
receive at least one volumetric image;
parse the received volumetric image into at least a single image frame field of view;
localize a present tooth inside the parsed volumetric image and identifying it by number;
extract the identified tooth and surrounding context within the localized volumetric image; and
classify a tooth's conditions based on the extracted volumetric image using at least one of a multi-task approach, one network per condition approach, or a sub-network approach, wherein the multi-task approach is a single network outputting a prediction for multiple tooth conditions, the one network per condition approach is a single network outputting a prediction for a single tooth condition, and the sub-network approach is multiple networks outputting predictions for multiple tooth conditions.
2. The system of claim 1 , wherein the at least one received volumetric image comprises a 3-D pixel array.
3. The system of claim 2 , further configured to pre-process by converting the 3-D pixel array into an array of Hounsfield Unit (HU) radio intensity measurements.
4. The system of claim 1 , further configured to pre-process at least one of the localization or classification steps by rescaling using linear interpolation.
5. The system of claim 1 , wherein the pre-processing comprises using any one of a normalization schemes to account for variations in image value intensity depending on at least one of an input or output of volumetric image.
6. The system of claim 1 , wherein the localization is achieved using a V-Net-based fully convolutional neural network.
7. The system of claim 1 , further configured to extract anatomical structure by finding a minimum bounding rectangle around the localized and identified tooth.
8. The system of claim 7 , wherein the bounding rectangle extends by at least 15 mm vertically and 8 mm horizontally (equally in all directions) to capture the tooth and surrounding context.
9. The system of claim 1 , wherein the classification is achieved using a DenseNet 3-D convolutional neural network.
10. A method for localizing a tooth and classifying a tooth condition, said method comprising the steps of:
receiving at least one volumetric image;
parsing the received volumetric image into at least a single image frame field of view;
localizing a present tooth inside the parsed volumetric image and identifying it by number;
extracting the identified tooth and surrounding context within the localized volumetric image; and
classify a tooth's conditions based on the extracted volumetric image using at least one of a multi-task approach, one network per condition approach, or a sub-network approach, wherein the multi-task approach is a single network outputting a prediction for multiple tooth conditions, the one network per condition approach is a single network outputting a prediction for a single tooth condition, and the sub-network approach is multiple networks outputting predictions for multiple tooth conditions.
11. The method of claim 10 , wherein the at least one received volumetric image comprises a 3-D pixel array.
12. The method of claim 10 , further includes a step of: pre-processing by converting the 3-D pixel array into an array of Hounsfield Unit (HU) radio intensity measurements.
13. The method of claim 10 , wherein the pre-processing for at least one of the localization or classification steps comprises rescaling using linear interpolation.
14. The method of claim 10 , wherein the pre-processing comprises using any one of a normalization schemes to account for variations in image value intensity depending on at least one of an input or output of volumetric image.
15. The method of claim 10 , wherein the localization is achieved using a V-Net-based fully convolutional neural network.
16. The method of claim 10 , further comprises a step of: achieving extraction by finding a minimum bounding rectangle around the localized and identified tooth.
17. The method of claim 16 , wherein the bounding rectangle extends by at least 15 mm vertically and 8 mm horizontally (equally in all directions) to capture the tooth and surrounding context.
18. The method of claim 10 , wherein the classification is achieved using a DenseNet 3-D convolutional neural network.