I have Images of the medical lab reports captured by the android device. All Images contain text data. I want to summarize the each and every report. summarization means for example in the USG report , the summary is like "No abnormality seen"., I want this text for each report . The structure of the model: Input is Image and gives summary in terms of text . I thought of Image captioning , but there is lot of similarities in the images. How can I do it ?
any ideas ?
What I can make out from your question is that you are looking for some kind of OCR technique. There are various available in the literature, using which you can convert the image into text, and then apply some NLP algorithms to caption each image as normal and abnormal. If these are medical lab reports, and are captured in DICOM format (structured report), the textual information can be directly derived from the .dcm file by accessing the relevant data elements. Or else if it is any kind of secondary capture image (in DICOM format again), some textual information might be present in the metadata section. Anyway for DICOM SC images too, OCR will be required to derive the information.
I hope this is in line with what you asked for.
The Dicom format allow this. Matlab have a package that's allow an easy way to manipulate it!
I don't have dicom format actually , I have report like Urine analysis , Allergy test . THe Images are captured using the android device.
Try to change the direction of Image captioning to avoid/reduce the similarities.
- Badges
- Science topic