AI in Diagnosing Miscarriages
Author Information
Author(s): Mahmood Tahir, Ullah Zeeshan, Latif Atif, Sultan Binish Arif, Zubair Muhammad, Ullah Zahid, Ansari AbuZar, Zehra Talat, Ahmed Shahzad, Dilshad Naqqash, Chen Dechang
Primary Institution: Atia General Hospital, Karachi, Pakistan
Hypothesis
Can AI techniques improve the accuracy and reliability of histopathological examination of tissue phenotypes in products of conception after spontaneous abortion?
Conclusion
AI techniques can enhance the accuracy and reliability of diagnosing tissue phenotypes in miscarriages compared to traditional methods.
Supporting Evidence
- AI techniques can reduce subjectivity and inter-observer variability in histopathological examinations.
- The dataset HistoPoC is the first of its kind focusing on spontaneous abortion in early pregnancy.
- AI models demonstrated improved diagnostic accuracy compared to traditional methods.
- The study provides a publicly available resource for further research in histopathology.
- AI can assist pathologists by providing consistent assessments and reducing human error.
Takeaway
This study shows that using computers can help doctors look at tissue samples from miscarriages more accurately, which can help understand why they happen.
Methodology
The study created a dataset of 5666 annotated histopathology images and evaluated AI techniques against traditional methods for detecting tissue phenotypes.
Potential Biases
The accuracy of AI models relies heavily on the quality of annotations from pathologists, which could introduce biases.
Limitations
The study's findings may not be generalizable due to the dataset being sourced from a single institution and potential biases in expert annotations.
Participant Demographics
The dataset includes tissue samples from 120 patients at Atia Hospital, Karachi, Pakistan.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website