Novel Logistic Regression Model of Chest CT Attenuation Coefficient Distributions for the Automated Detection of Abnormal (Emphysema or ILD) Versus Normal Lung

Abstract

Rationale and Objectives We evaluated the role of automated quantitative computed tomography (CT) scan interpretation algorithm in detecting Interstitial Lung Disease (ILD) and/or emphysema in a sample of elderly subjects with mild lung disease.ypothesized that the quantification and distributions of CT attenuation values on lung CT, over a subset of Hounsfield Units (HU) range [−1000 HU, 0 HU], can differentiate early or mild disease from normal lung. Materials and Methods We compared results of quantitative spiral rapid end-exhalation (functional residual capacity; FRC) and end-inhalation (total lung capacity; TLC) CT scan analyses in 52 subjects with radiographic evidence of mild fibrotic lung disease to 17 normal subjects. Several CT value distributions were explored, including (i) that from the peripheral lung taken at TLC (with peels at 15 or 65mm), (ii) the ratio of (i) to that from the core of lung, and (iii) the ratio of (ii) to its FRC counterpart. We developed a fused-lasso logistic regression model that can automatically identify sub-intervals of [−1000 HU, 0 HU] over which a CT value distribution provides optimal discrimination between abnormal and normal scans. Results The fused-lasso logistic regression model based on (ii) with 15 mm peel identified the relative frequency of CT values over [−1000, −900] and that over [−450,−200] HU as a means of discriminating abnormal versus normal, resulting in a zero out-sample false positive rate and 15%false negative rate of that was lowered to 12% by pooling information. Conclusions We demonstrated the potential usefulness of this novel quantitative imaging analysis method in discriminating ILD and/or emphysema from normal lungs.

A novel CT-responsive hydrogel for the construction of an organ simulation phantom for the repeatability and stability study of radiomic features

“…2, soft tissues in normal adults typically exhibit CT values in the range of 20-80 HU. [38][39][40][41] Bone tissues, on the other hand, have a higher density due to the significant presence of calcium, resulting in a CT value higher than 200 HU. 35,[38][39][40][41][42][43] Calcium ions, known for their excellent CT signals, are abundant in the human body, with approximately 60% of them being calcium-containing inorganic salts.…”

“…[38][39][40][41] Bone tissues, on the other hand, have a higher density due to the significant presence of calcium, resulting in a CT value higher than 200 HU. 35,[38][39][40][41][42][43] Calcium ions, known for their excellent CT signals, are abundant in the human body, with approximately 60% of them being calcium-containing inorganic salts. Among these, bone salts primarily consist of calcium and phosphorus chemicals, distributed in the organic matter as crystalline hydroxyapatite and amorphous calcium phosphate.…”

Editorial: Sarcoidosis: now and into the future

“…Imaging patterns on CT scans have hitherto played a key role in the diagnosis and prognostication of ILDs however there is significant interobserver variability and deep learning techniques have been developed to minimize this variability and improve diagnostic certainty. Furthermore, recent studies suggest that automated quantitative interpretative CT algorithms can detect areas of abnormal lung tissue, including areas of fibrotic lung disease [13] and quantification of fibrosis has been harnessed to build predictive models for patients with fibrotic pulmonary sarcoidosis [14]. With the advent of effective therapy for patients with PPF [10,15] (which also includes patients with progressive fibrotic pulmonary sarcoidosis), there is a need for a rapid and easily accessible CT quantification systems to aid with the quantification of progressive fibrosis for which treatment may be beneficial.…”

High-dimensional inference of ordinal data with medical applications

“…This study is similar as Chan et al (2016), but with ordinal responses. The approach introduced above leverages on the fact that scarring in the periphery of Table 6.4: Airway score: out-sample classification frequencies using V 3 (method 2) @ @ @ @ @ @ yŷ 0 1 2 3 4 0 12 5 0 0 0 1 2 9 1 7 1 2 0 3 1 2 0 3 0 7 2 6 2 4 0 3 1 5 0 Table 6.5: Parenchymal score: out-sample classification frequencies using V 2 (method 1) @ @ @ @ @ @ Table 6.6: Parenchymal score: out-sample classification frequencies using V 2 (method 2) @ @ @ @ @ @ yŷ 0 1 2 3 4 0 12 4 1 0 0 1 1 3 7 2 0 2 2 2 8 1 3 3 0 1 3 3 3 4 0 1 4 3 5 disease, which is better than Chan et al (2016) analysis for not using the information pooling strategy. Due to the superior performance of the cumulative logit model in the simulation study of high-dimensional setting, a future modification of the current approach is to use the logit link to select variables.…”