Tool box of lung nodule predictors


Introduction

Lung cancer has become a common clinical problem, which always performs as a pulmonary nodule/masses on imaging.

The accurate assessing of the pulmonary nodules/masses decides the diagnosis and treatment of patients.

Recently, numbers of clinical models combining patient clinical characteristics with other features evaluate the malignancy and benignancy of the pulmonary nodules/masses.

Here, a tool estimating the probability of malignancy has been developed.

Tool box of lung nodule predictors is a scalable and one-stop web platform, with SCHC model (Sichuan Hospital of Cancer model), BU model (Brock University model), BU simple model, MC model (Mayo Clinic Model), VA model and TCRnoseek model (TCR nodule seek model).

SCHC model combines platelet features with nodule imaging features to estimate a probability of malignancy.

BU model, BU simple model, MC model, VA model calculate the malignant probability with chest radiographic features, and clinical characteristics.

TCRnoseek model provides the malignant probability based on the clinical characteristics and TCR ShannonIndex.

It provides a simpler web, which can be performed to estimate a probability of malignancy by clinicians, as well as suspicious patients.

If you apply this above model in academic work, please cite corresponding work below:

BU model : Chung K, Mets OM, Gerke PK, Jacobs C, den Harder AM, Scholten ET, Prokop M, de Jong PA, van Ginneken B, Schaefer-Prokop CM. Brock malignancy risk calculator for pulmonary nodules: validation outside a lung cancer screening population. Thorax 2018, 73(9): 857-863.

MC model: Swensen SJ, Silverstein MD, Ilstrup DM, Schleck CD, Edell ES. The probability of malignancy in solitary pulmonary nodules. Application to small radiologically indeterminate nodules. Archives of internal medicine 1997, 157(8): 849-855.

VA model : Gould MK, Ananth L, Barnett PG. A clinical model to estimate the pretest probability of lung cancer in patients with solitary pulmonary nodules. Chest 2007, 131(2): 383-388.

SCHC model: Waiting accepted

TCRnodseek model: Waiting accepted

The team members of tool box of lung nodule predictors mainly come from Sichuan Cancer Hospital & Institute and University of Electronic Science and Technology of China.

If you need more details about the models and tool, or have any comments, feel free to contact us.

Company name and address: Department of Clinical Laboratory,

Sichuan Cancer Hospital & Institute,

Sichuan Cancer Center,

School of Medicine,

University of Electronic Science and Technology of China,

No 55, Section 4, South People's Road, Chengdu 610041,

Sichuan, People's Republic of China;

Tel: + 862885420344; Email: hcluo@std.uestc.edu.cn.

SCHC model

Description: SCHC model was constructed using eXtreme Gradient Boosting (XGBoost) based on patient’s age, nodule/mass size and platelet features.

Required input



Enter the required features in the box. Age represents the age of patient in years; pPLT represents platelet counts in platelet rich plasma sample (×109/L); pPCT represents plateletcrit in platelet rich plasma sample; size represents the largest diameter of nodule/mass in millimeter (mm); bPCT represents plateletcrit in whole blood sample.

After clicking on the bottom of the web page, the probability of malignancy would display as the bar box. And in the blank box, it would display “malignant” if the probability is over than cut-off value, otherwise it would display “benign”.



BU model

Description: The BU model also named Brock model was developed from the Pan-Canadian Early Detection of Lung Cancer Study, which included two sets of models: a parsimonious model and a full model. The BU model calculated the probabilities using multivariable logistic regression based on the variables including age, gender, family history of lung cancer, presence of COPD, nodule size, nodule location, nodule counts, and nodule characteristics.

Required input





Enter the required features in the box. Age is age in years; nodule count represents the count of the nodule/mass on imaging; size represents the largest diameter of nodule/mass in millimeter (mm); PSN, Gender, COPD, GGN, Up, Spiculation box are selectable in the list (click the

button).

PSN Option

Y represents the nodule/mass is semisolid/part-solid on imaging, N represents the nodule/mass is solid on imaging.

Gender Option

There are male and female can be chosen in the electable list.

COPD Option

If the patient has COPD presently, Y should be chosen; otherwise, N should be chosen.

GGN Option

Y represents the nodule/mass is ground-glass on imaging, N represents otherwise.

Up Option

Y represents the nodule/mass locates in an upper lobe; N represents otherwise.

Speculation Option

Y represents if the edge of nodule/mass has spicules, N represents otherwise.

After clicking on the bottom of the web page, the probability of malignancy would display as the bar box.



BU simple model

Description: The BU simple model is the parsimonious model of BU model, which includes size, gender, up, and speculation.

Required input



Enter the required features in the box. Size represents the largest diameter of nodule/mass in millimeter (mm); Gender, Up, Spiculation box are selectable in the list (click the button).

Gender Option

There are male and female can be chosen in the electable list.

Up Option

Y represents the nodule/mass locates in an upper lobe; N represents otherwise.

Speculation Option

Y represents if the edge of nodule/mass has spicules, N represents otherwise.

After clicking on the bottom of the web page, the probability of malignancy would display as the bar box.



MC model

Description: The MC model was proposed by Swensen at the Mayo Clinic, which expresses the probability of malignant nodules as a function of the variables including age, smoking history, cancer history and characteristics of the nodules.

The MC model was constructed using multivariable logistic regression.

Required input



Enter the required features in the box. Age is age in years; size represents the largest diameter of nodule/mass in millimeter (mm); smoke, cancer history, up, spiculation box are selectable in the list (click the button).

Smoke Option

Current/former represents that the patient has been smoking; never represents that the patient has never smoked.

Cancer history Option

Y presents that the patient has a history of extrathoracic cancer diagnosed within 5 years; N represents otherwise.

Up Option

Y represents the nodule/mass locates in an upper lobe; N represents otherwise.

Speculation Option

Y represents if the edge of nodule/mass has spicules; N represents otherwise.

After clicking on the bottom of the web page, the probability of malignancy would display as the bar box.



VA model

Description: The VA model was developed using the data from Department of Veterans Affairs (VA), which calculated the probability of malignancy in patients with solid pulmonary nodules (SPNs) using multivariable logistic regression based on the patients’ smoking history, age and nodule size.

Required input

Enter the required features in the box. Age is age in years; size represents the largest diameter of nodule/mass in millimeter (mm); Years since quitting represents the years that the patients has been quitting smoking; smoke box is selectable in the list (click the button).

Smoke Option

Current/former represents that the patient has been smoking; never represents that the patient has never smoked.

After clicking on the bottom of the web page, the probability of malignancy would display as the bar box.



TCRnoSeek model

Description: TCRnoseek model integrates TCR diversities and clinical information to distinguish indeterminate lung nodules as benign or malignant. TCRnoseek model was constructed based on logistics m regression, after Lasso regression selecting reliable features.





Age is age in years; nodule Shannon index represents the diversity of TCR, which is calculated based on the clonal abundance of all productive TCR sequences indexes; Shannon index = , n represents clone type counts, i represents one specific clone type, and pi represents the reads for i, which were divided by all read counts; size represents the largest diameter of nodule/mass in millimeter (mm); number represents the count of the nodule/mass on imaging;

Gender, smoking, Up, GGN, Spiculation box are selectable in the list (click the button).

Gender Option

There are male and female can be chosen in the electable list.

Smoking Option

Y represents that the patient has been smoking; N represents that the patient has never smoked.

Up Option

Y represents the nodule/mass locates in an upper lobe; N represents otherwise.

GGN Option

Y represents the nodule/mass is ground-glass on imaging, N represents otherwise.

Speculation Option

Y represents if the edge of nodule/mass has spicules, N represents otherwise.

After clicking on the bottom of the web page, the probability of malignancy would display as the bar box.


60%

benign

Malignant

-


60%






60%








60%


60%














60%