Software

ML/AI Engineer & PI. Building a Nigerian chest X-ray benchmark to measure how well Western AI models work on African patients.

[Software @software0408]

Building Africa's first chest X-ray AI benchmark from Nigerian hospital.

What happens when AI trained on American chest X-rays is deploy in a Nigerian hospital? No one has rigorously measured how well they work on our patients.

I'm 9 weeks into a research project to find out.

[Software @software0408]

@MrCharlesky Interested

[Software @software0408]

@MrCharlesky Interested

[Software @software0408]

@famous_dyl AI / Machine Learning Engineer

[Software @software0408]

@JobFound5 Interested

[Software @software0408]

https://t.co/XmXHHlmIbo

[Software @software0408]

The lesson: don't choose rules OR transformers.

Use both. Rules give you interpretable precision. BERT gives you semantic reach.

Our ensemble: rules primary on 11 labels, BERT primary on 2.

495 reports ready for radiologist validation.

#MedicalAI #NLP #ChestXray #Africa #BuildInPublic

[Software @software0408]

Week 10/48. This week I audited our entire NLP pipeline.

The question: should you use rules or transformers for medical text?

The answer surprised me. Thread.

[Software @software0408]

Also verified negation detection across 5,517 reports.

0 failures out of 134 negated mentions.

And reviewed 617 reports with 3+ labels — 89.9% of all label assignments supported by text evidence.

[Software @software0408]

@heyrimsha This is great

[MIT CSAIL @MIT_CSAIL]

A prompt you can use in Claude to help build & launch apps, v/@milesdeutscher.

[Software @software0408]

@ellen_in_sf literature review

[Software @software0408]

The pipeline: two extractors working in parallel.

Rule-based: 447-phrase bilingual dictionary with negation detection
BioClinicalBERT: fine-tuned transformer, Normal as 14th label

Rules handle 11 labels. BERT handles 2 that rules can't detect.

They vote. Flag disagreements.

[Software @software0408]

@lilyyang169 community

[Software @software0408]

The dataset: 5,517 chest X-rays from FMC Ebute-Metta, Lagos.

Free-text radiology reports. English mixed with local medical terms. 48% had no section headers.

Step 1: turn those reports into structured labels 13 pathologies + Normal.

That took Weeks 1–8.

[Software @software0408]

@beginnersblog1 ROADMAPS

[Software @software0408]

@cyrotrading Freebies gang, lolz 73 likes 38 comments under 6 hours. See turn ups when it comes to 1Sol