Pink and Blue Glass Art Work Boston Deaconess Hospital

• Journal List
• Springer Open up Selection
• PMC5717124

Eur Radiol. 2018; 28(1): 235–242.

Software-based risk stratification of pulmonary adenocarcinomas manifesting as pure ground glass nodules on computed tomography

Ursula Nemec

ⁱDepartment of Biomedical Imaging and Image-guided Therapy, Vienna General Hospital, Medical University of Vienna, Waehringer Guertel 18-20, Vienna, Austria

Benedikt H. Heidinger

²Radiology, Beth State of israel Deaconess Medical Heart, Harvard Medical Schoolhouse, Boston, MA USA

Kevin R. Anderson

ⁱⁱⁱPathology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA USA

Michael S. Westmore

⁴Imbio, Delafield, WI Us

Paul A. VanderLaan

³Pathology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA Us

Alexander A. Bankier

²Radiology, Beth Israel Deaconess Medical Centre, Harvard Medical School, Boston, MA USA

Received 2017 Feb 17; Revised 2017 May viii; Accepted 2017 Jun 8.

Supplementary Materials

ESM 1: (GIF fourteen kb)

GUID: 60FD934A-7434-43EA-99CD-327F9D5E7D6A

Loftier resolution prototype (TIF 1463 kb)

GUID: 0EDA25F4-03D8-4F26-B338-7A7A928DE204

Abstruse

Objectives

To assess the performance of the "Computer-Aided Nodule Assessment and Risk Yield" (CANARY) software in the differentiation and run a risk assessment of histological subtypes of lung adenocarcinomas manifesting equally pure ground drinking glass nodules on computed tomography (CT).

Methods

64 surgically resected and histologically proven adenocarcinomas manifesting as pure footing-drinking glass nodules on CT were assessed using CANARY software, which classifies voxel-densities into 3 risk components (depression, intermediate, and high risk). Differences in risk components between histological adenocarcinoma subtypes were analysed. To determine the optimal threshold reflecting the presence of an invasive focus, sensitivity, specificity, negative predictive value, and positive predictive value were calculated.

Results

28/64 (44%) were adenocarcinomas in situ (AIS); 26/64 (41%) were minimally invasive adenocarcinomas (MIA); and 10/64 (xvi%) were invasive ACs (IAC). The software showed significant differences in risk components betwixt histological subtypes (P<0.001–0.003). A relative book of 45% or less of depression-risk components was associated with histological invasiveness (specificity 100%, positive predictive value 100%).

Conclusions

CANARY-based hazard assessment of ACs manifesting as pure footing glass nodules on CT allows the differentiation of their histological subtypes. A threshold of 45% of low-risk components reflects invasiveness in these groups.

Key points

• CANARY-based risk assessment allows the differentiation of their histological subtypes.

• 45% or less of low-take a chance component reflects histological invasiveness.

• CANARY has potential part in suspected adenocarcinomas manifesting as pure ground-drinking glass nodules.

Electronic supplementary material

The online version of this article (doi:10.1007/s00330-017-4937-2) contains supplementary cloth, which is available to authorized users.

Keywords: Lung adenocarcinoma, Pure basis glass nodule, Risk stratification, Computed tomography, Software based

Introduction

The "Computer Aided Nodule Assessment and Risk Yield" (CANARY) software packet has recently been introduced for risk stratification of pulmonary nodules of the lung adenocarcinoma spectrum [1]. CANARY is based on assessing voxel densities of pulmonary nodules and the subsequent assignment of hazard, as inferred from histology, according to voxel proportion and clustering. While this software has been successful in assessing both histology and outcome in a representative but morphologically diverse number of lung adenocarcinomas (AC) [1–3], it has non notwithstanding been tested in lung adenocarcinomas with morphologically more uniform appearance on computed tomography (CT).

A singled-out subgroup of lung adenocarcinomas with a more compatible CT appearance, are those manifesting as pure ground glass nodules. They are considered to reverberate a specific subgroup of adenocarcinomas with tailored management recommendations according to the 'Guidelines for Management of Incidental Pulmonary Nodules Detected on CT Images: From the Fleischner Society 2017 [4]. These adenocarcinomas manifesting as pure ground glass nodules are important because they are mutual and can represent different histologic subtypes, ranging from noninvasive adenocarcinoma in situ (AIS) to invasive subtypes such every bit minimally invasive ACs (MIAs) and invasive ACs (IACs) [v]. A tool to differentiate amidst these histological subtypes in lung ACs could help to reduce diagnostic uncertainty in the management of such nodules. Therefore, the specific purpose of our study was to investigate the operation of CANARY software in the differentiation and hazard assessment of histological subtypes of lung adenocarcinomas manifesting as pure ground glass nodules on CT.

Materials and methods

The protocol for this retrospective study (#15-020) was approved past our institutional review board with the need for written informed consent waived.

Study population

This retrospective study was based on a radiology-pathology information repository, which has served for other studies in the past, with a partial overlap of lesions and patients [6–8]. We reviewed the medical records of patients undergoing surgical resection for master pulmonary adenocarcinoma (Ac) at our institution betwixt January 2005 and July 2016. Inclusion criteria were: (1) histologically verified primary pulmonary AC; (ii) bachelor pre-operative computed tomography (CT) examination; and (three) AC manifesting as pure ground glass nodule (GGN) on CT. Exclusion criteria were (one) histologic diagnosis other than Air-conditioning; (2) pre-operative CT examination not available; (3) Air-conditioning manifesting as a solid or function-solid nodule on CT; and (four) pathologically described nodule non identifiable on CT.

The medical record review resulted in 698 resected ACs, of which 89 (12.4%) were excluded because pre-operative CT examinations were non available. After further radiologic consensus review by ii thoracic radiologists (AAB and BHH, with 20 and 2 years of experience, respectively), 546/611 (89.four%) adenocarcinomas were excluded considering either the nodules were non identifiable on CT or they manifested as solid or part-solid lesions. Thereafter, the surgical resected specimens of the 65 remaining ACs were re-reviewed past a subspecialty-trained thoracic pathologist (PAV) and a senior pathology resident (KRA). This histopathological consensus review excluded one/65 (1.5%) nodule because it was an atypical adenomatous hyperplasia (AAH). This left 64 adenocarcinomas manifesting equally pure GGNs for CANARY analysis.

CT acquisitions

CT examinations were performed using various CT scanner units and acquisition protocols, which were all considered state-of-the-art at the time of acquisition. The most frequently used CT units were Aquilion Ane (Toshiba, 320-detector row), Discovery CT750 HD (GE Medical Systems, 64-detector row), and Lightspeed VCT (GE Medical Systems, 64-detector row).

All CT examinations were performed in the supine body position, at total inspiration, covering the entire lung. Examinations before April 2007 were performed with fixed mAs (range: 130-340 mAs) and 120 kVp. After April 2007, automatic exposure control and other dose reduction algorithms were used. Transverse images were reconstructed with 0.625 to 1.v mm section thickness using standard reconstruction kernels in lung window settings (mean, −500 HU; width, 1500 HU). CT examinations for staging purposes (north=21, 36%) were performed using intravenous contrast textile. In the remaining 38 (64%) examinations, no contrast fabric was administered.

CT

The CT examinations were anonimised and presented in a random social club on a picture archiving and communication organisation (GE Healthcare, Centricity) to two thoracic radiologists independently. Both radiologists were unaware of clinical and histological information. The long-centrality and short-axis diameters of all 64 ACs were measured. Measurements were performed in lung window setting on the transverse CT department that displayed the largest nodule dimensions. Measurements were recorded in millimetres. Average CT diameters were calculated based on the long- and curt-axis diameter.

Reckoner Aided Nodule Cess and Risk Yield (CANARY)

All included nodules were likewise analysed by CANARY, a software for automated take a chance cess of ACs based on the reduction of voxel density histograms to 9 natural clusters. In a pilot report, these clusters were identified past selecting 774 regions of interest (two dimensional area ROIs, 9 x 9 voxels) in 37 ACs and comparing all ROIs to one some other using Affinity Propagation and pairwise similarity metrics [1]. Exemplars were generated from these nine natural clusters and color-coded equally violet (Five), indigo (I), blue (B), green (G), yellow (Y), orangish (O), red (R), cyan (C) and pink (P). These exemplars were and then used for allocation of nodule run a risk components. To allocate nodule risk components, the 9 x 9 region around each voxel inside the segmented nodule is compared to these nine exemplars to decide which cluster the voxel is most similar to. This results in a specific combination of color codes for each nodule. A representative nodule is shown in Fig.1.

(a) Transverse computed tomography prototype of an adenocarcinoma in situ manifesting as pure ground drinking glass nodule in the left upper lobe (b) Colour-coded CANARY output overlay shows components of low risk group (blue-green–cyan) (depression risk 99.8%, intermediate adventure 0.2%, high risk 0%) (c) histologic image showing purely lepidic tumour growth blueprint without areas of invasion (H&Eastward stain, 200x original magnification)

For lung nodule characterisation, all 64 GGNs were located by i observer (United nations, iv years of experience in general radiology) using OsiriX (v8.0.1, Pixmeo SARL, Bernex, Switzerland). Initial data processing steps involved the segmentation of the nodule with the brush tool and pre-determined pixel values for inside and outside the ROIs. Predetermined pixel-values were arbitrarily defined numbers required by the CANARY algorithm to decide what is within or exterior of the ROI. ROIs were and so saved as DICOMs and uploaded for CANARY analysis. Time needed for CANARY analysis averaged 10 minutes per lesion.

For each nodule, the CANARY analysis yielded the accented and relative volume of each colour. These nine colours were then combined into three components as determined past Maldonado et al. [1]: (one) depression chance – bluish-green-cyan; (2) intermediate risk – pink-yellow; and (3) high risk – violet-indigo-red-orangish.

The accented and relative volumes of the three components of each nodule were so used for analysis.

Histologic Review

The original histologic glass slides from all 64 resected nodules were retrieved and re-reviewed past a subspecialty-trained thoracic pathologist (PAV, v years of experience) and a senior pathology chief resident (KRA, PGY4). The histologic growth patterns (including the non-invasive/inferred in-situ lepidic pattern too every bit whatsoever invasive patterns including acinar, papillary, micropapillary, and solid growth) were recorded in semiquantitative 5% increments according to current practice recommendations [ix]. The size and number of any invasive foci (past definition, any histologic pattern other than lepidic) were assessed. Based on this information, these adenocarcinomas were classified using electric current WHO terminology and diagnostic criteria as AIS, MIA, or IAC [9]. AIS is defined as a lonely lung adenocarcinoma ≤3.0 cm demonstrating a pure lepidic growth blueprint without bear witness of stromal, vascular, or pleural invasion, tumour necrosis, or spread of tumour through the airspaces. MIA is defined as a solitary lung adenocarcinoma ≤three.0 cm with a predominantly lepidic growth pattern only that contains an invasive focus (i.e., growth design other than lepidic) that is ≤0.5 cm in greatest dimension, and also lacks lymphatic, vascular, or pleural invasion, tumour necrosis, or spread of tumour through the airspaces. In contrast, IAC are lung adenocarcinomas that either are greater than 3.0 cm, have an invasive growth pattern greater than 0.v cm, or that demonstrate any of the more aggressive features lacking in AIS or MIA (lymphatic, vascular, or pleural invasion, tumour necrosis, or spread of neoplasm through the airspaces). Any discrepancies in neoplasm component measurement or nomenclature between PAV and KRA were resolved via multihead microscope consensus review.

Statistical analysis

All statistical analyses and graphs were performed using commercially available software (STATA 12.0, StatCorp, College Station, TX, USA). The normality of distributions was assessed by the Shapiro-Wilk test. Normally distributed information were expressed as mean ± SD. Not-normally distributed information were expressed as median together with their [interquartile range]. Statistical Power Analysis was performed using STATA module for simulation-based power analysis. A p value less than 0.05 was considered statistically significant.

Commencement, nosotros calculated the distribution of AIS, MIA, and IACs of the 64 GGNs included in this study.

Second, we calculated the medians of the low adventure, intermediate chance, and loftier risk CANARY components for AIS, MIA, and IACs, respectively. Potential differences between the CANARY components in the histological subtypes, namely AIS, MIAs and IACs, were assessed for statistical significance using a Kruskal-Wallis test, with private differences additionally assessed with Conover-Iman tests.

Third, nosotros tested the strength of the relation of the size of invasive foci with both volume and percentage of the three CANARY components past computing Spearman-rank correlation coefficients. All P-values obtained from Spearman correlation were Bonferroni corrected.

Fourth, to assess the presence of invasive foci, we grouped the ii histological components with an invasive component, namely MIAs and IACs, and plotted individual percentages of the low chance and intermediate risk grouping, respectively.

Based on those private percentages, we defined threshold values for low risk group in intervals of five percentages starting at twoscore%. To make up one's mind the optimal threshold reflecting the presence of an invasive focus, we calculated sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV) for each threshold value.

To assess the influence of contrast material administration on the CANARY analysis, we performed a linear regression assay with CANARY components as dependent variables and dissimilarity administration and section thickness as independent variables.

Results

The 64 ACs manifesting as pure GGNs on CT were detected in 59 patients, 5/59 (8%) of whom had more than 1 resected GGN, just no patient had more than than two resected nodules.

In that location were 40/59 (67.8%) women (hateful age 67±9 years; range, 45-84 years), and nineteen/59 (32.ii%) men (mean age 69±9 years; range, 49-86 years). There was no significant departure in historic period between women and men (P=0.439). The median duration between the CT examination and surgical resection was 1 month [IQR 0-2].

GGNs were located equally follows: right upper lobe (22; 34%); left upper lobe (18; 28%); right lower lobe (13; 20%), left lower lobe (9; 14%) and the correct centre lobe (2; 3%).

Twenty-five (39%) of the nodules were removed past lobectomy, 28 (44%) by wedge resection, and 11 (17%) past segmentectomy.

Using an alpha level of 0.05, retrospective power analysis showed a statistical power of 0.95 for low risk components, 0.94 for intermediate hazard component and 0.55 for loftier take chances component.

Of the 64 GGNs included in our report, 28 (44%) were AIS, 26 (41%) were MIAs, and 10 (sixteen%) were IACs. The average CT diameter of the nodules was 14.iv± 5.iii mm.

Examples of CANARY assessment for AIS, MIA, and IAC are displayed in Figs.two, ​ three and ​ 4. The iii CANARY components for AIS, MIA, and IACs are displayed in Table ​ one. The volumes of the depression take a chance CANARY component did not differ significantly between AIS, MIAs, and IACs (P=0.408). The volumes of the intermediate take a chance group were significantly lower in AIS than in MIAs (P=0.003), and lower in AIS than in IACs (P<0.001). Nonetheless, the book departure between MIAs and IACs was non statistically meaning (P=0.055). The volumes of the high risk group were significantly lower in AIS than in MIAs (P=0.040), in AIS than in IACs (P<0.001), and in MIAs than IACs (P=0.001).

(a) Color-coded CANARY output overlay of adenocarcinoma in situ (blue-light-green-cyan-pink) (low adventure 94.9%, intermediate take chances 5.i%, high risk 0%);. (b) histologic image showing patchy interstitial chronic inflammation, but a purely lepidic tumour grown pattern without areas of invasion (H&E stain, 100x original magnification)

(a) Color-coded CANARY output overlay of minimally invasive adenocarcinoma (blue-light-green-cyan-pink-xanthous) (low run a risk 55.7%, intermediate hazard 44.3%, loftier risk 0%) (b) histologic image showing the transition zone (dotted line) of peripheral not-invasive lepidic growth blueprint (upper right) and the more central invasive acinar pattern component (lower left)(H&East stain, 200x original magnification)

(a) Color-coded CANARY output overlay of invasive adenocarcinoma (yellow-pink-cyan) (low risk 32.0%, intermediate take a chance 66.9%, high risk ane%)(b) histologic image from the central area of the tumour shows the invasive acinar growth design (c) histologic epitome from the neoplasm periphery showing non-invasive lepidic growth pattern and adjacent uninvolved lung parenchyma (H&Due east stain, 200x original magnification)

Table 1

Book and percentages of CANARY components for adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA) and invasive adenocarcinoma (IAC)

	Low risk		Intermediate risk		High risk
	Volume	Percentage	Volume	Percentage	Volume	Pct
	(mL)	(%)	(mL)	(%)	(mL)	(%)
AIS, n=28	i.1±1.2	78.five±17.6	0.3±0.5	21.5±17.vi	0.0±0.0	0.0±0.0
	(0.2,7.ane)	(45.two,100.0)	(0.0,2.7)	(0.0,54.8)	(0.0,0.0)	(0.0,0.one)
MIA, n=26	i.five±one.7	65.iii±27.0	0.9±ane.6	32.3±23.5	0.1±0.three	ii.4±6.6
	(0.ii,seven.1)	(15.5,100)	(0.0,8.2)	(0.0,71.0)	(0.0,1.6)	(0.0,29.6)
IAC, north=10	2.3±2.eight	46.6±22.9	iii.9±six.nine	49.9±xix.8	0.7±ane.nine	3.four±5.5
	(0.3,seven.4)	(twenty.1,xc.0)	(0.1,23.iii)	(10.0,66.9)	(0.0,6.3)	(0.0,17.ane)

Information are displayed as hateful ± standard difference, together with (range)

The percentages of the low take a chance CANARY component were significantly college in AIS than MIAs (P=0.034), in AIS than IACs (P<0.001), and in MIAs than IACs (P=0.024). The percentages of the intermediate adventure component were significantly lower in AIS than MIAs (P=0.037), in AIS than IACs (P<0.001), and in MIAs than IACs (P=0.016). The percentages of the high take chances group were significantly lower in AIS than in MIAs (P=0.037), in AIS than in IACs (P<0.001), and in MIAs than IACs (P=0.002).

The relations of the size of the invasive focus and the volumes of the intermediate and high hazard components were statistically significant (r=0.477, P<0.001, and r=0.471, P<0.001, respectively). Even so, the relation of the size of the invasive focus and the book of the low risk component was not statistically meaning (r=0.200, P=0.679). The relations between the percentage of the three CANARY components with the size of invasive focus were statistically significant (low risk: r=-0.406, P=0.005; intermediate adventure: r=0.407, P=0.005; loftier risk: r=0.467, P<0.001), respectively.

Individual percentages of the low risk group are displayed in Fig.5. For each threshold value, sensitivity, specificity, NPV, and PPV are shown in Tabular array ​ 2.

Private percentages of the CANARY low risk components for not-invasive (adenocarcinoma in situ) and invasive (minimally invasive adenocarcinoma and invasive adenocarcinoma) groups Note.- AIS- adenocarcinoma in situ; MIA - minimally invasive adenocarcinoma: IAC- invasive adenocarcinoma; Sens – Sensitivity; Spec- Specificity; PPV-positive predictive value; NPV – negative predictive value

Tabular array ii

Sensitivity, specificity, positive predictive value and negative predictive value, together with their (95% conviction intervals) for predefined thresholds of CANARY component percentages

Cutoff	Sensitivity	Specificity	Positive predictive value	Negative predictive value
≤ forty%	xxx.half dozen% (16.three% - 48.1%)	100.0% (87.7% - 100.0%)	100.0% (71.5% - 100.0%)	52.8% (38.half dozen% - 66.7%)
≤ 45%	33.3% (18.half dozen% - 51.0%)	100.0% (87.7% - 100.0%)	100.0% (73.5% - 100.0%)	53.eight% (39.v% - 67.viii%)
≤ 50%	36.1% (xx.8% - 53.eight%)	96.iv% (81.vii% - 99.nine%)	92.ix% (66.1% - 99.eight%)	54.0% (39.3% - 68.ii%)
≤ 55%	44.iv% (27.9% - 61.9%)	85.vii% (67.3% - 96.0%)	80.0% (56.3% – 94.3%)	54.v% (38.8% - 69.6%)
≤ 60%	58.3% (forty.eight% - 74.five%)	82.1% (63.one% - 93.9%)	eighty.8% (60.half dozen% - 93.4%)	60.5% (43.iv% - 76.0%)
≤ 65%	58.iii% (twoscore.viii% - 74.5%)	71.4% (51.3% - 86.8%)	72.iv% (52.8% - 87.3%)	57.1% (39.iv% - 73.seven%)
≤ lxx%	63.ix% (46.two% - 79.2%)	67.9% (47.6% - 84.one%)	71.9% (53.3% - 86.iii%)	59.4% (xl.6% - 76.iii%)

Individual percentages of intermediate take a chance group are shown in Supplementary Fig. ane.

Linear regression assay showed no relation betwixt CANARY component percentages and contrast material administration (P=0.331 to 0.664) as well as department thickness (P=0.255 to 0.762).

Word

Our study showed that in ACs manifesting as pure ground glass nodules on CT, CANARY software immune the differentiation of the histological subtypes AIS, MIA and IAC. The analysis software indeed showed statistically significant differences in take a chance components betwixt these three subtypes. Previously published studies also showed differences in risk components between histological Air-conditioning subtypes [i–3]. Still, these studies included morphologically heterogeneous groups of ACs, whereas our study showed this in a very homogeneous group of ACs. Therefore, to the best of our knowledge, our findings are novel in showing these differences in a homogenous group by exclusively including only ACs manifesting equally pure basis glass nodules. This is important because risk assessment of ACs manifesting as pure ground drinking glass nodules is essential for choosing a direction approach such equally CT follow-upwardly interval, time of biopsy, and choice of an advisable surgical method [x–15].

In our study, the percentage of low risk components decreased from AIS to MIA and from MIA to IAC. This likely mirrors the consistency of CANARY assessment with histological characteristics of AC subtypes. Simultaneously, the percentage of intermediate take a chance and loftier risk components increased from AIS to MIA and from MIA to IAC. Again, this may mirror the higher number of invasive foci in these histological subtypes [16, 17]. Annotation that in our study, AIS did not evidence high risk components which should be expected equally AIS is by definition a pre-cancerous neoplasm without invasive foci [16, 17]. Furthermore, the percentage of high run a risk components was, in general, likewise depression in MIA and IAC, which may be due to our morphologically homogeneous group of ACs and its like attenuation profiles [18]. Our study showed that analysis software was able to notice differences fifty-fifty in this very homogeneous group.

Our written report also showed that a low risk component threshold of 45% provided a 100% specificity and positive predictive value for invasiveness of ACs manifesting as pure ground drinking glass nodules. At a threshold of l%, the PPV was still 93% with a specificity of 96% and only a minimal gain in sensitivity. This reflects that below a threshold of 50%, in that location is a shut to 100% likelihood of the nodule being invasive. This further indicates that the PPV for invasiveness is the force of this software when used in ACs manifesting as pure basis glass nodules.

On the other paw, our results showed a low sensitivity of approximately 30% for histological invasiveness. A previously published airplane pilot study reported a sensitivity of 95.four to 98.seven% and a negative predictive value of 87.five to 96.8% for a unlike histological subcategorization of pulmonary adenocarcinomas [1]. Therefore, and considering this study also included both, solid and role-solid lesions, the diagnostic performance of these two studies are difficult to compare.

Previous studies reported lesion size to exist an of import predictor for invasiveness in ACs manifesting as pure ground glass nodules [19, 20]. Lee et al. showed that an overall lesion size of < 10 mm can be used to differentiate between pre-invasive and invasive lesions [21]. In contrast, Liu et al. reported lesions smaller than 10 mm likewise to be invasive ACs [10]. In our written report, we focused on component size rather than on overall lesion size and our results showed a rather weak but statistically meaning relation between the size of the invasive focus and the relative percentages of the three CANARY components. This suggests that size can only must not be considered a relevant factor for predicting invasiveness. Moreover, size thresholds are controversial. Therefore, the analysis software could better facilitate in the prediction of invasiveness of ACs. This prediction is important as patients with suspected invasive nodules are more than probable to undergo more follow-upwardly examinations, earlier biopsies, and more radical surgical resections [17, 22]. In particular, the software may exist used in high risk patients.

Our study has several limitations. First, our study included just surgically resected ACs manifesting every bit pure ground glass nodules that were surgically resected and histologically proven to be ACs. For this reason, our nodule sample might have been skewed towards larger, morphologically more conspicuous or aggressively behaving ground drinking glass nodules. However, this particular written report design warranted a pathologically homogeneous sample of nodules. Prospective study information will be required to confirm our electric current results. Second, CANARY analysis is a semiautomatic software tool, requiring the manual drawing of regions of interest, a time consuming cistron that may limit its clinical utilize. Third, our written report may be limited by the variability in CT protocols. However, at the time of the CT acquisitions, each individual CT scanner was considered to be state-of-the-fine art. All image data were reconstructed with standard reconstruction algorithms, which accept been described to be similar between unlike scanners [23]. Furthermore, to exclude the influence of dissimilarity media administration and section thickness, we included these variables in our statistical analyses. 4th, we did not accost whether the software is able to differentiate between beneficial and malign lesions. However, the density reduction algorithm was developed by using but pulmonary nodules that were histologically proven adenocarcinomas of the lung. Although we believe that the differentiation between beneficial or malign lesions is of import in daily routine, the focus of our study was to test the ability of CANARY software to differentiate betwixt histological subtypes of adenocarcinomas manifesting equally pure basis glass nodules. Further studies will have to assess whether CANARY can differentiate betwixt beneficial and malignant lesions.

In conclusion, our study showed that CANARY-based hazard stratification of ACs manifesting as pure ground drinking glass nodules allows the differentiation of AIS, MIA and IACs, the histological subtypes of these lesions. Moreover, CANARY software reflects invasiveness in this group of ACs manifesting every bit pure basis drinking glass nodules using a threshold of 45% or lower of low risk components. While the precise role of CANARY in the work-up of ACs remains to be determined, our study suggests a potential role for this software in suspected ACs manifesting as pure ground glass nodules on CT, notably for patients with loftier hazard of invasiveness.

Electronic supplementary fabric

Below is the link to the electronic supplementary textile.

ESM 1^{(15K, gif)}

(GIF 14 kb)

Acknowledgments

Open up access funding provided by Medical University of Vienna. The authors wish to acknowledge Ms. Donna Wolfe (Beth Israel Deaconess Medical Center, Boston, MA, USA), for her outstanding support in editing the manuscript.

Compliance with ethical standards

Guarantor

The scientific guarantor of this publication is Prof. Dr. Alexander A. Bankier.

Disharmonize of interest

The authors of this manuscript declare relationships with the following companies: Michael Westmore is employed at Imbio, LLC (Minneapolis, MN).

Funding

The authors country that this work has not received any funding.

Statistics and biometry

Ane of the authors has meaning statistical expertise.

Informed consent

Written informed consent was waived by the Institutional Review Board.

Upstanding approval

Institutional Review Board approving was obtained.

Study subjects or cohorts overlap

Our report population has been previously reported in "Heidinger BH, et al. Size measurement and T-staging of lung adenocarcinomas manifesting as solid nodules ≤ xxx mm on CT: radiology-pathology correlation. Acad Radiol. 2017", "Anderson KR, et al. Measurement Bias of Gross Pathologic every bit Compared to Radiologic Tumor Size of Resected Lung Adenocarcinomas; Implications for the T-Phase Revisions in the 8th Edition of the AJCC Cancer Staging Transmission. American Journal of Clinical Pathology 147 (6):641-648. 2017" and "Heidinger BH, et al. Lung adenocarcinoma manifesting as pure footing-glass nodules: Correlating CT size, book, density, and shape with histopathologic invasion. Journal of Thoracic Oncology. 2017".

Methodology

• retrospective

• observational

• performed at i establishment

Footnotes

Ursula Nemec and Benedikt H. Heidinger contributed equally to this work.

Electronic supplementary cloth

The online version of this article (doi:ten.1007/s00330-017-4937-two) contains supplementary textile, which is available to authorized users.

References

ane. Maldonado F, Boland JM, Raghunath S, et al. Noninvasive characterization of the histopathologic features of pulmonary nodules of the lung adenocarcinoma spectrum using figurer-aided nodule cess and risk yield (CANARY)--a pilot written report. J Thorac Oncol. 2013;viii:452–460. doi: x.1097/JTO.0b013e3182843721. [PMC gratis article] [PubMed] [CrossRef] [Google Scholar]

2. Raghunath S, Maldonado F, Rajagopalan S, et al. Noninvasive risk stratification of lung adenocarcinoma using quantitative computed tomography. J Thorac Oncol. 2014;ix:1698–1703. doi: 10.1097/JTO.0000000000000319. [PMC free article] [PubMed] [CrossRef] [Google Scholar]

3. Maldonado F, Duan F, Raghunath SM, et al. Noninvasive Computed Tomography-based Risk Stratification of Lung Adenocarcinomas in the National Lung Screening Trial. Am J Respir Crit Care Med. 2015;192:737–744. doi: 10.1164/rccm.201503-0443OC. [PMC free article] [PubMed] [CrossRef] [Google Scholar]

4. MacMahon H, Naidich DP, Goo JM, et al. Guidelines for Management of Incidental Pulmonary Nodules Detected on CT Images: From the Fleischner Society 2017. Radiology. 2017 [PubMed] [Google Scholar]

5. Lantuejoul Southward, Rouquette I, Brambilla E, Travis WD. New WHO nomenclature of lung adenocarcinoma and preneoplasia. Ann Pathol. 2016;36:five–14. doi: ten.1016/j.annpat.2015.11.010. [PubMed] [CrossRef] [Google Scholar]

6. Kevin R. Anderson, Benedikt H. Heidinger, Yigu Chen, Alexander A. Bankier, Paul A. VanderLaan, (2017) Measurement Bias of Gross Pathologic Compared With Radiologic Tumor Size of Resected Lung Adenocarcinomas. American Journal of Clinical Pathology 147 (6):641-648 [PubMed]

7. Benedikt H. Heidinger, Kevin R. Anderson, Eoin Thousand. Moriarty, Daniel B. Costa, Sidhu P. Gangadharan, Paul A. VanderLaan, Alexander A. Bankier, (2017) Size Measurement and T-staging of Lung Adenocarcinomas Manifesting as Solid Nodules ?30?mm on CT. Academic Radiology [PubMed]

8. Benedikt H. Heidinger, Kevin R. Anderson, Ursula Nemec, Daniel B. Costa, Sidhu P. Gangadharan, Paul A. VanderLaan, Alexander A. Bankier, (2017) Lung adenocarcinoma manifesting equally pure ground-glass nodules: Correlating CT size, book, density, and roundness with histopathologic invasion and size. Periodical of Thoracic Oncology [PubMed]

ix. Travis WD BE, Shush AP, Marx A, Nicholson AG (2015) WHO Nomenclature of Tumours of the Lung, Pleura, Thymus and Center. Fourth edition. IARC [PubMed]

10. Liu LH, Liu M, Wei R, et al. CT findings of persistent pure ground glass opacity: tin we predict the invasiveness? Asian Pac J Cancer Prev. 2015;sixteen:1925–1928. doi: x.7314/APJCP.2015.16.5.1925. [PubMed] [CrossRef] [Google Scholar]

11. Cho J, Ko SJ, Kim SJ, et al. Surgical resection of nodular ground-glass opacities without percutaneous needle aspiration or biopsy. BMC Cancer. 2014;14:838. doi: 10.1186/1471-2407-fourteen-838. [PMC gratuitous article] [PubMed] [CrossRef] [Google Scholar]

12. Garfield DH, Cadranel JL, Wislez Thousand, Franklin WA, Hirsch FR. The bronchioloalveolar carcinoma and peripheral adenocarcinoma spectrum of diseases. J Thorac Oncol. 2006;1:344–359. doi: 10.1016/S1556-0864(15)31593-8. [PubMed] [CrossRef] [Google Scholar]

13. Tsukada H, Satou T, Iwashima A, Souma T. Diagnostic accurateness of CT-guided automated needle biopsy of lung nodules. AJR Am J Roentgenol. 2000;175:239–243. doi: 10.2214/ajr.175.1.1750239. [PubMed] [CrossRef] [Google Scholar]

14. Detterbeck FC, Mazzone PJ, Naidich DP, Bach Pb. Screening for lung cancer: Diagnosis and management of lung cancer, 3rd ed: American College of Chest Physicians evidence-based clinical practice guidelines. Chest. 2013;143:e78S–92S. doi: x.1378/chest.12-2350. [PMC costless article] [PubMed] [CrossRef] [Google Scholar]

xv. Kobayashi Y, Mitsudomi T. Management of basis-glass opacities: should all pulmonary lesions with ground-drinking glass opacity be surgically resected? Transl Lung Cancer Res. 2013;two:354–363. [PMC free article] [PubMed] [Google Scholar]

16. Tang ER, Schreiner AM, Pua BB. Advances in lung adenocarcinoma nomenclature: a summary of the new international multidisciplinary classification system (IASLC/ATS/ERS) J Thorac Dis. 2014;half dozen:S489–501. [PMC costless article] [PubMed] [Google Scholar]

17. Naidich DP, Bankier AA, MacMahon H, et al. Recommendations for the direction of subsolid pulmonary nodules detected at CT: a statement from the Fleischner Society. Radiology. 2013;266:304–317. doi: 10.1148/radiol.12120628. [PubMed] [CrossRef] [Google Scholar]

18. Lim HJ, Ahn S, Lee KS, et al. Persistent pure basis-glass opacity lung nodules >/= 10 mm in bore at CT browse: histopathologic comparisons and prognostic implications. Chest. 2013;144:1291–1299. doi: 10.1378/chest.12-2987. [PubMed] [CrossRef] [Google Scholar]

xix. Mao H, Labh Thou, Han F, Jiang S, Yang Y, Sun X. Diagnosis of the invasiveness of lung adenocarcinoma manifesting as ground glass opacities on high-resolution computed tomography. Thorac Cancer. 2016;seven:129–135. doi: 10.1111/1759-7714.12269. [PMC costless article] [PubMed] [CrossRef] [Google Scholar]

20. Xiang W, Xing Y, Jiang South, et al. Morphological factors differentiating betwixt early lung adenocarcinomas appearing as pure ground-drinking glass nodules measuring </=10 mm on sparse-department computed tomography. Cancer Imaging. 2014;14:33. doi: 10.1186/s40644-014-0033-ten. [PMC free commodity] [PubMed] [CrossRef] [Google Scholar]

21. Lee SM, Park CM, Goo JM, Lee HJ, Wi JY, Kang CH. Invasive pulmonary adenocarcinomas versus preinvasive lesions appearing equally ground-glass nodules: differentiation by using CT features. Radiology. 2013;268:265–273. doi: ten.1148/radiol.13120949. [PubMed] [CrossRef] [Google Scholar]

22. Van Schil PE, Balduyck B, De Waele M, Hendriks JM, Hertoghs Thou, Lauwers P. Surgical treatment of early on-phase non-pocket-sized-jail cell lung cancer. EJC Suppl. 2013;11:110–122. doi: 10.1016/j.ejcsup.2013.07.021. [PMC free article] [PubMed] [CrossRef] [Google Scholar]

23. Boedeker KL, McNitt-Grayness MF, Rogers SR, et al. Emphysema: effect of reconstruction algorithm on CT imaging measures. Radiology. 2004;232:295–301. doi: 10.1148/radiol.2321030383. [PubMed] [CrossRef] [Google Scholar]

merrittwoughat.blogspot.com

Source: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5717124/

Share this post