Competition projectChest X-ray · Explainability

Explainable Cardiomegaly Screening

A frontal chest X-ray system that combines anatomical segmentation, cardiothoracic measurements, CNN evidence, and tabular classification instead of relying on a single opaque score.

Project record

This page reports the maturity and evidence documented in Ethosoft's working archive. It does not imply peer review, regulatory clearance, or clinical validation.

Methods and evidence

Competition projectChest X-ray · Explainability

Explainable Cardiomegaly Screening

A frontal chest X-ray system that combines anatomical segmentation, cardiothoracic measurements, CNN evidence, and tabular classification instead of relying on a single opaque score.

Data: Project materials describe PadChest, MIMIC-CXR, CheXpert, VinDr-CXR, NIH ChestX-ray14, and CheXmask-derived masks.
Models: U-Net++ for the heart, DeepLabV3+ with ResNet101 for the lungs, and LightGBM over CTR, area features, and CNN probability.
Reported results: Heart IoU 87.26%; right/left lung IoU 91.11%/89.53%; classifier F1 91.4%; CTR SMAPE 7.5%.

Interpretability: the report includes cardiothoracic ratio, heart/lung area features, Grad-CAM, and SHAP. It also documents StyleGAN2-ADA synthetic-data experiments.

Expanded project brief

Anatomy-first chest X-ray reasoning.

The cardiomegaly system is organized as a chain rather than a single classifier. Heart and lung regions are segmented first, the cardiothoracic ratio and area relationships are calculated, and a classifier combines those measurements with image-model evidence. The local competition materials describe U-Net++ for the heart, DeepLabV3+ with a ResNet101 backbone for lungs, and LightGBM for the final tabular decision.

This structure is intended to make an output inspectable: a reviewer can see the predicted anatomy, the cardiothoracic measurements, Grad-CAM-style image evidence, and SHAP-style feature contributions. The archive also records experiments with StyleGAN2-ADA synthetic images and multiple public chest X-ray sources, which makes dataset provenance and leakage controls central to any reproduction.

Evidence boundary: reported IoU, F1, and ratio-error figures come from the project materials. They are competition-study results, not a clinical validation or a substitute for radiologist review.

Verbatim project material

Original source text

The following text is reproduced exactly from the supplied project files. It is preserved for provenance; it may be in Turkish, use Markdown notation, or contain draft statements.

teknofest2025syz/aciklama.txt

Download original file

Teknofest 2025 Sağlıkta Yapay Zeka yarışmasında geliştirdiğimiz tamamen özgün kardiyomegali tespit ve polip segmentasyonu projelerimizle lise alanında yarışma sunum puanı rekoru kırdık ve türkiye 3.sü olduk.

teknofest2025syz/kardiyomegali/aciklama.txt

Download original file

sistem sunum ve pdfte detaylıca açıklanmıştır.