Postoperative Pulmonary Function Assessment Based on Deep Learning Study

Multimodal Machine Learning Integration of Pulmonary Function Parameters and 3D-CT Imaging for Postoperative Outcome Prediction in Early-Stage NSCLC

Improvements in low-dose CT screening have led to an increase in early-stage NSCLC diagnoses, with surgical resection-usually lobectomy or segmentectomy-remaining the primary curative option. In Hong Kong, however, patients frequently present with comorbidities such as chronic respiratory disease or cardiovascular issues, making the preservation of healthy lung tissue crucial for their long-term quality of life. Traditional surgical resection, such as lobectomy or segmentectomy, has certain limitations in terms of functional preservation. In contrast, robotic/navigational bronchoscopic ablation has emerged in recent years as a novel minimally invasive endoscopic treatment strategy. This approach has been implemented in select centers and demonstrates potential advantages, including faster postoperative recovery, reduced trauma, and improved preservation of pulmonary function. By leveraging advanced navigation systems, bronchoscopic ablation enables precise localization and ablation of pulmonary nodules, avoiding the extensive resection of healthy lung tissue required in traditional surgery. These benefits hold promise for enhancing patients' long-term quality of life and survival rates.

Moreover, conventional pulmonary function tests like FEV₁ and diffusing capacity of the lung for carbon monoxide provide only a global assessment of respiratory capacity, which may not fully capture the regional changes in pulmonary function that occur following segmentectomy or lobectomy. Likewise, basic CT volumetry overlooks finer anatomical details such as segmental airway distribution, microvascular networks, and local alveolar compliance. Furthermore, there is currently a paucity of direct comparative studies between robotic/navigational bronchoscopic ablation and traditional surgical resection regarding postoperative pulmonary function and long-term outcomes. Supported by Research Grants Council, our work since 2019 has validated the feasibility and safety of this technique, leading to widespread recognition and numerous publications. However, most existing research is retrospective or derived from single-center data, with a primary focus on short-term safety and technical feasibility.

To address these limitations, an integrative approach leveraging 3D-CT imaging and ML is proposed. Machine learning is a technology that uses algorithms to automatically learn from data and make predictions or decisions. Deep learning, a subfield of ML, utilizes multi-layer neural networks to effectively extract features and recognize patterns in complex, high-dimensional data. ML, and particularly DL, has demonstrated remarkable potential in various medical imaging applications, including lesion detection, tissue segmentation, and outcome prediction. By automatically learning complex patterns in high-dimensional data, ML and DL models can interpret subtle radiologic characteristics that may be missed by conventional analyses. In the context of 3D-CT imaging for NSCLC, DL architectures-such as convolutional neural networks-can extract detailed features from volumetric scans, enabling robust quantification of tumor size, shape, and location as well as refined assessment of lung parenchyma. When integrated with pulmonary function parameters and clinical data, these algorithms provide a powerful means to generate predictive models, identify at-risk patients earlier, and guide individualized treatment planning. Moreover, ML-driven approaches can adapt to evolving datasets over time, continuously refining and improving their performance. This scalability and adaptability are especially valuable in prospective studies, where large, multimodal datasets are collected to evaluate the long-term impact of different treatment strategies. Consequently, incorporating ML and DL in this research not only enhances the precision of outcome prediction but also contributes to a standardized framework for dynamic, personalized assessment of pulmonary function, guiding more informed clinical decision-making.

The primary aim of this study is to determine whether segmentectomy truly offers better functional preservation than lobectomy, whether robotic/navigation-guided bronchoscopic ablation indeed achieves superior pulmonary function preservation compared to traditional surgical resection, and under which specific patient conditions each approach may yield the greatest benefit. By undertaking a prospective, well-designed investigation, the research will fill a critical gap in evidence regarding long-term functional outcomes, providing clearer criteria for selecting the most appropriate resection type. Moreover, the introduction of a standardized, integrative assessment tool has the potential to optimize surgical decision-making and postoperative care, ultimately improving survival and quality of life for early-stage NSCLC patients in Hong Kong and potentially informing best practices in other healthcare contexts.

Study Overview

• Status: Not yet recruiting
• Conditions: Lung Cancer (NSCLC)
• Intervention / Treatment: Procedure: Segmentectomy; Procedure: Lobectomy; Procedure: Bronchoscopic Ablation

• Study Type: Interventional
• Enrollment (Estimated): 192
• Phase: Not Applicable

Contacts and Locations

• Study Contact: Name: Calvin Sze Hang Ng; Phone Number: +852 3505 2618; Email: calvinng@surgery.cuhk.edu.hk

Study Locations

• Hong Kong
  • Shatin, Hong Kong
    • Prince of Wales Hospital
    • Contact: Calvin Sze Hang Ng; Phone Number: +852 3505 2618; Email: calvinng@surgery.cuhk.edu.hk

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

Description

Inclusion Criteria:

1. Age ≥18 years
2. Histologically or cytologically confirmed stage IA NSCLC (T1N0M0, tumor ≤3 cm)
3. Scheduled for surgical treatment (segmentectomy or lobectomy or robotic/navigation-guided bronchoscopic ablation)
4. Ability to complete postoperative visits for up to 2 years
5. Provided informed consent

Exclusion Criteria:

1. Presence of lymph node involvement (N1 or higher) or distant metastases (M1)
2. Concomitant primary malignancies that could confound outcome analysis
3. Insufficient or missing key data points (e.g., tumor size, treatment type, survival status)
4. Duplicate or overlapping records from different data sources
5. Contraindications to anesthesia or sedation for bronchoscopy or surgery
6. Inability or unwillingness to perform the PF tests
7. Pregnant or breastfeeding women (for prospective phase)
8. Inability or unwillingness to comply with study procedures

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Other
• Allocation: Non-Randomized
• Interventional Model: Parallel Assignment
• Masking: None (Open Label)

Number of Arms

3

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Experimental: Segmentectomy Group	Procedure: Segmentectomy; with 64 participants
Experimental: Lobectomy Group	Procedure: Lobectomy; with 64 participants
Experimental: Bronchoscopic Ablation Group	Procedure: Bronchoscopic Ablation; with 64 participants

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Develop and validate a composite PF score	Develop and validate a composite PF score by applying ML algorithms to integrate pulmonary function test results (FEV₁, FVC, DLCO), 3D-CT volumetric data (segmental volumes, vascular density) (Figure 1), and basic demographic variables (age, BMI). The primary objective is to predict changes in postoperative pulmonary function at 1, 3, 6, 12, 18, and 24 months.	1, 3, 6, 12, 18, and 24 months (post-operation)

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
To compare the time trajectory of the PF score	Compare the time trajectory of the PF score among patients undergoing different treatment approaches-segmentectomy, lobectomy, and robotic/navigation-guided bronchoscopic ablation-to determine which method offers a superior advantage in long-term postoperative pulmonary function preservation.	at 1, 3, 6, 12, 18, and 24 months (post-operation)
To Correlate the PF score with overall survival, cancer-specific survival, and postoperative complications	Correlate the PF score with overall survival, cancer-specific survival, and postoperative complications, assessing its utility as a clinical endpoint.	at 1, 3, 6, 12, 18, and 24 months (post-operation)
To identify key predictive factors-including tumor location, histology, and baseline comorbidities	Identify key predictive factors-including tumor location, histology, and baseline comorbidities-that may influence postoperative functional decline or survival. Additionally, conduct a preliminary investigation into the application of robotic/navigation-guided bronchoscopic ablation in high-risk patients.	at 1, 3, 6, 12, 18, and 24 months (post-operation)

Collaborators and Investigators

• Sponsor: Chinese University of Hong Kong

Study record dates

Study Major Dates

• Study Start (Estimated): July 1, 2026
• Primary Completion (Estimated): June 30, 2029
• Study Completion (Estimated): December 31, 2029

Study Registration Dates

• First Submitted: November 20, 2025
• First Submitted That Met QC Criteria: November 20, 2025
• First Posted (Actual): December 1, 2025

Study Record Updates

• Last Update Posted (Actual): December 1, 2025
• Last Update Submitted That Met QC Criteria: November 20, 2025
• Last Verified: November 1, 2025

More Information

Terms related to this study

• Keywords: Non-small cell lung cancer; Machine learning; Segmentectomy; Lobectomy; Robotic bronchoscopy; Navigational bronchoscopy; Postoperative outcomes; Pulmonary function score; 3D-CT imaging
• Additional Relevant MeSH Terms: Neoplasms by Site; Neoplasms; Respiratory Tract Diseases; Lung Diseases; Respiratory Tract Neoplasms; Thoracic Neoplasms; Carcinoma, Bronchogenic; Bronchial Neoplasms; Lung Neoplasms; Carcinoma, Non-Small-Cell Lung; Surgical Procedures, Operative; Mastectomy; Neurosurgical Procedures; Mastectomy, Segmental; Anterior Temporal Lobectomy
• Other Study ID Numbers: Protocol version 1.0 (UW Madison)

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No