Predictive dual machine translation

Patent metadata

Field	Value
Publication number	US-11699514-B2
Application number	US-202016885219-A
Country	US
Kind code	B2
Filing date	May 27, 2020
Priority date	May 27, 2020
Publication date	Jul 11, 2023
Grant date	Jul 11, 2023

Abstract

Dual machine translators are trained by generating a translated medical image by operation of an illustrative model on an original medical record, generating information based on whether the translated medical image is natural in a modality of medical imaging, producing a back-translated medical record by operation of an interpretive model on the translated medical image, calculating a reward by comparing the back-translated medical record to the original medical record, updating parameters of the illustrative model in response to the information and the reward, and updating parameters of the interpretive model in response to the reward.

First claim

What is claimed is:

1. An apparatus comprising:
a memory embodying computer executable instructions; and
at least one processor, coupled to the memory, and operative by the computer executable instructions to facilitate a method of:
generating a translated medical image by operation of an illustrative model on an original medical record;
generating information based on whether the translated medical image is natural in a modality of medical imaging;
producing a back-translated medical record by operation of an interpretive model on the translated medical image;
calculating a reward by comparing the back-translated medical record to the original medical record;
updating parameters of the illustrative model in response to the information and the reward;
updating parameters of the interpretive model in response to the reward;
training the illustrative model and the interpretive model by repeatedly generating, producing, calculating, and updating until the parameters of the illustrative model and the parameters of the interpretive model converge; and
generating a plurality of translated medical images by operation of the trained illustrative model on a plurality of original medical records.

2. The apparatus of claim 1 wherein the method further comprises producing a prognostic medical record by operation of a text-based disease progression model on the original medical record.

3. The apparatus of claim 2 wherein the method further comprises producing a prognostic medical image by operation of the illustrative model on the prognostic medical record.

4. An apparatus comprising:
a memory embodying computer executable instructions; and
at least one processor, coupled to the memory, and operative by the computer executable instructions to facilitate a method of:
generating a translated medical image by operation of an illustrative model on an original medical record;
generating information based on whether the translated medical image is natural in a modality of medical imaging;
producing a back-translated medical record by operation of an interpretive model on the translated medical image;
calculating a reward by comparing the back-translated medical record to the original medical record;
updating parameters of the illustrative model in response to the information and the reward;
updating parameters of the interpretive model in response to the reward;
training the illustrative model and the interpretive model by repeatedly generating, producing, calculating, and updating until the parameters of the illustrative model and the parameters of the interpretive model converge; and
producing a translated medical record by operation of the trained interpretive model on an original medical image, then producing a prognostic medical record by operation of a text-based disease progression model on the translated medical record.

5. An apparatus comprising:
a memory embodying computer executable instructions; and
at least one processor, coupled to the memory, and operative by the computer executable instructions to facilitate a method of:
generating a translated medical image by operation of an illustrative model on an original medical record;
generating information based on whether the translated medical image is natural in a modality of medical imaging;
producing a back-translated medical record by operation of an interpretive model on the translated medical image;
calculating a reward by comparing the back-translated medical record to the original medical record;
updating parameters of the illustrative model in response to the information and the reward;
updating parameters of the interpretive model in response to the reward;
training the illustrative model and the interpretive model by repeatedly generating, producing, calculating, and updating until the parameters of the illustrative model and the parameters of the interpretive model converge; and
producing a translated medical record by operation of the trained interpretive model on an original medical image, then producing a prognostic medical record by operation of a text-based disease progression model on the translated medical record.

Assignees

• IBM

Inventors

• Kozloski James R
• Gurev Viatcheslav
• Parikh Jaimit
• Di Achille Paolo
• Shahn Zachary
• Sow Daby

Classifications

• G16H50/30

  for calculating health indices; for individual health risk assessment · CPC title

• G06F18/2133

  based on naturality criteria, e.g. with non-negative factorisation or negative correlation · CPC title

• G16H10/60

  for patient-specific data, e.g. for electronic patient records · CPC title

• G06N20/00

  Machine learning · CPC title

• G06V10/82

  using neural networks · CPC title