Patent
Hearing state prediction apparatus and method based on diffusion tensor image
| العنوان: | Hearing state prediction apparatus and method based on diffusion tensor image |
|---|---|
| Patent Number: | 12183,002 |
| تاريخ النشر: | December 31, 2024 |
| Appl. No: | 18/634483 |
| Application Filed: | April 12, 2024 |
| مستخلص: | Disclosed are a hearing state prediction apparatus and method based on a diffusion tensor image, including: obtaining a diffusion tensor image to be processed, wherein the diffusion tensor image comprises a diffusion-weighted image; generating a diffusion index image based on the diffusion-weighted image; determining a white matter microstructural feature corresponding to the diffusion tensor image, and determining a hearing state corresponding to the diffusion tensor image according to a mapping relationship between the white matter microstructural feature and the hearing state. The apparatus may generate the diffusion index image through the diffusion-weighted image included in the diffusion tensor image, and then the white matter microstructural feature is determined based on the diffusion index image, so that the white matter microstructural feature can be identified more accurately, and the accuracy of a hearing state evaluation result is improved. Meanwhile, the relation between hearing disorder and brain microstructure change is disclosed. |
| Inventors: | BEIJING FRIENDSHIP HOSPITAL, CAPITAL MEDICAL UNIVERSITY (Beijing, CN) |
| Assignees: | BEIJING FRIENDSHIP HOSPITAL, CAPITAL MEDICAL UNIVERSITY (Beijing, CN) |
| Claim: | 1. A hearing state prediction apparatus based on a diffusion tensor image, comprising: an obtaining module, configured to obtain a diffusion tensor image to be processed, wherein the diffusion tensor image comprises a diffusion-weighted image; a first determining module, configured to generate a diffusion index image according to the diffusion-weighted image; and a second determining module, configured to perform feature extraction on the diffusion index image to obtain a first depth feature; perform segmentation on the diffusion index image to obtain multiple segmented regions; perform feature extraction on the multiple segmented regions to obtain a second depth feature; and input the first depth feature and the second depth feature to a first deep learning model to obtain a hearing state corresponding to the depth feature, and the first deep learning model is trained to determine the hearing state. |
| Claim: | 2. The apparatus according to claim 1 , wherein when the first determining module generates a diffusion index image according to the diffusion-weighted image, it is specifically configured to: determine a diffusion tensor according to the diffusion-weighted image; determine a diffusion eigenvalue according to the diffusion tensor; determine the diffusion index data according to the diffusion eigenvalue; and generate the diffusion index image according to the diffusion index data. |
| Claim: | 3. The apparatus according to claim 1 , wherein when the second determining module determines the white matter microstructural feature corresponding to the diffusion tensor image according to the diffusion index image, it is specifically configured to: perform filtering processing on the diffusion index image to obtain a processed diffusion index image; perform feature extraction on the processed diffusion index image to obtain a third depth feature; perform segmentation on the processed diffusion index image to obtain multiple segmented regions; perform feature extraction on the multiple segmented regions to obtain a fourth depth feature; and determine the white matter microstructure feature corresponding to the diffusion tensor image according to the third depth feature and the fourth depth feature. |
| Claim: | 4. The apparatus according to claim 3 , wherein the second determining module performs feature extraction on the processed diffusion index image to obtain a third depth feature, is specifically configured to: perform analyzing on the processed diffusion index image using a first machine learning model to obtain a depth feature generation graph corresponding to the processed diffusion index image, and the first machine learning model is trained to determine the depth feature generation graph; perform downsampling on the depth feature generation graph to obtain a depth feature graph; extract a gray level co-occurrence matrix based on the depth feature graph; and the gray level co-occurrence matrix is determined as the third depth feature. |
| Claim: | 5. The apparatus according to claim 4 , wherein the first machine learning model comprises an adversarial diffusion model including mixed noise. |
| Claim: | 6. The apparatus according to claim 5 , wherein the second determining module performing feature extraction on the multiple segmented regions to obtain fourth depth feature, is specifically configured to: perform analyzing on the multiple segmented regions using a second machine learning model to obtain fourth depth feature information corresponding to each of the multiple segmented regions, the second machine learning model is trained to obtain the fourth depth feature information corresponding to the segmented regions. |
| Claim: | 7. The apparatus according to claim 3 , the second determining module determining the corresponding diffusion tensor image according to the mapping relationship between the white matter microstructural feature and the hearing state, is specifically configured to: determine the hearing state to which the white matter microstructure belongs according to the white matter microstructural feature and a white matter microstructural feature threshold corresponding to each hearing state. |
| Claim: | 8. The apparatus according to claim 1 , wherein the diffusion index image comprises one or a combination of the following: an anisotropic fraction graph, an average diffusion coefficient graph, an axial diffusion coefficient graph, and a radial diffusion coefficient graph. |
| Claim: | 9. A hearing state prediction method based on a diffusion tensor image, comprising: obtaining a diffusion tensor image to be processed, wherein the diffusion tensor image comprises a diffusion-weighted image; generating a diffusion index image based on the diffusion-weighted image; performing feature extraction on the diffusion index image to obtain a first depth feature; performing segmentation on the diffusion index image to obtain multiple segmented regions; performing feature extraction on the multiple segmented regions to obtain a second depth feature; and inputting the first depth feature and the second depth feature to a first deep learning model to obtain a hearing state corresponding to the depth feature, and the first deep learning model is trained to determine the hearing state. |
| Claim: | 10. A hearing state prediction model training apparatus, comprising: an obtaining module, configured to obtain a diffusion tensor image sample and a preset hearing state corresponding to the diffusion tensor image sample, wherein the diffusion tensor image sample comprises a diffusion-weighted image; a first determining module, configured to generate a diffusion index image according to the diffusion-weighted image; a second determining module, configured to perform feature extraction on the diffusion index image to obtain a first depth feature; perform segmentation on the diffusion index image to obtain multiple segmented regions; perform feature extraction on the multiple segmented regions to obtain a second depth feature; and input the first depth feature and the second depth feature to a first deep learning model to obtain a hearing state corresponding to the depth feature, and the first deep learning model is trained to determine the hearing state; and generating module, configured to determine a consistency between the hearing state and the preset hearing state corresponding to the diffusion tensor image sample, when the consistency between the hearing state and the preset hearing state is greater than or equal to a preset threshold, generating a hearing state prediction model. |
| Claim: | 11. A training method for a hearing state prediction model, comprising: obtaining a diffusion tensor image sample and a preset hearing state corresponding to the diffusion tensor image sample, wherein the diffusion tensor image sample comprises a diffusion-weighted image; generating a diffusion index image according to the diffusion-weighted image; performing feature extraction on the diffusion index image to obtain a first depth feature; performing segmentation on the diffusion index image to obtain multiple segmented regions; performing feature extraction on the multiple segmented regions to obtain a second depth feature; and inputting the first depth feature and the second depth feature to a first deep learning model to obtain a hearing state corresponding to the depth feature, and the first deep learning model is trained to determine the hearing state; and determining a consistency between the hearing state and the preset hearing state corresponding to the diffusion tensor image sample, when the consistency between the hearing state and the preset hearing state is greater than or equal to the preset threshold, a hearing state prediction model is generated. |
| Patent References Cited: | 6853189 February 2005 Pipe 8059879 November 2011 Tsukimoto 11393087 July 2022 Parikh 2021/0082135 March 2021 Xu 2024/0075320 March 2024 Rezai 2024/0185498 June 2024 Francis 2024/0221162 July 2024 Xie 101287410 October 2008 107330267 November 2017 |
| Other References: | Brown, Colin J., and Ghassan Hamarneh. “Machine learning on human connectome data from MRI.” arXiv preprint arXiv:1611.08699 (2016). (Year: 2016). cited by examiner CN Office Action dated Sep. 1, 2023 as received in Application No. 202310436275.9. cited by applicant |
| Primary Examiner: | Safaipour, Bobbak |
| Attorney, Agent or Firm: | Maschoff Brennan |
| رقم الانضمام: | edspgr.12183002 |
| قاعدة البيانات: | USPTO Patent Grants |
| الوصف غير متاح. |