CT negative attenuation pixel distribution and texture analysis for detection of fat in small angiomyolipoma on unenhanced CT

Naoki Takahashi, Mitsuru Takeuchi, Kohei Sasaguri, Shuai Leng, Adam Froemming, Akira Kawashima

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

Purpose: The purpose of the paper is to evaluate if CT pixel distribution and texture analysis can identify fat in angiomyolipoma (AML) on unenhanced CT. Methods: Thirty-seven patients with 38 AMLs and 75 patients with 83 renal cell carcinomas (RCCs) were evaluated. Region of interest (ROI) was manually placed over renal mass on unenhanced CT. In-house software generated multiple overlapping small-ROIs of various sizes within whole-lesion-ROI. Maximal number of pixels under cutoff attenuation values in the multiple small-ROIs was calculated. Skewness of CT attenuation histogram was calculated from whole-lesion-ROI. Presence of fat in renal mass was also evaluated subjectively. Performance of subjective evaluation and objective methods for identifying fat was compared using McNemar test. Results: Macroscopic fat was identified in 15/38 AMLs and 1/83 RCCs by both subjective evaluation and by CT negative pixel distribution analysis (p = 1.0). Optimal threshold was ≥6 pixels below −30 HU within 13-pixel-ROI. Skewness of < −0.4 in whole-lesion-ROI identified fat in 10/38 AMLs and 0/83 RCCs. By combining CT negative pixel distribution analysis and skewness, fat was identified in 20/38 AMLs and 1/83 RCCs, but the difference to the subjective method was not statistically significant (p = 0.07). Conclusion: CT negative attenuation pixel distribution analysis does not identify fat in AML beyond subjective evaluation. Addition of skewness by texture analysis may help improve identifying fat in AML.

Original languageEnglish (US)
Pages (from-to)1142-1151
Number of pages10
JournalAbdominal Radiology
Volume41
Issue number6
DOIs
StatePublished - Jun 1 2016

Fingerprint

Angiomyolipoma
Fats
Renal Cell Carcinoma
Kidney
Software

Keywords

  • Angiomyolipoma
  • Computed tomography
  • Kidney
  • Renal cell carcinoma

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging
  • Gastroenterology
  • Urology
  • Radiological and Ultrasound Technology

Cite this

CT negative attenuation pixel distribution and texture analysis for detection of fat in small angiomyolipoma on unenhanced CT. / Takahashi, Naoki; Takeuchi, Mitsuru; Sasaguri, Kohei; Leng, Shuai; Froemming, Adam; Kawashima, Akira.

In: Abdominal Radiology, Vol. 41, No. 6, 01.06.2016, p. 1142-1151.

Research output: Contribution to journalArticle

Takahashi, Naoki ; Takeuchi, Mitsuru ; Sasaguri, Kohei ; Leng, Shuai ; Froemming, Adam ; Kawashima, Akira. / CT negative attenuation pixel distribution and texture analysis for detection of fat in small angiomyolipoma on unenhanced CT. In: Abdominal Radiology. 2016 ; Vol. 41, No. 6. pp. 1142-1151.
@article{266da7b00d5b4b73b56f26a00a5a53ee,
title = "CT negative attenuation pixel distribution and texture analysis for detection of fat in small angiomyolipoma on unenhanced CT",
abstract = "Purpose: The purpose of the paper is to evaluate if CT pixel distribution and texture analysis can identify fat in angiomyolipoma (AML) on unenhanced CT. Methods: Thirty-seven patients with 38 AMLs and 75 patients with 83 renal cell carcinomas (RCCs) were evaluated. Region of interest (ROI) was manually placed over renal mass on unenhanced CT. In-house software generated multiple overlapping small-ROIs of various sizes within whole-lesion-ROI. Maximal number of pixels under cutoff attenuation values in the multiple small-ROIs was calculated. Skewness of CT attenuation histogram was calculated from whole-lesion-ROI. Presence of fat in renal mass was also evaluated subjectively. Performance of subjective evaluation and objective methods for identifying fat was compared using McNemar test. Results: Macroscopic fat was identified in 15/38 AMLs and 1/83 RCCs by both subjective evaluation and by CT negative pixel distribution analysis (p = 1.0). Optimal threshold was ≥6 pixels below −30 HU within 13-pixel-ROI. Skewness of < −0.4 in whole-lesion-ROI identified fat in 10/38 AMLs and 0/83 RCCs. By combining CT negative pixel distribution analysis and skewness, fat was identified in 20/38 AMLs and 1/83 RCCs, but the difference to the subjective method was not statistically significant (p = 0.07). Conclusion: CT negative attenuation pixel distribution analysis does not identify fat in AML beyond subjective evaluation. Addition of skewness by texture analysis may help improve identifying fat in AML.",
keywords = "Angiomyolipoma, Computed tomography, Kidney, Renal cell carcinoma",
author = "Naoki Takahashi and Mitsuru Takeuchi and Kohei Sasaguri and Shuai Leng and Adam Froemming and Akira Kawashima",
year = "2016",
month = "6",
day = "1",
doi = "10.1007/s00261-016-0714-y",
language = "English (US)",
volume = "41",
pages = "1142--1151",
journal = "Abdominal Radiology",
issn = "2366-004X",
publisher = "Springer New York",
number = "6",

}

TY - JOUR

T1 - CT negative attenuation pixel distribution and texture analysis for detection of fat in small angiomyolipoma on unenhanced CT

AU - Takahashi, Naoki

AU - Takeuchi, Mitsuru

AU - Sasaguri, Kohei

AU - Leng, Shuai

AU - Froemming, Adam

AU - Kawashima, Akira

PY - 2016/6/1

Y1 - 2016/6/1

N2 - Purpose: The purpose of the paper is to evaluate if CT pixel distribution and texture analysis can identify fat in angiomyolipoma (AML) on unenhanced CT. Methods: Thirty-seven patients with 38 AMLs and 75 patients with 83 renal cell carcinomas (RCCs) were evaluated. Region of interest (ROI) was manually placed over renal mass on unenhanced CT. In-house software generated multiple overlapping small-ROIs of various sizes within whole-lesion-ROI. Maximal number of pixels under cutoff attenuation values in the multiple small-ROIs was calculated. Skewness of CT attenuation histogram was calculated from whole-lesion-ROI. Presence of fat in renal mass was also evaluated subjectively. Performance of subjective evaluation and objective methods for identifying fat was compared using McNemar test. Results: Macroscopic fat was identified in 15/38 AMLs and 1/83 RCCs by both subjective evaluation and by CT negative pixel distribution analysis (p = 1.0). Optimal threshold was ≥6 pixels below −30 HU within 13-pixel-ROI. Skewness of < −0.4 in whole-lesion-ROI identified fat in 10/38 AMLs and 0/83 RCCs. By combining CT negative pixel distribution analysis and skewness, fat was identified in 20/38 AMLs and 1/83 RCCs, but the difference to the subjective method was not statistically significant (p = 0.07). Conclusion: CT negative attenuation pixel distribution analysis does not identify fat in AML beyond subjective evaluation. Addition of skewness by texture analysis may help improve identifying fat in AML.

AB - Purpose: The purpose of the paper is to evaluate if CT pixel distribution and texture analysis can identify fat in angiomyolipoma (AML) on unenhanced CT. Methods: Thirty-seven patients with 38 AMLs and 75 patients with 83 renal cell carcinomas (RCCs) were evaluated. Region of interest (ROI) was manually placed over renal mass on unenhanced CT. In-house software generated multiple overlapping small-ROIs of various sizes within whole-lesion-ROI. Maximal number of pixels under cutoff attenuation values in the multiple small-ROIs was calculated. Skewness of CT attenuation histogram was calculated from whole-lesion-ROI. Presence of fat in renal mass was also evaluated subjectively. Performance of subjective evaluation and objective methods for identifying fat was compared using McNemar test. Results: Macroscopic fat was identified in 15/38 AMLs and 1/83 RCCs by both subjective evaluation and by CT negative pixel distribution analysis (p = 1.0). Optimal threshold was ≥6 pixels below −30 HU within 13-pixel-ROI. Skewness of < −0.4 in whole-lesion-ROI identified fat in 10/38 AMLs and 0/83 RCCs. By combining CT negative pixel distribution analysis and skewness, fat was identified in 20/38 AMLs and 1/83 RCCs, but the difference to the subjective method was not statistically significant (p = 0.07). Conclusion: CT negative attenuation pixel distribution analysis does not identify fat in AML beyond subjective evaluation. Addition of skewness by texture analysis may help improve identifying fat in AML.

KW - Angiomyolipoma

KW - Computed tomography

KW - Kidney

KW - Renal cell carcinoma

UR - http://www.scopus.com/inward/record.url?scp=84975259888&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84975259888&partnerID=8YFLogxK

U2 - 10.1007/s00261-016-0714-y

DO - 10.1007/s00261-016-0714-y

M3 - Article

C2 - 27015866

AN - SCOPUS:84975259888

VL - 41

SP - 1142

EP - 1151

JO - Abdominal Radiology

JF - Abdominal Radiology

SN - 2366-004X

IS - 6

ER -