Neonatal TNE reporting tool

🙏 thank you

I’m trying to do the same thing right now and will look into this!

Hi @Mo7 , such a cool tool. I think @Stefan Johansson could feature it in the newsletter later this month?

On my side, I'm really missing some info on the tool developer, but I don't know if that's a standard expectation. How does it work? I understand that it collects the input data, and there's just an LLM that constructs the summary?

Besides, I played with it a bit, and while I run it with only limited values and left many fields incomplete, the summary was very assertive of normal findings- eg. it says " The indexed outputs are adequate, and there are no indications of pulmonary hypertension or impaired filling pressures, though continuous monitoring will be essential given the patient's prematurity and hemodynamic vulnerability." - I'm not an echo expert here, but I'm wondering if the parameters I entered really support these statements sufficiently. So that would be my comment - that it uses reassuring language when there are gaps due to lack of input. Maybe you could consider flagging values that are missing in the interpretation, or pushing it to give statements like "interpretation on presence of absence of XXX cannot be made due to missing values"?

Thanks so much, really appreciate you taking the time to explore it and for the thoughtful feedback.

Just to clarify, I’m the one who built the tool as part of my own clinical workflow in neonatal haemodynamics. It’s not intended as a standalone AI system, the calculations are rule-based, and the summary is generated from those inputs.

You’ve raised a very important point about the summary being overly reassuring when data is incomplete, I completely agree. At the moment, missing inputs are not sufficiently accounted for, which can lead to overconfident statements. I’m working on improving this so that:

• missing parameters are explicitly flagged

• the interpretation reflects uncertainty

• and statements such as “cannot assess due to missing values” are included where appropriate

The aim is to better reflect how we actually interpret TNE in practice, where a multi-parameter approach is essential and absence of data should never imply normality.

I’ll also add a short section explaining who developed the tool and how it works, as I agree it currently feels a bit like a black box.

Thanks again, this is exactly the kind of feedback I was hoping for.

And yes, very happy for it to be featured or shared more widely.

I shall try this

Good