Across UK lung cancer care in 2026, artificial intelligence no longer sits in a single place in the patient's journey. It now operates across five sequential decision points — from the moment a shadow appears on a screening CT scan, through risk stratification, surgical planning, robotic biopsy and resection, and into an emerging phase of intraoperative anatomical overlay that exists today only in prototype. Each step on its own is incremental. The clinical change in patient outcomes comes from the integration — what becomes possible when these decision points chain together into one continuous pathway, sometimes a single anaesthetic. This is a UK consultant thoracic surgeon's account of where the AI-augmented pathway stands in 2026, what is delivered now in NHS and private practice, and where the next phase is heading. Mr Lawrence Okiror operates at Guy's and St Thomas' NHS Foundation Trust, London Bridge Hospital, and The Lister Hospital Chelsea.
Last reviewed: May 2026 · Mr Lawrence Okiror FRCS(CTh) FRCSEd(CTh) · GMC 6150382
AI is now active at five sequential decision points in the lung cancer pathway from screening shadow to operating theatre. The first four are deployed in UK practice in 2026; the fifth is in early prototype.
The clinical change is not any single technology. It is the sequencing of AI across the entire pathway — moving what was a three-month, three-decision journey toward a continuous pathway, sometimes a single anaesthetic.
The integrated combined-anaesthetic pathway is offered privately at London Bridge Hospital. GSTT delivered 635 ION robotic bronchoscopy procedures and 153 anatomic resections by Mr Okiror in 2024–25.
The AI conversation in lung cancer has been told as a sequence of point solutions — better detection here, better risk scoring there, better navigation somewhere else. The clinical change in patient outcomes does not come from any one of these.
It comes from sequencing them across the pathway, end to end, so that what was a three-month, three-decision journey becomes one continuous patient pathway, sometimes a single anaesthetic. That is the position from which this page is written: integration as the discriminator, not point-solution accumulation. The next four-thousand words explain what that means clinically, where it is delivered in the UK in 2026, and where the next phase of the pathway is heading. Request a consultation at London Bridge Hospital within 2–3 working days →
For most of the modern history of lung cancer, patients reached the thoracic surgeon late. The presenting symptoms — cough that did not settle, weight loss, breathlessness, haemoptysis — were the manifestations of disease that had already moved beyond what surgery could cure. Stage at presentation was the single greatest determinant of outcome, and stage at presentation was, for most patients, advanced. The operation could be performed; the operation could not always change what was going to happen.
CT screening of high-risk smokers and ex-smokers inverts this pattern. Lee and colleagues, reporting five-year outcomes from the NHS England Lung Cancer Screening Programme in Nature Medicine in March 2026, document a stage shift of historic magnitude [1]. Of the cancers detected through the programme, 76% were Stage I or II at diagnosis — the earliest, most curable, and smallest stages, for which anatomic surgical resection is generally the principal curative modality. Comparable London-specific data from the SUMMIT cohort, reported by Bhamani and colleagues in Lancet Oncology in 2025, broadly confirms the pattern in a different metropolitan population [2]. The change in who reaches the thoracic surgeon, and at what stage, is unambiguous.
The pathway consequence is a different bottleneck. The National Lung Cancer Audit 2026 (NATCAN State of the Nation report) records that UK lung cancer surgical resection volume rose from 6,547 in 2023 to 7,878 in 2024 — a 20% single-year increase directly driven by screening [3]. The same audit reports that 88% of early-stage cancer patients waited longer than the 49-day target from referral to surgery. The system is processing more patients than ever before, but not fast enough to meet its own published standards. Capacity, not detection, is the new constraint.
A further structural problem sits upstream of the operation. Around 15% of people scanned in the screening programme are found to have a pulmonary nodule — the vast majority benign, a small proportion early cancer, and an intermediate proportion indeterminate on the initial scan. The conventional surveillance pathway involves interval CT at three to six to twelve months, with most patients reassured eventually by stability. For some, the interval is the wrong intervention: a higher-risk nodule on the first scan that warranted earlier biopsy spends months in surveillance instead. For others, the interval is excessive: a clearly benign appearance that did not require months of follow-up generates anxiety and repeat radiation exposure.
This is the front-end volume problem that the AI-augmented pathway exists to address. It is not principally about replacing radiologists or reading scans faster. It is about applying objective quantitative tools to the heterogeneity of risk at the moment a shadow first appears, so that the right pathway — surveillance, biopsy, or resection — is chosen earlier and with less variance between centres. The remainder of this page is an account of the five sequential AI decision points that now sit between that first abnormal CT and the operative outcome.
Figure 1. The five-phase AI-augmented lung cancer pathway in UK practice, 2026.
The clinical problem AI risk stratification addresses is the indeterminate pulmonary nodule. A perfectly clear-cut benign nodule does not need a probability score; nor does a clearly malignant one. The challenge is the middle ground — the 6 mm or 8 mm nodule whose size, density, and morphology sit in the range where Brock-derived clinical probability ranges from 5% to 25%, where the next decision is plausibly surveillance, plausibly biopsy, and plausibly straight to resection depending on context. In conventional UK pathways the decision is made at lung nodule MDT, applying British Thoracic Society guidelines and clinician judgement. The challenge is variability: the same nodule with the same imaging can yield different decisions at different centres or with different clinicians.
The Brock model and Mayo Clinic model were the major advances in this area before AI [4, 5]. Both combine clinical variables — age, smoking history, family history, nodule diameter, spiculation, upper-lobe location — into a single malignancy probability. They are useful and widely embedded in clinical guidelines. They have two limitations. First, they capture morphological information through coarse categorical variables, missing the rich image-derived features that radiologists describe verbally but cannot codify. Second, they perform less well at the smallest nodule sizes (typically under 8 mm) where size alone — the dominant predictor in clinical models — is too low to be informative.
AI algorithms trained on large nodule datasets — typically several thousand patients with histology-confirmed outcomes — learn image-derived features that are difficult or impossible to specify by hand. The resulting probability score draws on texture, surface roughness, internal heterogeneity, vessel relationships, and pattern features that human readers describe but do not quantify. The clinical purpose is twofold: to reduce inter-reader variability in how indeterminate nodules are characterised, and to apply quantitative scoring to small nodules in the size range where clinical models alone are uninformative.
Two algorithms are now FDA-cleared and/or EU-MDR-certified for clinical use in this step. Both have published peer-reviewed evidence on diagnostic performance against histology-confirmed outcomes, demonstrating area-under-curve metrics that outperform clinical models alone in the indeterminate-nodule range. UK reimbursement frameworks for AI-augmented radiology are now in development, with NICE assessment of risk-stratification AI ongoing. This page is descriptive of the category and does not name specific vendors; both leading platforms are referenced in current peer-reviewed literature and in regulator-approved labelling.
In January 2026, NHS England announced a national pilot integrating AI-based risk stratification of nodules with the Targeted Lung Health Check screening programme, alongside concurrent rollout of robotic bronchoscopy access for nodules requiring biopsy [6]. Guy's and St Thomas' is among the lead participating centres. The pilot is structured to evaluate two questions in parallel: whether AI risk stratification reduces unnecessary surveillance and unnecessary biopsy compared with current practice, and whether it accelerates the route to treatment for the higher-risk subset. The pilot is not yet a basis for national reimbursement or routine NHS adoption; that decision will follow the published evaluation.
The clinical placement of AI risk stratification is at the lung nodule MDT, where the score sits alongside the radiologist's read, the patient's clinical context, and pre-existing risk model output. The decision — surveillance, biopsy, or direct resection — is made by the MDT, not by the algorithm. The added value of the algorithm is the quantitative input on image-derived features, particularly in the small-nodule range and in the borderline cases where clinical risk models alone leave the next decision uncertain.
Lung segmental anatomy is not standard. Each patient has their own configuration of segmental bronchi, pulmonary arteries, and pulmonary veins, and the variations between patients are clinically significant. For a lobectomy, this matters less — the entire lobe is removed along its lobar pedicle, and intralobar variation is encountered during the operation rather than planned around it. For a segmentectomy — where the operation removes a single anatomic segment along its own vascular and airway pedicle, preserving the remaining segments of the lobe — the variation matters more. The operative plane runs between two adjacent segments along their individual anatomy. Anticipating that anatomy before the operation, rather than discovering it intraoperatively, is the difference between a planned operation and an unplanned one.
3D anatomical reconstruction software takes the patient's pre-operative CT and produces a colour-coded virtual model of their own lung. The pulmonary arteries are one colour, the pulmonary veins another, the bronchi a third, the lobar fissures a fourth. The tumour can be rendered separately and shown in relation to the surrounding segmental anatomy. The model can be rotated, sectioned, and inspected on a tablet or workstation. For the surgeon planning a segmentectomy, this allows three things to be done before the operation: identification of the specific segmental vessels and bronchi that will need to be divided, estimation of the safe parenchymal margin around the tumour, and anticipation of variant anatomy — a non-standard branching pattern or an anomalous vein — before encountering it on the operating table.
Several manufacturers now offer 3D reconstruction platforms for thoracic surgical planning. The technology has matured from the experimental stage of the mid-2010s into a commercial offering with multiple competing products, integrated with picture archiving and communication systems and increasingly available to UK thoracic units. The category is moving; the page is descriptive of the category, not of any single platform.
Mr Okiror uses 3D anatomical reconstruction in selected segmentectomy cases — particularly where the patient's segmental anatomy is unusual, where the tumour sits close to a segmental boundary and the safe-margin question is non-trivial, or where the choice between two adjacent segmental approaches is anatomically delicate. Not every segmentectomy needs a 3D reconstruction. For straightforward apical or basal segments with a standard vascular pattern, the operation is planned from the 2D CT in the same way it has been for years.
The cases where 3D reconstruction earns its keep are the complex ones: combined-segment resections, segmentectomy abutting a fissure or a major vessel, patients with prior lung resection or significant emphysema where the residual anatomy is distorted, and bilateral cases where the operative plan needs to consider both lungs together. In these cases, the reconstruction is shared with the patient at the pre-operative consultation, which materially improves how the operation is understood. Patients can see their own anatomy, the location of the tumour, the segments that will be removed, and the segments that will be preserved.
The screening cohort is the population in which 3D reconstruction is increasingly relevant. Screen-detected cancers are typically small and peripheral — the cancers JCOG0802 and CALGB 140503 have established as candidates for segmentectomy [7, 8]. As the proportion of resections performed as segmentectomy rather than lobectomy continues to rise — at GSTT, 24.2% of anatomic resections in 2024–25 were segmentectomies, with 88.9% performed robotically — the case for routine 3D planning in selected cases strengthens. Patients having a parenchyma-sparing operation benefit most from the parenchyma-sparing being precise.
The diagnostic problem in screen-detected lung cancer is reaching the lesion. The cancers being found are increasingly small — 8 to 20 mm is now common — and increasingly peripheral, sitting in the outer third of the lung that conventional flexible bronchoscopy cannot access. CT-guided percutaneous needle biopsy can reach these lesions but at substantial cost: pneumothorax rates of 20–30% in published series, with hospitalisation rates that are non-trivial. For some peripheral nodules, particularly those in deep lung parenchyma or near major vessels, CT-guided biopsy is technically possible but practically unsafe.
Robotic navigational bronchoscopy was developed to solve this problem. The ION platform, the principal system in UK practice, uses shape-sensing technology to track the position of a slim catheter (3.5 mm) inside the airway with sub-millimetre accuracy. Before the procedure, the platform processes the patient's CT and generates a three-dimensional airway tree, mapping the bronchial branching pattern out to its peripheral terminations. The operator selects the target lesion and the platform plans the optimal navigation route — the airway path that will bring the catheter closest to the lesion. During the procedure the operator drives the catheter along this path, with real-time visual feedback of the catheter position relative to the airway and the planned target.
Cone beam CT, available in the same operating suite, provides intraoperative confirmation of positional accuracy. A short scan is performed with the catheter in place; the image is reconstructed in seconds; the operator verifies that the catheter tip is at the lesion before the biopsy is taken. This step matters particularly in the small-lesion range, where a positional error of a few millimetres is the difference between a diagnostic and a non-diagnostic biopsy. The combined ION-plus-cone-beam-CT approach now reports diagnostic yields of 76–90% across published series, against pneumothorax rates of 2–3% — substantially safer than CT-guided needle biopsy.
A cytopathologist is present in the operating theatre during ION procedures used in the combined biopsy-to-resection pathway. The biopsy sample is assessed under the microscope within minutes of being taken, providing a preliminary indication of whether malignant cells are present. ROSE is not a definitive histological diagnosis — the formal pathology follows in the laboratory in the usual way — but it provides the in-theatre signal that determines whether the combined-anaesthetic pathway proceeds to immediate resection or stops at the bronchoscopy. The combination of ION navigation, cone beam CT confirmation, and ROSE assessment is the structural reason a single-anaesthetic biopsy-to-resection pathway is feasible for selected patients.
If the preliminary ROSE assessment signals non-small cell lung cancer and the decision is made to proceed to resection, the lesion is dye-marked through the same ION catheter. A small volume of dye is injected into the parenchyma at the lesion site, becoming visible on the pleural surface during subsequent robotic surgery. This intraoperative localisation guides the surgeon to the correct lung segment, eliminating the need for separate pre-operative localisation procedures (such as hookwire placement) that have their own complication rates.
Mr Okiror heads the GSTT interventional bronchoscopy and airways service, which performs 635 ION procedures per year and is among the highest-volume robotic bronchoscopy services in Europe. Privately, ION is available at London Bridge Hospital, the principal private venue for the combined biopsy-to-resection pathway. The integration of ION with subsequent robotic resection — under the same anaesthetic where appropriate — is detailed in the combined biopsy and robotic surgery pathway page.
The clinical evidence base for robotic anatomic resection has matured substantially over the past five years. Two landmark trials — JCOG0802 (Saji and colleagues, Lancet 2022) and CALGB 140503 (Altorki and colleagues, New England Journal of Medicine 2023) — established segmentectomy as equivalent to lobectomy in the right patient population: small (≤2 cm) peripheral non-small-cell lung cancers without nodal involvement [7, 8]. JCOG0802 showed superior overall survival with segmentectomy in this group; CALGB 140503 confirmed non-inferiority for disease-free survival. The clinical implication is significant: as the screen-detected cancer population grows, segmentectomy — preserving more healthy lung parenchyma — becomes the preferred operation for the larger proportion of cases that meet the trial-defined criteria.
The da Vinci Xi robotic platform enables anatomic resection through 3 to 4 small incisions of 1–2 cm. The surgeon sits at a console and controls every instrument movement in real time. No ribs are spread or divided. The platform's three-dimensional vision, articulating wristed instruments, and tremor-filtered movement allow precise dissection around segmental vessels and bronchi — precision that matters particularly in segmentectomy, where the operative plane runs between two adjacent anatomical units.
At GSTT in 2024–25, 71.3% of all anatomic lung cancer resections were performed robotically — the highest robotic share in the UK by SCTS national audit volume. The numbers behind that share are detailed in the table below.
| Metric (GSTT, SCTS 2024–25) | Value |
|---|---|
| Primary lung cancer resections — total | 969 (approximately 12% of the UK total) |
| Anatomic resections | 892 (646 lobectomy + 216 segmentectomy + 17 sleeve + 13 pneumonectomy) |
| Robotic share of anatomic resections | 71.3% |
| Lobectomy — robotic share | 66.6% |
| Segmentectomy — robotic share | 88.9% (192 of 216) |
| Operative survival rate | 99.59% (4 deaths in 969 cases) |
| Total thoracic surgical volume | 2,218 cases — 39.4% robotic |
Mr Okiror personally performed 153 anatomic resections in 2024–25 — approximately 1.94% of the UK national total — with a personal career experience exceeding 1,000 lung cancer operations and over 500 da Vinci robotic cases. The personal robotic segmentectomy count stands at 95 cases. These figures matter not as recitation but because they are the volume-dependent variable on which the safety and quality of robotic anatomic resection depend. Operative survival in 2024–25 at GSTT was 99.59% — the institutional outcome on which the personal practice operates.
Segmentectomy is the operation in which the AI-augmented pathway most coherently expresses itself. The screening cohort produces predominantly small peripheral cancers; AI risk stratification identifies the higher-probability nodules for biopsy; AI 3D reconstruction informs the segmentectomy plan; ION biopsies and dye-marks the lesion; robotic anatomic resection completes the cure. Each phase informs the next. The patient who entered the pathway with a 12 mm shadow on a screening CT leaves it, in the right case, with the affected lung segment removed, the remaining segments preserved, and the operation completed in one anaesthetic via three small incisions.
For a fuller account of robotic segmentectomy — selection, technique, evidence base, and outcomes — see the dedicated robotic segmentectomy page. The Toki et al. real-world series of locally advanced lung cancer at GSTT, on which Mr Okiror is co-author (Lung Cancer 2025), describes outcomes at the other end of the staging spectrum — patients who have completed neoadjuvant chemoimmunotherapy and proceed to anatomic resection [9].
The clinical problem at the back end of the pathway is this: the surgeon, looking at the live operative field, sees the visceral pleural surface of the lung. The underlying segmental anatomy — the bronchi, the segmental arteries, the segmental veins, the lobar fissures — is invisible. The surgeon operates by holding a mental three-dimensional model of that anatomy in their head, built from the pre-operative CT and the 3D reconstruction, and continuously updating it from intraoperative dissection findings. The mental model is good; the integration of pre-operative imaging into the live operative view is, in 2026, still principally cognitive.
The parallel from orthopaedic and neurosurgical operating is informative. Image-guided navigation has been routine in pedicle screw insertion in spine surgery, in deep brain stimulation electrode placement, and in tumour resection in the brain, for the better part of two decades. The pattern in those fields is consistent: a pre-operative imaging dataset is registered to the patient's intraoperative position, and the surgeon's view of the operative field is augmented in real time with the underlying anatomy projected into it. The technology arrived in orthopaedics and neurosurgery before it arrived in thoracic surgery because the anatomical targets in those fields are bony (and therefore more easily registered) and because the patient's anatomy is relatively static during the operation. Thoracic surgery is technically harder — the lung deflates and reflates, the chest wall moves, and the soft-tissue anatomy deforms with each surgical step — but the principle is the same.
For robotic anatomic resection, AR overlay would deliver real-time projection of the patient's segmental anatomy onto the operative field through the robotic console. The surgeon would see, superimposed onto the visceral pleural surface, the position of the segmental artery they are about to clip, the path of the segmental bronchus they are about to divide, and the boundary of the segment they are removing — in the patient's individual anatomy, not a textbook diagram. The cognitive load of holding the mental three-dimensional model would fall. The precision of segmentectomy — particularly in complex or variant anatomy — would improve.
Several manufacturers have prototype platforms in development across the global surgical AR landscape. The current stage is industry demonstration and early evaluation by selected surgical users. There is no peer-reviewed evidence yet that AR overlay improves clinical outcomes in thoracic surgery; the evidence base will need to be built before the technology enters routine clinical use. The reasonable expectation, on the orthopaedic and neurosurgical precedent, is that limited clinical evaluation begins over the next several years, with broader adoption following published evidence and regulatory clearance.
Mr Okiror has seen early prototypes of AR overlay demonstrated at an industry event in London. The page reflects what is in development, not what is in deployment. Patients should not expect AR-guided thoracic surgery to be a routine option in UK practice in 2026; they should expect it to be one in the foreseeable future. This is the phase of the pathway where the integration completes — not because the rest is unfinished, but because the operative-field projection of patient-specific anatomy is the missing piece of an otherwise sequenced system.
The case for the AI-augmented pathway is not made by any single phase. AI risk stratification, taken on its own, is an incremental improvement in nodule MDT decision-making. 3D anatomical reconstruction, taken on its own, is a useful surgical planning aid in selected cases. ION robotic bronchoscopy, taken on its own, is a safer diagnostic technique for peripheral lesions. Robotic anatomic resection, taken on its own, is an established operative approach with well-documented outcomes. Each phase, in isolation, is an incremental gain. The clinical change in patient outcomes comes from sequencing them.
Figure 2. Traditional staged pathway versus integrated AI-augmented pathway. Days replace months for selected patients.
Three things change when the phases chain together rather than running as separate, sequential, multi-week steps.
Time-to-treatment falls. The conventional pathway runs across multiple appointments, multiple departments, and multiple decisions, with administrative latency between each. The integrated pathway compresses these. In its most advanced expression — the combined same-anaesthetic biopsy-to-resection pathway at London Bridge Hospital — the diagnostic and surgical components of the journey happen in one theatre session, with recovery and follow-up to follow. The NLCA 2026 finding that 88% of early-stage patients waited longer than the 49-day target is the measurement of latency in the conventional NHS pathway; the integrated pathway exists in part as a structural answer.
Decision quality improves. Each phase informs the next with quantitative information rather than verbal handover. The AI risk score sits in the MDT documentation. The 3D reconstruction sits in the surgical plan. The intraoperative ROSE result determines whether resection proceeds. The dye marking guides the resection itself. The information flow is structural, not contingent on individual clinicians remembering to pass along key facts. This is the operational characteristic that distinguishes an integrated pathway from a series of well-performed but disconnected procedures.
Patient experience changes. The patient who entered the pathway with a 12 mm shadow on a screening CT does not spend three months in surveillance, waiting, surveillance, waiting, MDT, waiting, surgery. They have a clear pathway, structured around their specific nodule, with the next decision visible at each step. For the patient whose ROSE assessment does not confirm cancer, the integrated pathway delivers a definitive answer in one anaesthetic with no surgical incisions and same-day discharge. For the patient whose ROSE assessment does confirm cancer, the operation that follows is the operation that was planned, with the parenchyma-sparing target identified by the 3D reconstruction and the lesion marked by the dye injection. This is detailed in the combined biopsy and robotic surgery page.
Integration is the discriminator. Point solutions accumulate. Pathways perform.
| Phase | UK availability in 2026 | Where the pathway is delivered |
|---|---|---|
| 1. AI risk stratification of nodules | Deployed in select trusts under the NHS England pilot; private radiology adoption growing | GSTT (among lead pilot sites); select UK lung nodule MDTs |
| 2. AI 3D anatomical reconstruction | Deployed by selected operators in selected cases across UK thoracic centres | Used by Mr Okiror in selected complex segmentectomy cases at GSTT and London Bridge Hospital |
| 3. ION robotic bronchoscopy | Established at high-volume centres; GSTT 635 procedures in 2024–25 | NHS: GSTT. Private: London Bridge Hospital (first centre in Europe outside trials) |
| 4. Robotic anatomic resection | Widely deployed and accelerating; GSTT 71.3% of anatomic resections robotic in 2024–25 | NHS: GSTT and most major UK thoracic centres. Private: London Bridge Hospital and The Lister Hospital Chelsea (both hold the da Vinci Xi platform) |
| 5. AR overlay onto operative field | Early prototype only — not clinically deployed | Industry demonstration and early evaluation by selected surgeons |
| Combined same-anaesthetic pathway | Phases 1–4 integrated under one general anaesthetic for selected fit patients with small peripheral lesions | Privately at London Bridge Hospital — the only UK facility holding all five required technologies in one theatre session (ION, cone beam CT, onsite ROSE, dye marking, da Vinci Xi) |
NHS access to the components of the AI-augmented pathway depends on local trust capacity, MDT design, and screening programme coverage. GSTT, as one of the highest-volume UK thoracic centres, delivers the operative phases of the pathway at scale; the AI risk stratification component is in pilot under the NHS England programme. NHS waiting times from screen-detected nodule through diagnostic work-up to surgery are reported in the NLCA 2026 to exceed the 49-day target in 88% of early-stage cancer cases.
Private access at London Bridge Hospital compresses the pathway substantially. Consultations are typically available within 2–3 working days. Where the patient and the lesion meet the suitability criteria, the combined biopsy-to-resection pathway can be delivered in a single anaesthetic, with same-day discharge if the ROSE preliminary assessment does not confirm cancer. The same consultant, the same MDT discussion, the same operative platform, and the same post-procedure follow-up apply across NHS and private settings. Most major private medical insurers are recognised; self-pay is straightforward to arrange.
For consultations and selected surgical cases not requiring ION bronchoscopy or multimodality planning, The Lister Hospital Chelsea is Mr Okiror's second private operating base. Both hospitals hold the da Vinci Xi robotic platform. Outpatient consultations are also available at HCA outpatients in Canary Wharf and the City of London.
The clinical case for the AI-augmented pathway is technical, but its consequence is experienced by the patient in straightforward terms. The patient with a shadow on a CT scan, told they need a biopsy and then probably surgery, has historically faced a journey measured in months and consisting of multiple separate hospital attendances. The integrated pathway, where suitability allows, replaces this with a journey measured in weeks or days, often consisting of one principal hospital admission.
In its most compressed expression — the combined same-anaesthetic pathway at London Bridge Hospital — the patient is admitted on the day of the procedure. Under one general anaesthetic, the lesion is reached and biopsied via robotic bronchoscopy; the biopsy is preliminarily assessed by an onsite cytopathologist; and either resection proceeds immediately under the same anaesthetic, or the procedure ends at bronchoscopy with same-day discharge and no surgical incisions. The patient wakes either with an answer that did not require surgery, or with the operation already complete and recovery beginning. The cognitive and emotional load of months in surveillance, waiting for the next appointment, waiting for the next result, is removed.
Not every patient is suitable for the most compressed expression of the pathway. The combined-anaesthetic pathway is offered to a carefully selected group: fit patients, with small peripheral lesions, without complex staging requirements, with imaging characteristics consistent with a single curative operation. The full account of suitability is set out on the combined biopsy and robotic surgery page. For patients who are not suitable for the combined pathway, the individual phases — ION bronchoscopy, robotic anatomic resection — remain available as discrete steps with the same operator and the same MDT supporting the journey, with the integration applying where it can.
A practical detail of the patient experience: the 3D reconstruction, when used, is shared with the patient at the pre-operative consultation. Patients see their own lung, their own tumour, and the segments that will be removed and preserved. This materially changes how an operation is understood — from an abstract phrase ("segmentectomy of the left upper lobe lingular segment") to a concrete image of one's own anatomy. The consultations that include 3D reconstruction routinely take longer and produce more questions, both of which are good outcomes. The integration is intellectual as well as procedural.
The data on workforce capacity is now unambiguous. The Society for Cardiothoracic Surgery and Specialty Advisory Committee Workforce Report 2025 documents a UK thoracic surgical consultant headcount of 153, against a modelled demand of 39–77 additional consultants by 2030 to meet projected surgical volume from the maturing screening programme [10]. The current rate of UK thoracic surgical certification — approximately 9 to 10 Certificates of Completion of Training per year — is well below the rate that would be required to close the gap on the timescale that screening rollout is generating it. The structural mismatch between detection capacity and surgical capacity is becoming visible in the audit data: the NLCA 2026 figure of 88% breach of the 49-day target reflects, in part, capacity-limited operating lists rather than diagnostic delay.
The pathway-level analysis from LungIMPACT (Woznitza and colleagues, Nature Medicine, March 2026) reaches a related conclusion from a different angle: that AI accelerates the diagnostic end of the pathway without changing the structural delays elsewhere in the journey [11]. AI risk stratification reduces the time spent in surveillance for higher-risk nodules; ION robotic bronchoscopy reduces the time and risk of biopsy; pathway integration reduces the administrative latency between phases. The accumulating consequence is that more patients reach the operating theatre, faster — and the operating theatre's capacity is the binding constraint that everything upstream pushes against.
The implication is not that the AI-augmented pathway is wrong to pursue, but that its full clinical benefit requires the workforce and theatre capacity downstream to match. This is the policy conversation that the SCTS Bulletin piece on workforce in the screening era addresses [12] — not separately discussed on this page, which is principally clinical, but acknowledged as the system-level frame within which the pathway operates. The bottleneck that the integration moves toward is the one the system must now solve.
For the UK to realise the full clinical benefit of the AI-augmented pathway, three things will need to move in parallel: continued AI deployment across NHS centres beyond the lead pilot sites; theatre and workforce expansion in thoracic surgery, beyond the current rate of CCT completion; and integration discipline in the design of cancer pathways at trust and regional levels, so that AI's incremental gains at each phase are not lost to administrative latency between them. The technology question is broadly answered. The pathway and capacity questions are the remaining work.
Across the UK lung cancer surgical pathway in 2026, artificial intelligence is now active at five sequential decision points: AI risk stratification of indeterminate nodules at the front end of the pathway; AI 3D anatomical reconstruction for surgical planning, particularly in segmentectomy; AI-planned robotic bronchoscopy for biopsy and dye-marking of peripheral lesions; robotic anatomic resection on the da Vinci Xi platform; and an emerging fifth phase — intraoperative AR overlay of patient-specific anatomy onto the live surgical field — that exists today in prototype only.
Each phase, in isolation, is an incremental gain. The clinical change in patient outcomes does not come from any single one. It comes from sequencing them across the pathway, end to end, so that what was a three-month, three-decision journey becomes one continuous patient pathway — in its most compressed expression, the combined same-anaesthetic biopsy-to-resection pathway delivered privately at London Bridge Hospital. The screening rollout has changed who reaches the thoracic surgeon and at what stage. The AI-augmented pathway is the operational response to that change.
The remaining UK constraint is no longer detection or technology. It is workforce and theatre capacity, against a modelled demand growing faster than the rate at which new thoracic surgical consultants are being certified. The pathway question and the capacity question are the work that remains.
For private patients with a shadow on a CT scan, indeterminate nodule, or suspected early-stage lung cancer who would benefit from a structured assessment within the integrated pathway, Mr Okiror offers consultations at London Bridge Hospital and The Lister Hospital Chelsea within 2–3 working days. Self-referrals welcome.
What does it mean that AI is now part of the lung cancer surgical pathway in 2026?
Artificial intelligence is no longer a single intervention in lung cancer care. In 2026 it sits at five sequential decision points across the pathway: risk-stratifying which indeterminate nodules on a screening CT are likely to be cancer; reconstructing the patient's own CT into a 3D anatomical model used for surgical planning; planning the airway navigation route for robotic bronchoscopy to reach and biopsy peripheral lesions; guiding robotic anatomic resection; and, in early prototype, overlaying that reconstructed anatomy onto the live surgical field during the operation. The clinical change comes from the integration across these five points, not from any single tool. In UK practice in 2026, the first four are deployed; the fifth exists in prototype.
Is the NHS using AI to interpret lung CT scans for cancer?
Yes, in selected centres. The NHS Targeted Lung Health Check (lung cancer screening) programme has been rolled out across England with five-year outcome data published in Nature Medicine in March 2026 (Lee et al.), showing 76% of screen-detected cancers diagnosed at Stage I or II. In January 2026, NHS England announced a national pilot integrating AI-based risk stratification of indeterminate nodules with low-dose CT screening, alongside robotic bronchoscopy access for nodules requiring biopsy. Guy's and St Thomas' is one of the lead participating centres. AI nodule risk stratification is not yet universal NHS practice but is moving from select-site pilot toward broader adoption.
What is AI risk stratification of lung nodules and how does it compare to existing scoring systems?
AI risk stratification analyses a CT scan of an indeterminate lung nodule and produces a quantitative probability that the nodule is malignant. It complements but does not replace existing scoring systems such as the Brock model and the Mayo Clinic model, which use clinical variables to estimate malignancy probability. AI tools add image-derived features that are difficult or impossible to capture by human reading alone, particularly in small nodules under 8 mm where size on its own is uninformative. The clinical decision — surveillance, biopsy, or resection — remains a multidisciplinary one made at MDT, informed by the AI score alongside clinical context and imaging review. Two algorithms are now FDA-cleared and/or MDR-certified for use in this clinical step; this page is descriptive of the category and does not name vendors.
How does AI plan navigation through the lung's airways for biopsy?
Modern robotic bronchoscopy platforms reconstruct a three-dimensional airway tree directly from the patient's own pre-operative CT scan. The reconstruction maps the branching pattern of the bronchi out to the lung periphery and identifies the safest route to a peripheral nodule. During the procedure the operator drives a slim shape-sensing catheter along this planned path while the platform displays the position relative to the airway in real time. Cone beam CT confirms positional accuracy at the lesion before biopsy. This allows reliable biopsy of small peripheral nodules in the outer lung that conventional flexible bronchoscopy cannot reach. The ION platform (Intuitive Surgical) is the principal system in UK practice; GSTT delivered 635 ION procedures in 2024–25.
What is 3D anatomical reconstruction in lung cancer surgery and when is it used?
3D anatomical reconstruction takes the patient's own CT scan and generates a patient-specific virtual model of the lung — airways, pulmonary arteries, pulmonary veins, and the tumour itself — that can be rotated and inspected before the operation. It is particularly useful in planning anatomic segmentectomy, where the operative plane runs between two adjacent lung segments along their individual vascular and airway anatomy. The reconstruction allows the surgeon to identify and confirm the relevant segmental vessels and bronchi, to estimate the safe margin around the tumour, and to anticipate variant anatomy before encountering it intraoperatively. Several manufacturers now offer 3D reconstruction platforms for thoracic surgery. Mr Okiror uses 3D reconstruction in selected cases — particularly complex segmentectomy planning where the patient's segmental anatomy is unusual or where the tumour sits close to a segmental boundary.
Can I have my biopsy and lung cancer surgery under the same anaesthetic?
For carefully selected patients with a small peripheral lung lesion, yes. Under a single general anaesthetic, ION robotic bronchoscopy is used to reach and biopsy the lesion, with cone beam CT positional confirmation and onsite rapid cytopathology (ROSE) providing a preliminary tissue assessment within minutes. If the ROSE assessment signals non-small cell lung cancer, dye marking and robotic anatomic resection follow immediately. If it does not, the patient wakes with no surgical incisions and is discharged the same day. This combined same-anaesthetic pathway is offered privately at London Bridge Hospital. It is not appropriate for all patients; suitability is assessed individually at consultation.
What is AR overlay of anatomy onto the surgical field — is it available in the UK?
Augmented reality overlay refers to real-time projection of the patient's reconstructed anatomy — airways, vessels, and tumour — onto the surgeon's view of the live operative field. The principle is the same as image-guided navigation that has transformed orthopaedic and neurosurgical practice over the past two decades. Several manufacturers have prototype platforms in development. The technology is not yet clinically deployed in the UK or in routine practice anywhere; it exists at the stage of industry demonstration and early evaluation. Mr Okiror has seen early prototypes demonstrated at an industry event in London. The reasonable expectation is that AR overlay enters limited clinical evaluation over the next several years before broader adoption.
Which UK hospitals currently offer the integrated AI-augmented lung cancer pathway?
The components of the AI-augmented pathway are at different stages of UK availability. AI risk stratification of nodules is deployed in select trusts including GSTT under the NHS England pilot. 3D anatomical reconstruction is used by selected operators across UK thoracic centres in selected cases. ION robotic bronchoscopy is established at high-volume centres including GSTT (635 procedures in 2024–25). Robotic anatomic resection is widely available across UK thoracic surgery, with GSTT performing 71.3% of anatomic resections robotically in 2024–25. The combined same-anaesthetic biopsy-to-resection pathway is offered privately at London Bridge Hospital. AR overlay is in early prototype only and not deployed.
How does private access to this pathway differ from NHS access?
NHS access to the components of the AI-augmented pathway depends on local trust capacity, MDT pathway design, and screening programme coverage. NHS waiting times from screen-detected nodule through diagnostic work-up to surgery are reported in the NLCA 2026 to exceed the 49-day target in 88% of early-stage cancer cases. Private access at London Bridge Hospital compresses this. Consultations are typically available within 2–3 working days. Where appropriate, the combined biopsy-to-resection pathway can be delivered in a single anaesthetic, with same-day discharge if the ROSE preliminary assessment does not confirm cancer. The same consultant, the same MDT discussion, and the same post-procedure follow-up apply across both settings.
Is Mr Okiror financially associated with any of these AI companies?
Mr Okiror has no employment, consulting, or equity relationships with the AI software companies or surgical platform manufacturers whose products are described on this page, beyond those declared in his published peer-reviewed research and on the relevant pages of this site (specifically: speaking and consulting fees from Pulmonx Corporation, related to endobronchial valve therapy and disclosed in full on the emphysema surgery page). He has attended industry-organised demonstration and educational events as a practising consultant evaluating emerging technology. This page is a descriptive account of the category, not endorsement of any specific product. Treatment recommendations are made by the multidisciplinary team based on clinical evidence and patient phenotype, not by individual clinician preference.
Mr Okiror sees private patients within 2–3 working days at London Bridge Hospital and The Lister Hospital Chelsea. NHS referrals through Guy's and St Thomas'. Second opinion service available for patients who have been told elsewhere that no further options exist.
Request a consultation →Disclosures and publication practice
Mr Okiror has no employment, consulting, or equity relationships with the artificial intelligence software companies or surgical platform manufacturers described on this page, beyond those declared in his published peer-reviewed research and on the relevant pages of this site (specifically: speaking and consulting fees from Pulmonx Corporation, related to endobronchial valve therapy and disclosed in full on the emphysema surgery page). Mr Okiror has attended industry-organised demonstration and educational events as a practising consultant evaluating emerging technology. This page is a descriptive account of the category, not an endorsement of any specific product. Treatment recommendations are made by the multidisciplinary team based on patient phenotype and the published evidence base, not by individual clinician preference. The information is provided in line with GMC Good Medical Practice and ABPI Code of Practice standards on transparency of industry relationships.
Publication practice. Where Mr Okiror's own research contributes to the evidence base summarised on this page, papers are deposited as preprints on medRxiv simultaneously with journal submission, supporting transparent and timely scrutiny and ensuring the work is openly accessible from the point of completion.
Companion clinical flagship — the Stages I–IIIA pathway and chemoimmunotherapy era
Combined Biopsy & Robotic SurgeryThe integrated same-anaesthetic pathway in clinical detail — selection, technique, outcomes
ION Robotic BronchoscopyPhase 3 in detail — navigational bronchoscopy for peripheral nodule biopsy
Robotic SegmentectomyPhase 4 in detail — lung-sparing anatomic resection, evidence, technique, outcomes
Lung Nodule SurgeryFront-end of the pathway — assessment, biopsy and surgical treatment for suspicious nodules
Shadow on a Lung ScanPatient orientation — what an indeterminate shadow on CT means and what happens next
For GPs & PhysiciansReferral information, pathways, and the 24-hour acknowledgement standard
Second Opinion ServiceFor patients told elsewhere that no further options are possible