Method for enhancing medical report generation by utilizing segmentation model and comparative learning
The invention discloses a method for enhancing medical report generation by using a segmentation model and comparative learning, and the method carries out the fine-grained segmentation of a medical image through the segmentation model SAM, and focuses on a meaningful region which may contain an anomaly. And then image features are extracted from the segmentation, so that the model pays more attention to areas where diseases possibly exist, and the analysis accuracy is improved. In order to relieve the problem of text data prejudice, supervision and comparison losses are introduced in the training process. The loss function encourages the model to distinguish between the target report and the error report, and gives a higher weight to the report which accurately describes the anomaly. By emphasizing comparisons between different report instances, the dominant position of normal area description can be mitigated, and generation of more balanced and information-rich reports can be facilitated..
| Medienart: |
|
|---|
Patent| Erscheinungsjahr: |
2024 |
|---|---|
| Erschienen: |
2024 |
| Enthalten in: |
Europäisches Patentamt - (2024) vom: 30. Jan. Zur Gesamtaufnahme - year:2024 |
|---|
| Sprache: |
Englisch |
|---|
| Beteiligte Personen: |
LI PIJI [VerfasserIn] |
|---|
| Links: |
Volltext [kostenfrei] |
|---|
| Themen: |
Sonstige Themen: |
|---|
| Anmerkungen: |
Source: www.epo.org (no modifications made), First posted: 2024-01-30, Last update posted on www.tib.eu: 2024-05-06, Last updated: 2024-05-10 |
|---|
| Patentnummer: |
CN117476162 |
|---|
| Förderinstitution / Projekttitel: |
|
|---|
| PPN (Katalog-ID): |
EPA002749920 |
|---|
| LEADER | 01000caa a22002652 4500 | ||
|---|---|---|---|
| 001 | EPA002749920 | ||
| 003 | DE-627 | ||
| 005 | 20240510105506.0 | ||
| 007 | cr uuu---uuuuu | ||
| 008 | 240329s2024 xx |||||o 00| ||eng c | ||
| 035 | |a (DE-627)EPA002749920 | ||
| 035 | |a (EPA)CN117476162 | ||
| 035 | |a (EPA)89628131 | ||
| 040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
| 041 | |a eng | ||
| 100 | 0 | |a LI PIJI |e verfasserin |4 aut | |
| 245 | 1 | 0 | |a Method for enhancing medical report generation by utilizing segmentation model and comparative learning |
| 264 | 1 | |c 2024 | |
| 336 | |a Text |b txt |2 rdacontent | ||
| 337 | |a Computermedien |b c |2 rdamedia | ||
| 338 | |a Online-Ressource |b cr |2 rdacarrier | ||
| 500 | |a Source: www.epo.org (no modifications made), First posted: 2024-01-30, Last update posted on www.tib.eu: 2024-05-06, Last updated: 2024-05-10 | ||
| 520 | |a The invention discloses a method for enhancing medical report generation by using a segmentation model and comparative learning, and the method carries out the fine-grained segmentation of a medical image through the segmentation model SAM, and focuses on a meaningful region which may contain an anomaly. And then image features are extracted from the segmentation, so that the model pays more attention to areas where diseases possibly exist, and the analysis accuracy is improved. In order to relieve the problem of text data prejudice, supervision and comparison losses are introduced in the training process. The loss function encourages the model to distinguish between the target report and the error report, and gives a higher weight to the report which accurately describes the anomaly. By emphasizing comparisons between different report instances, the dominant position of normal area description can be mitigated, and generation of more balanced and information-rich reports can be facilitated. | ||
| 650 | 4 | |a G06V | |
| 650 | 4 | |a G06N: Computer systems based on specific computational models | |
| 650 | 4 | |a G16H: Healthcare informatics, i.e. information and communication technology [ict] specially adapted for the handling or processing of medical or healthcare data | |
| 650 | 4 | |a inf | |
| 650 | 4 | |a 615 | |
| 700 | 0 | |a ZHAO RUOQING |e verfasserin |4 aut | |
| 700 | 0 | |a WANG XI |e verfasserin |4 aut | |
| 700 | 0 | |a DAI HONGLIANG |e verfasserin |4 aut | |
| 773 | 0 | 8 | |i Enthalten in |t Europäisches Patentamt |g (2024) vom: 30. Jan. |
| 773 | 1 | 8 | |g year:2024 |g day:30 |g month:01 |
| 856 | 4 | 0 | |u https://worldwide.espacenet.com/patent/search/family/89628131/publication/CN117476162A1?q=CN117476162 |x 0 |z kostenfrei |3 Volltext |
| 912 | |a GBV_EPA | ||
| 951 | |a AR | ||
| 952 | |j 2024 |b 30 |c 01 | ||