Welcome to Medisha.com OCR API Preview

This site provides a public preview of transcription and analytics capabilities from Medisha.com that are used for building custom data analytics pipelines from medical records.

Typical pipeline would start from medical documents which are then:

• Text Recognized and Extracted i.e. Transcription / OCR / HTR
• Extracted Text for each page is Analyzed and summarized
• Document summaries are assembled and further analyzed into an analytical report
• The analytical output powers further operational workflow

Preview on this site demonstrates the capabilities of initial Transcription and Summarization using our API backend

• Upload your handwritten/printed Medical document image (scan or smartphone) after visiting the upload page link below.
• The Result page will periodically monitor the status of your request and display the result once ready.
• Output: AI generated Summary of the document followed by raw text as extracted by our OCR trained on handwritten prescriptions and reports.

Currently the web based preview of API capabilities is free and rate limited. Subscription to a connectable API service will be available shortly.

PS: Our text extraction accuracy averages 95% per character on handwritten medical documents that we have tested on (5% CER, ~25% Word Error Rate), and is ~100% (<1% CER) on printed medical documents and reports. All training and testing is done in English Language, in Indian healthcare context. Documents have text and handwriting readable by trained medical professionals. Your mileage may vary depending on handwriting legibility, lighting, image quality and vocabulary used. Expect this to improve with time as we are always working on enhancements.

Usage: This preview service is available for free and as-is. We do not claim that it is fit for use for any purpose. Your use of these services is at your own risk.
Research Use: We will like to use the documents uploaded on this screen to help improve our service (e.g. Handwriting training) in an aggregated manner without identifying individuals. However for extra precaution, we would still recommend that you hide personal and sensitive details like name, phone number, etc. before uploading.
Privacy: We track the IP Address and use Cookies to ensure that there is no spamming or abuse of this service.
Data Security: For protection of your privacy, beyond general best practices, we follow these additional data security practices:

• Transcription results can only be accessed from the system browser and session the image was uploaded from.
• Results expire after 48 hours so invertant access later from the same system is not possible.
• All communication with our servers is via SSL/TLS-1.2+.
• Our servers and data storage are on facilities certified ISO-27001 compliant and are situated within India.

DPDP Compliance: We certify that our services are compliant with the requirements of DPDP Act. Any PHI (Personal Health Information) submitted within the images is incidental and is not specifically retained. The images themselves are used for anonymized research as a permitted exception under the act. We do not link the images uploaded to any identified user and no individual level tracking or analysis is conducted by us.