PDAC prediction based on plain CT images

A Novel Multiresolution-Statistical Texture Analysis Architecture: Radiomics Aided Diagnosis of PDAC Based on Plain CT Images

Mithunjha Anandakumar
2 min readMar 6, 2022


Pancreatic Ductal Adenocarcinoma (PDAC ) is a malignant tumor that is usually diagnosed in later stages and has a very poor prognosis. CT imaging is a non-invasive PDAC diagnostic technique. It is hard to diagnose the condition with plain CT scans because the human eye cannot detect many grayscale variations. Although contrast-enhanced CT can be used to visually diagnose the condition, it also involves several limitations: Allergic reactions to contrast agents, interpretations are subjective, and so on.

The paper titled “A Novel Multiresolution-Statistical Texture Analysis Architecture: Radiomics Aided Diagnosis of PDAC Based on Plain CT Images” proposes a multiresolution statistical texture analysis architecture to discriminate PDAC slices from healthy pancreas (HP) slices using Radiomics and AI.

Radiomics can be defined as a process that extracts quantitative features from medical images, especially radiology images. Radiomics study is based on the fact that medical images reflect the information about the underlying pathological/ physiological condition. Radiomics together with AI plays an important role in the early diagnosis of PDAC based on plain CT.

Radiomics Scheme used in the paper [1]

Multiresolution-Statistical Texture Analysis (MSTA) architecture captures and utilizes the grayscale variations as much as possible to represent the ROI. MSTA architecture includes several steps: multiresolution analysis, analysis of coefficients, discretization of coefficients, and Statistical analysis methods to extract mathematical descriptors that can be used as texture features. Extracted features then go through feature selection and finally classification of PDAC. In addition, significance testing was conducted to determine features that are statistically significant difference between both PDAC and HP.

I’ve implemented the MSTA architecture on python, to classify PDAC and HP from ROIs. Code available here :


[1] Qiu, J.J., Yin, J., Qian, W., Liu, J.H., Huang, Z.X., Yu, H.P., Ji, L. and Zeng, X.X., 2020. “A Novel Multiresolution-Statistical Texture Analysis Architecture: Radiomics-Aided Diagnosis of PDAC Based on Plain CT Images”. IEEE Transactions on Medical Imaging, 40(1), pp.12–25.