Image-guided blood vessel preparation

Abstract

The present disclosure provides systems and methods for AI models that are trained layers to identify characteristics associated with a lesion in a blood vessel and provide suggested procedures associated with the lesion. The AI model may receive as input intravascular and / or extraluminal image data associated with the blood vessel and provide as output characteristics associated with the blood vessel, such as characterization. The characterization may include, for example, form or tissue properties (e.g., microcalcification, punctiform, fragment-like, sheet-like, nodule-like, or others), location within the vessel (e.g., inner layer, superficial layer, middle layer, depth, etc.), size (e.g., length, depth, thickness, etc.), severity, and / or the like. The characterization may be used as input to the same and / or different models to provide a suggested procedure associated with a lesion. The proposed procedures may include, for example, orbital atherectomy, endovascular lithotripsy, rotary atherectomy, cutting and / or scoring balloons, and / or the like.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 667,166, entitled “Image Guided Vessel Preparation,” filed July 3, 2024, the entire disclosure of which is hereby incorporated by reference herein. TECHNICAL FIELD

[0002] The technology disclosed herein relates to using intravascular imaging to provide vessel condition data for determining vessel preparation and treatment strategies. BACKGROUND

[0003] Percutaneous coronary intervention (PCI) planning uses visualization techniques (e.g., intravascular ultrasound (IVUS), intravascular optical coherence tomography (OCT), micro-OCT, and near-infrared spectroscopy (NIRS)). OCT is an intravascular imaging modality that provides high-definition cross-sectional images and three-dimensional images of the microscopic structure of a blood vessel using near-infrared light during PCI. An OCT catheter emits near-infrared light into a blood vessel to produce high-resolution real-time images of the blood vessel. To perform an imaging scan, referred to as an OCT pullback, the OCT catheter is inserted into the blood vessel and an infrared laser is used to scan the blood vessel wall in a spiral or spiral-like manner (e.g., rotated about an axis) while moving longitudinally along the axis. OCT components include an OCT designed catheter and an imaging system for receiving image data captured by the OCT probe, processing the image data, and providing images for display.

[0004] Patients with insufficient stent expansion are at high risk for adverse outcomes (e.g., in-stent thrombosis and in-stent restenosis). Several publications have proposed OCT-based quantitative scoring systems to predict future stent expansion. See Fujino, A., et al., A New Optical Coherence Tomography-based Calcium Scoring System to Predict Stent Under-expansion, EuroIntervention, 13, e2182-e2189, (2018). Machine learning models and development have also been proposed to identify blood vessels at risk for under-expansion, which would be candidates for atherectomy lesion preparation. See Gharaibeh, et al., Prediction of Stent Under-expansion in Calcified Coronary Arteries Using Machine Learning on Intravascular Optical Coherence Tomography Images, Sci Rep, 13, 18110, (2023). https: / / doi.org / 10.1038 / s41598-023-44610-9 .

[0005] Artificial intelligence (AI) models can be trained to identify calcification of blood vessels. Lesion calcification can be characterized based on the location of the vascular bed or in the body, including, for example, coronary vascular beds, peripheral vascular beds, carotid vascular beds, renal vascular beds, and other vascular beds. Generally, the type of treatment regimen used for a particular calcified lesion includes orbital atherectomy, rotational atherectomy, laser, and intravascular lithotripsy. Each treatment method is generally selected based on the characteristics of the blood vessel.

[0006] Medial calcification is one of two main categories of vascular calcification. Medial calcification is associated with atherosclerotic plaques, which are thought to be caused by altered accumulation of lipids, pro-inflammatory cytokines, and apoptosis within the plaque, which induce osteoblast differentiation. The result of atherosclerosis is often luminal stenosis. Medial calcification associated with atherosclerosis is characterized by calcified plaque-like or spotty areas on the luminal side of the internal elastic lamina of the blood vessel, and often affects the coronary arteries, carotid arteries, and larger arteries. Advanced forms of medial calcification can lead to vascular obstruction or occlusion.

[0007] Medial calcification is considered the second of two major categories of vascular calcification. It tends to be more widespread in the lower abdominal region and extremities and is caused by osteogenic differentiation of vascular smooth muscle cells within the medial layer of the vessel wall. Calcium accumulation begins as amorphous mineral deposits and undergoes progressive remodeling, potentially progressing to mineralization into mature bone. Medial calcification is more common in small to medium-sized arteries and can lead to increased arterial stiffness. Although medial calcification is generally not associated with luminal obstruction, decreased elasticity and compliance of the arterial wall can ultimately lead to atherosclerosis, reduced perfusion, and eventually peripheral artery disease (PAD), and in later forms, left ventricular hypertrophy and / or heart failure. Medial calcification is commonly associated with diabetes, chronic kidney disease, and metabolic dysfunction, and is characterized as described above, accompanied by elastin layer rupture and calcification. This type of medial calcification can appear as smooth, continuous calcified areas, often appearing as parallel lines or “railway” tracks on X-rays.

[0008] The presence of vascular calcification (whether in the inner or middle layer) can hinder optimal outcomes following conventional interventional therapy. Inner layer calcification can obstruct or block the vascular lumen, making it difficult to pass through this portion of the vessel, and passing through such an obstructed or blocked lesion can lead to overstretching injury, which is prone to restenosis. Both inner and middle layer calcification can also impair the ability to achieve full device dilation to restore flow and contribute to higher vascular backflow, both of which result in higher residual stenosis. Vascular calcification also hinders drug diffusion, further increasing the likelihood of restenosis.

[0009] In areas of blood vessels with severe calcification, stents may be difficult to expand fully. Once implanted in tissue that is highly resistant to expansion, there is no opportunity to apply pre-stent treatment options or procedures (e.g., laser ablation, rotational plaque resection, scoring balloon, angioplasty balloon, or orbital plaque resection). After stent implantation, the only option is risky high-pressure (e.g., up to 30 atm) balloon expansion, which carries the risk of vessel rupture.

[0010] Specialized balloon-based techniques include cutting balloons and scoring balloons, which are commonly used for eccentric calcification. Cutting balloons consist of multiple microblades mounted on the balloon, while scoring balloons consist of a semi-compliant balloon with several nitinol wires wound around it. Both create incisions in the calcification and improve the compliance of the vessel, allowing for dilation. Their design allows them to grip the calcification, resulting in less slippage, also known as the "melonseeding effect," which avoids dissection of adjacent vessels. However, in the presence of severe calcification, cutting balloons have been found to have lower surgical success rates than rotational plaque resection (RA), although they are indeed practical as an adjunct to RA. Very high-pressure balloons consist of a bilayered non-compliant balloon with a rated rupture pressure of approximately 30–35 atm. This technique has its place as an adjunct to other techniques, such as, in conjunction with adjunctive RA.

[0011] Devices used for intravascular lithotripsy (IVL) consist of a balloon-based delivery system containing numerous emitters that generate short, high-energy electrical sparks. Although the balloon itself only inflates to 4 atm, each brief pulse delivers a localized spike of pressure outside the balloon. To date, IVL has been more effective in concentric calcified lesions. However, IVL can still be effective in eccentric calcifications, but when the calcified lesion is only eccentric, it requires more pulses.

[0012] IVL devices may include electrodes or lithotripsy emitters that generate sound waves through an electric arc discharge between electrode components, but may also include devices that generate acoustic energy within a balloon via a laser energy source. Examples of such laser systems are described in U.S. Patent Nos. 11,058,492 and 11,246,569 (the entire contents of which are incorporated herein by reference). Examples of inductive systems are described in U.S. Patent Nos. 8,728,091, 9,642,673, and 10,850,078, and published U.S. Patent Application No. 2022-0054194 (the entire contents of which are incorporated herein by reference).

[0013] Rotational plaque excision (RA) uses a diamond-tipped file that rotates at very high speeds (140,000–160,000 rpm), resulting in differential ablation of calcified lesions. RA was previously used for aggressive debulking of calcium, which led to numerous complications, including non-return and vascular perforation (due to embolism of particulate matter). However, modifications to RA techniques (e.g., shorter RA stroke, use of pecking motion at the lesion site, smaller burr size, and combinations of auxiliary and complementary techniques) have significantly improved its use.

[0014] Orbital plaque resection (OA) involves an off-center-mounted diamond-coated crown that uses centrifugal force to run along a orbit (at 80,000 or 120,000 rpm), resulting in preferential calcification abrasion while bending away from elastic healthy tissue. With OA, it is hoped that diameter stenosis can be reduced to 50% in >98% of lesions. Because OA achieves a wider rotational orbit and deeper calcification modification, it is preferred over resection of radiculopathy (RA) in larger vessels with concentric or nodular calcification.

[0015] Greater specificity and techniques using OCT are described in U.S. Patent Application 2020 / 0294659A1, published September 17, 2020; U.S. Patent Application 2021 / 0042927A1, published February 11, 2021; and U.S. Patent Application 2023 / 0054891A1, published February 23, 2023, the entire contents of each of these publications are incorporated herein by reference. Summary of the Invention

[0016] This disclosure generally relates to one or more AI models trained to identify characteristics associated with lesions within blood vessels and to provide suggested surgical procedures associated with those lesions. The AI ​​models may receive intravascular and / or extravascular image data associated with the blood vessel as input and provide characteristics associated with the blood vessel as output. Characterizations may include, for example, morphology or tissue properties, location within the blood vessel, size, severity, and / or similar features. These characteristics may be used as input to the same and / or different models to provide suggested surgical procedures associated with the lesion. Suggested surgical procedures may include, for example, orbital plaque resection, intravascular lithotripsy, rotational plaque resection, cutting balloon and / or scarifying balloon, and / or similar procedures. Treatment of coronary artery disease (CAD) and peripheral artery disease (PAD) requires proficiency in all types of modification techniques, an understanding of which procedures and / or tools are best suited for a given patient, and familiarity with intracoronary imaging used to guide surgery. An increasing combination of lesion modification techniques is being used in clinical practice. Therefore, the systems and methods described herein can provide efficient and patient-specific suggested surgical procedures in a rapid and accurate manner.

[0017] One aspect of this disclosure relates to a system comprising one or more processors for receiving vascular data associated with at least a portion of a blood vessel from an imaging modality, for determining initial characteristics of a lesion within at least a portion of the blood vessel based on the vascular data, for providing the determined initial characteristics of the lesion as input to a model, for identifying a proposed surgical procedure by executing the model for modifying the lesion within at least a portion of the blood vessel, and for providing the proposed surgical procedure as output.

[0018] After the recommended surgery is completed, one or more processors in the system can further receive postoperative vascular data from the imaging modality and determine the updated characteristics of the lesion based on the postoperative vascular data.

[0019] One or more processors in the system can further store the initial features, suggested procedures, and updated features as training data for the model in memory, and update the model based on the initial features, suggested procedures, and updated features.

[0020] In some examples, the imaging modality could be optical coherence tomography or intravascular ultrasound.

[0021] In some examples, the characteristics associated with a lesion may include at least one of the lesion’s total angle, thickness, or length.

[0022] In some examples, the recommended procedures could be orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

[0023] In some examples, the lesion can be a calcified lesion.

[0024] One or more processors in the system can further determine, based on vascular data, whether the lesion includes a medial or medial lesion, and based on this determination, provide an indication of whether the lesion includes a medial or medial lesion for output.

[0025] Another aspect of this disclosure relates to a method comprising: receiving vascular data associated with at least a portion of a blood vessel from an imaging modality by one or more processors; determining, based on the vascular data, initial characteristics of a lesion within at least a portion of the blood vessel by one or more processors; providing the determined initial characteristics of the lesion as input to a model by one or more processors; identifying, by executing the model, a proposed surgical procedure for modifying the lesion within at least a portion of the blood vessel by one or more processors; and providing the proposed surgical procedure as output by one or more processors.

[0026] The method may further include: after the recommended surgery is completed, receiving postoperative vascular data from the imaging modality by one or more processors, and determining updated characteristics of the lesion based on the postoperative vascular data by one or more processors.

[0027] The method may further include: storing initial features, suggested procedures, and updated features as training data for the model in memory by one or more processors, and updating the model by one or more processors based on the initial features, suggested procedures, and updated features.

[0028] In some examples, the imaging modality could be optical coherence tomography or intravascular ultrasound.

[0029] In some examples, the characteristics associated with a lesion may include at least one of the lesion’s total angle, thickness, or length.

[0030] In some cases, the recommended procedures could be orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

[0031] In some examples, the lesion can be a calcified lesion.

[0032] The method may further include: one or more processors determining, based on vascular data, whether a lesion comprises a medial or medial lesion, and one or more processors providing an indication, based on the determination, whether a lesion comprises a medial or medial lesion for output.

[0033] Another aspect of this disclosure relates to one or more non-transitory computer-readable storage media encoding instructions that, when executed by one or more processors, cause one or more processors to perform operations including: receiving vascular data associated with at least a portion of a blood vessel from an imaging modality; determining initial characteristics of a lesion within at least a portion of the blood vessel based on the vascular data; providing the determined initial characteristics of the lesion as input to a model; identifying a proposed surgical procedure by executing the model for modifying the lesion within at least a portion of the blood vessel; and providing the proposed surgical procedure as an output.

[0034] The one or more non-transitory computer-readable storage media may further include: receiving postoperative vascular data from an imaging modality by one or more processors after the proposed surgery has been completed, and determining updated characteristics of the lesion based on the postoperative vascular data by one or more processors.

[0035] The one or more non-transitory computer-readable storage media may further include: storing initial features, suggested procedures, and updated features as training data for the model in memory by one or more processors, and updating the model by one or more processors based on the initial features, suggested procedures, and updated features.

[0036] In some examples, the imaging modality could be optical coherence tomography or intravascular ultrasound.

[0037] In some examples, the characteristics associated with a lesion may include at least one of the lesion’s total angle, thickness, or length.

[0038] In some examples, the recommended procedures could be orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

[0039] In some examples, the lesion can be a calcified lesion.

[0040] The one or more non-transitory computer-readable storage media may further include: determining, based on vascular data, whether a lesion comprises a medial or medial lesion by one or more processors, and providing, based on the determination, an indication of whether a lesion comprises a medial or medial lesion for output.

[0041] Another aspect of this disclosure relates to a system comprising one or more processors that receive intravascular imaging data of a blood vessel from an imaging modality, determine information associated with the blood vessel based on the intravascular imaging data, determine the presence of an intravascular lesion based on the intravascular imaging data, determine the characteristics of the lesion based on the intravascular imaging data, provide the characteristics of the lesion as input to a model, wherein the model is trained to identify a proposed surgery based at least in part on the information associated with the blood vessel and the determined characteristics of the lesion, identify a proposed surgery for modifying a lesion within at least a portion of the blood vessel by executing the model, provide a proposed surgery as output, store the information associated with the blood vessel, the determined characteristics of the lesion, and the proposed surgery in a memory from a database, and update the model based on the database.

[0042] One or more processors in the system may further: receive postoperative vascular data from the imaging modality after the recommended surgery has been completed, and determine the updated characteristics of the lesion based on the postoperative vascular data.

[0043] In some examples, the imaging modality could be optical coherence tomography or intravascular ultrasound.

[0044] In some examples, the characteristics associated with a lesion may include at least one of the lesion’s total angle, thickness, or length.

[0045] In some examples, the recommended procedures could be orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

[0046] In some examples, the lesion can be a calcified lesion.

[0047] One or more processors in the system can further determine, based on vascular data, whether the lesion includes a medial or medial lesion, and based on this determination, provide an indication of whether the lesion includes a medial or medial lesion for output.

[0048] Another aspect of this disclosure relates to a method comprising: receiving intravascular imaging data of a blood vessel from an imaging modality by one or more processors; determining information associated with the blood vessel based on the intravascular imaging data by one or more processors; determining the presence of an intravascular lesion based on the intravascular imaging data by one or more processors; determining the characteristics of the lesion based on the intravascular imaging data by one or more processors; providing the characteristics of the lesion as input to a model by one or more processors, wherein the model is trained to identify a proposed surgery based at least in part on the information associated with the blood vessel and the determined characteristics of the lesion; identifying the proposed surgery by executing the model by one or more processors, the proposed surgery being used to modify the lesion within at least a portion of the blood vessel; providing the proposed surgery as output by one or more processors; storing the information associated with the blood vessel, the determined characteristics of the lesion, and the proposed surgery from a database in a memory by one or more processors; and updating the model based on the database by one or more processors.

[0049] The method may further include: after the recommended surgery is completed, receiving postoperative vascular data from the imaging modality by one or more processors, and determining updated characteristics of the lesion based on the postoperative vascular data by one or more processors.

[0050] In some examples, the imaging modality could be optical coherence tomography or intravascular ultrasound.

[0051] In some examples, the characteristics associated with a lesion may include at least one of the lesion’s total angle, thickness, or length.

[0052] In some cases, the recommended procedures could be orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

[0053] In some examples, the lesion can be a calcified lesion.

[0054] The method may further include: one or more processors determining, based on vascular data, whether a lesion comprises a medial or medial lesion, and one or more processors providing an indication, based on the determination, whether a lesion comprises a medial or medial lesion for output.

[0055] Another aspect of this disclosure relates to one or more non-transitory computer-readable storage media encoding instructions that, when executed by one or more processors, cause one or more processors to perform operations including: receiving intravascular imaging data of a blood vessel from an imaging modality; determining information associated with the blood vessel based on the intravascular imaging data; determining the presence of an intravascular lesion based on the intravascular imaging data; determining the characteristics of the lesion based on the intravascular imaging data; providing the characteristics of the lesion as input to a model, wherein the model is trained to identify a proposed surgery based at least in part on the information associated with the blood vessel and the determined characteristics of the lesion; identifying the proposed surgery by executing the model for modifying the lesion within at least a portion of the blood vessel; providing the proposed surgery as output; storing the information associated with the blood vessel, the determined characteristics of the lesion, and the proposed surgery from a database into a memory; and updating the model based on the database.

[0056] The one or more non-transitory computer-readable storage media may further include: receiving postoperative vascular data from an imaging modality by one or more processors after the proposed surgery has been completed, and determining updated characteristics of the lesion based on the postoperative vascular data by one or more processors.

[0057] In some examples, the imaging modality could be optical coherence tomography or intravascular ultrasound.

[0058] In some examples, the characteristics associated with a lesion may include at least one of the lesion’s total angle, thickness, or length.

[0059] In some examples, the recommended procedures could be orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

[0060] In some examples, the lesion can be a calcified lesion.

[0061] The one or more non-transitory computer-readable storage media may further include: determining, based on vascular data, whether a lesion comprises a medial or medial lesion by one or more processors, and providing, based on the determination, an indication of whether a lesion comprises a medial or medial lesion for output.

[0062] Another aspect of this disclosure relates to a method comprising: receiving vascular data associated with at least a portion of a blood vessel by one or more processors; determining, based on the vascular data, initial characteristics of a lesion within at least a portion of the blood vessel by one or more processors; providing the determined initial characteristics of the lesion as input to a model by one or more processors; identifying a proposed surgery by executing the model by one or more processors for modifying the lesion within at least a portion of the blood vessel; and providing the proposed surgery as output by one or more processors.

[0063] The method may further include: receiving postoperative vascular data by one or more processors after the recommended surgery is completed, and determining updated characteristics of the lesion based on the postoperative vascular data by one or more processors.

[0064] The method may further include: storing initial features, suggested surgeries, and updated features as training data of the model in memory by one or more processors, and updating the model based on the initial features, suggested surgeries, and updated features by one or more processors.

[0065] In some examples, the characteristics associated with a lesion may include at least one of the lesion’s total angle, thickness, or length.

[0066] In some examples, the recommended procedures could be orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

[0067] In some examples, the lesion can be a calcified lesion.

[0068] The method may further include: one or more processors determining, based on vascular data, whether a lesion comprises a medial or medial lesion, and one or more processors providing an indication, based on the determination, whether a lesion comprises a medial or medial lesion for output.

[0069] Another aspect of this disclosure relates to a method comprising: updating a model by one or more processors based on postoperative intravascular imaging data of a blood vessel and on a database. The model can be trained to identify proposed surgeries based at least in part on determined vascular characteristics and characteristics of determined lesions. Updating the model may include: receiving postoperative intravascular imaging data of a blood vessel by one or more processors; determining vascular characteristics based on the postoperative intravascular imaging data by one or more processors; determining the presence of intravascular lesions based on the postoperative intravascular imaging data by one or more processors; determining the characteristics of the lesions based on the postoperative intravascular imaging data by one or more processors; providing the characteristics of the lesions as input to the model by one or more processors; identifying proposed surgeries for modifying lesions within at least a portion of the blood vessel by one or more processors; providing the proposed surgeries as output by one or more processors; and storing the determined vascular characteristics, the characteristics of the determined lesions, and the proposed surgeries from the database into memory by one or more processors.

[0070] The method may further include: receiving postoperative vascular data by one or more processors after the recommended surgery is completed, and determining updated characteristics of the lesion based on the postoperative vascular data by one or more processors.

[0071] In some examples, the characteristics associated with a lesion may include at least one of the lesion’s total angle, thickness, or length.

[0072] In some examples, the recommended procedures could be orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

[0073] In some examples, the lesion can be a calcified lesion.

[0074] The method may further include: one or more processors determining, based on vascular data, whether a lesion comprises a medial or medial lesion, and one or more processors providing an indication, based on the determination, whether a lesion comprises a medial or medial lesion for output. Attached Figure Description

[0075] Figure 1 It is an exemplary data collection system for blood vessels according to various aspects of this disclosure.

[0076] Figure 2 These are exemplary treatment devices according to various aspects of this disclosure, which can be used with... Figure 1 The data collection system is used in conjunction with and / or based on data collection systems. Figure 1 The data collection system determines the recommended surgical procedure to be used.

[0077] Figure 3 Exemplary methods according to various aspects of this disclosure are used for... Figure 1 The data collection system receives vascular data to determine information associated with blood vessels and / or the characteristics of lesions.

[0078] Figure 4 and Figure 5A Exemplary methods according to various aspects of this disclosure use OCT as an imaging-based modality for training a model.

[0079] Figure 5B The exemplary flowcharts are based on various aspects of this disclosure, which are at least in part based on... Figure 1 The data collection system receives vascular data to train the model, generating suggested vascular preparation options.

[0080] Figure 6 This is an exemplary flowchart based on various aspects of this disclosure, which is used for at least part based on... Figure 1 The data collection system receives vascular data to create a database for vascular preparation surgery and output.

[0081] Figure 7 This is an exemplary flowchart based on various aspects of this disclosure, which is used for at least part based on... Figure 1 The data collection system receives vascular data to create a database for vascular preparation surgery and output, and executes models trained on the database.

[0082] Figure 8 The illustration shows a surgical procedure recommended according to various aspects of this disclosure, which can be performed by... Figure 1 The data is determined by the model trained on the database of the data collection system.

[0083] Figures 9 to 15 The illustration shows an exemplary interface screen according to various aspects of this disclosure, which, for example, can be... Figure 1 Implemented within the context of a data collection system.

[0084] Figure 16 A flowchart is depicted illustrating an exemplary process according to various aspects of this disclosure, which uses... Figure 1 The data collection system is used to provide recommended surgeries.

[0085] Figure 17 A flowchart is depicted for another exemplary process according to various aspects of this disclosure, which uses Figure 1 The data collection system is used to provide recommended surgeries. Detailed Implementation

[0086] This technology generally involves one or more AI models trained to identify characteristics associated with lesions within blood vessels and to provide recommended surgical procedures related to those lesions. The AI ​​models may receive intravascular and / or extravascular image data associated with the blood vessel as input and provide characteristics (e.g., characterizations) associated with the blood vessel as output. Characterizations may include, for example, form or tissue properties (e.g., microcalcifications, punctate, fragmented, sheet-like, nodular, or others), location within the blood vessel (e.g., inner, superficial, middle, deep, etc.), size (e.g., length, depth, thickness, etc.), severity, and / or similar factors. Characterizations may be used as input to the same and / or different models to provide recommended surgical procedures associated with the lesion. Recommended surgical procedures may include, for example, orbital plaque resection, intravascular lithotripsy, rotational plaque resection, cutting balloon and / or scarifying balloon, and / or similar procedures.

[0087] The world's population will increasingly require treatment for increasingly complex coronary artery disease (CAD) and peripheral artery disease (PAD). Treatment of CAD and PAD necessitates proficiency in all types of lesion modification techniques, an understanding of which procedures and / or tools are best suited for a given patient, and familiarity with intracoronary imaging used to guide surgery. Combinations of lesion modification techniques will be increasingly used in clinical practice and can again benefit from the methods and systems described herein. In particular, using vascular data to characterize lesions and subsequently using this characterization in determining the recommended procedure associated with the lesion provides a variety of workflows and options to facilitate the treatment of CAD and PAD, as well as other diseases.

[0088] By using vascular data captured targeting regions of interest within a vessel (e.g., intravascular and / or extraluminal image data), the system can dynamically determine information associated with the region of interest within the vessel (including lesion characterization), which can be used to identify recommended surgical procedures for treating the lesion. Recommended surgical procedures can be provided automatically upon receiving vascular data. Since less user input may be received by the system when identifying recommended surgical procedures, the system's computational efficiency increases with reduced processing and network overhead. Furthermore, when characterizing the vessel and / or a vessel with lesions, the determined characterization can be provided for output along with the recommended surgical procedure. This allows for efficient and almost entirely automatic fusion of information (including recommended surgical procedures) between image data and information determined based on vascular data. For example, information associated with a vessel may include information about: plaque burden, fractional flow reserve (“FFR”) measurements at one or more locations along the vessel, calcium angle, external elastic layer (“EEL”) detection, calcium detection, proximal frame, distal frame, EEL-based indices, stent / stent-free decision, lumen diameter, mean diameter, percentage of stenosis, lesion type, and / or similar factors. Greater specificity and techniques for imaging and evaluating blood vessels are described in U.S. Patent No. 11,819,309, issued November 21, 2023, the entire contents of which are incorporated herein by reference.

[0089] Additional data (e.g., patient history data, and / or instructions for performing subsequent vascular preparation procedures on the imaged blood vessel) can also be provided as input to the model. This additional data can be stored in a database to generate additional training data for the AI ​​model, which can then be used for further fine-tuning to improve model performance.

[0090] The display may include one or more representations that include information associated with blood vessels. For example, the information associated with blood vessels may be provided as a digital representation, classification or labeling, color-coded visualization, and / or a coaxial loop, etc. The representation may include image data, a two-dimensional representation generated based on the image data, a three-dimensional representation, and / or the like.

[0091] When determining the recommended surgery, information associated with blood vessels and characteristics associated with lesions can be used. Therefore, the information associated with blood vessels used to generate a representation of the vessel (which provides that representation for output) can also be used to determine the characteristics associated with the lesion and to identify the recommended surgery. By using vascular data for multiple purposes (e.g., determining information associated with blood vessels, determining the characteristics of the lesion, and / or determining the recommended surgery associated with the lesion), the efficiency of the system is improved because multiple surgeries no longer need to be performed to capture vascular data for each specific purpose.

[0092] Based on some examples, recommended procedures may include lesion modification and / or therapeutic procedures. For example, an AI model may be trained to identify recommended modification and / or therapeutic procedures. Modifying treatments may be, for example, plaque excision, IVL, various forms of ablation, and / or balloons, etc. In yet another example, recommended procedures may be recommended implantation devices (e.g., type of stent, stent size, and / or the like), and / or recommended landing sites (e.g., recommended stent location, and / or the like). In some examples, recommended procedures may include a recommendation not to perform any procedure at all. As another example, recommended procedures may include postponing treatment and / or vascular preparation procedures to a later time.

[0093] The recommended procedure, equipment, and / or landing area (collectively referred to herein as "recommended procedure") are determined based on data information identified from the received vascular data. For example, the system receives vascular data (e.g., imaging data). Imaging data can be intravascular imaging data and / or extraluminal imaging data. In some examples, the recommended procedure may be further determined based on information stored in a database. This database is built over time from previous procedures, including pre-treatment data, patient history data, specific details of the selected and performed procedures, and / or data related to the determined treatment outcomes.

[0094] Recommended surgical procedures can be identified based on characteristics associated with the lesion. For example, treatments have been developed specifically for the expansion, fragmentation, and / or removal of plaques of particular type, size, and / or location. Such plaques include, for example, lipids, fibrin, cholesterol, and / or calcification. Plaques can present in many different forms around the vessel wall and / or within different layers of the vessel wall. Accordingly, in order to identify recommended surgical procedures associated with lesions within the vessel, the system determines the characteristics associated with the lesion (e.g., type, shape, and / or location (along the vessel wall and / or within the vessel layers), etc.) and uses the determined lesion characteristics when identifying recommended surgical procedures.

[0095] According to some examples, characteristics associated with a lesion (e.g., the characterization of the lesion) can be determined based on histopathological examination and / or microscopic examination of the lesion tissue. For example, the system can receive information from histopathological examination and / or microscopic examination of the tissue. This information may include, for example, form or tissue nature (e.g., microcalcifications, punctate, fragmented, sheet-like, nodular, and / or other forms of calcification). In some examples, the system can be configured to perform histopathological examination and / or microscopic examination of the tissue. Histopathological examination or microscopic examination of the lesion tissue can be performed on the coronary arteries of a cadaver and / or the coronary arteries and / or peripheral arteries of a living organism.

[0096] In some examples, characteristics associated with a lesion may include its location (e.g., inner layer (closest to the vessel lumen), superficial, middle, and / or deep). In some examples, the AI ​​model generates output that at least partially characterizes the identified lesion and its location. The output may form part of the input to the AI ​​model for outputting a suggested surgery. In some examples, the AI ​​model is split into separate models, with at least one model for classifying the lesion based on the input image, and at least one other model for suggesting vascular preparation surgery based on the lesion classification output and potential other vascular data. Before providing the lesion classification output, the system may (e.g., via a user interface) receive user input to confirm or correct the classification. The output classification may include inner or middle layer calcification, including microcalcifications, punctate calcifications, fragmented calcifications, sheet-like calcifications, and / or nodular calcifications. The classification may include information about the lesion's location (e.g., inner layer, superficial, middle, deep, and more); size (e.g., length, depth, thickness, and more); and / or severity. If needed, the output classification can be updated before it is fed into the model (to train the model to suggest vascular preparation procedures). In other examples, the AI ​​model is trained end-to-end (e.g., based on vascular data) to output suggested vascular preparations. In those examples, the AI ​​model can output lesion classifications. Other characteristics can be determined based on any number of differences between one lesion and another.

[0097] Based on some examples, AI models (e.g., machine learning (ML) models, such as convolutional neural networks (CNNs)) can be used to determine characteristics associated with lesions. An AI model can be trained to receive vascular data as input and provide characteristics (e.g., representations) associated with the lesion as output. In some examples, the same AI model can be trained to determine information associated with blood vessels, in addition to characteristics associated with the lesion.

[0098] As an example, a model can be trained to identify one or more markers as detectable characteristics of a lesion. A lesion can be, for example, calcium within or along the vessel wall. The identified markers can provide indications about the location of the lesion, whether it has been altered or destroyed, or similar information. The model can be trained on temporal data (e.g., vascular data corresponding to the same vessel captured at different time points). For example, the model can receive vascular data to generate a proposed surgery. The proposed surgery can then be performed (e.g., autonomously or manually). The proposed surgery can be paused, and new vascular data can be acquired (e.g., from an OCT retraction of a vessel that has been partially prepared for surgery). In some examples, additional imaging (e.g., extraluminal imaging) can be performed during the surgery, forming part of new vascular data. The model can receive the new vascular data as additional input and output a proposed surgery. In this example, the proposed surgery could be the same surgery as a previously proposed surgery, a different surgery, or a suggestion to stop the surgery because no further preparation is needed. For example, in the case of dissection, if a drug-coated balloon (DCB) is the primary recommendation, a secondary recommendation could be a drug-eluting stent (DES) or a drug-eluting reabsorbable stent (DRS). For instance, two instances of vascular data, one prior to and one during preparation, could be stored in a database as described herein and provided as training examples.

[0099] Therefore, the temporal nature of the training data allows the model to identify potential features or connections between features that change over time. For example, training data spanning different time points in the surgery can reflect changes in current stenosis, changes in calcium rupture, and / or other changes caused by the recommended vascular surgery.

[0100] According to some examples, vascular data can be used when determining a recommended procedure. The recommended procedure may be one that should be performed by a user (e.g., a physician) prior to another procedure (e.g., stent implantation). In other examples, the recommended procedure may be performed automatically, for example, by a surgical robot or other device configured to perform the procedure. Vascular data may include image data captured by extraluminal imaging devices and / or intravascular imaging devices. Extraluminal imaging devices include, for example, computed tomography (CT) scanners, magnetic resonance imaging (MRI) scanners, ultrasound probes, fluoroscopy, and / or the like. Intravascular imaging devices may be, for example, optical coherence tomography (“OCT”) probes, intravascular ultrasound (“IVUS”) catheters, miniature OCT probes, near-infrared spectroscopy (NIRS) sensors, optical frequency domain imaging (“OFDI”) probes, and / or any other device that can be used to image blood vessels. In some examples, in addition to vascular data, the type of treatment to be recommended may be determined based on other data obtained from any particular lesion by means other than imaging. As an example, the withdrawal of an OCT probe can capture vascular data (e.g., intravascular images). Vascular data captured by OCT probes provides clear boundaries of calcium regions or calcium zones associated with the vessel wall structure.

[0101] The system can use these intravascular images (e.g., vascular data) to determine the characteristics of lesions (e.g., calcification, formation, and / or interstitial vascular data). These lesion characteristics can then be used when determining the type of vascular preparation surgery recommended for the desired treatment of such lesions within the vessel. For example, characteristics associated with the lesion can be provided as input to an AI model trained to recognize recommended surgeries (e.g., for dilation, rupture, and / or removal of lesions within the vessel). As described herein, recommended surgeries may include recommendations not to perform vascular preparation and / or recommendations to reassess the need for vascular preparation surgery at a later time. For example, another round of imaging and vascular preparation recommendations may be performed at a later time. The AI ​​model trained to recognize recommended surgeries can be the same model or a different model compared to the AI ​​model trained to recognize characteristics associated with vessels and / or lesions. For example, the AI ​​model may include a sub-model or module for generating OCT image data, which is fed as input to another sub-model or module to generate recommendations. In other examples, the AI ​​model processes input in the form of OCT images in an end-to-end manner to generate recommended outputs. The AI ​​model described in this article can represent multiple models trained to perform these and other tasks.

[0102] As discussed above and in this article, such images (whether taken alone or together with other data for a specific lesion) can be used to identify the recommended procedure (e.g., lesion modification therapy or technique) associated with the lesion before another treatment (e.g., stent deployment, balloon angioplasty, stent placement, and / or any other procedure to remove the vascular obstruction). By effectively and accurately characterizing the vessel and / or lesion and using the characteristics of the vessel and / or lesion to identify the recommended procedure, further modifications and / or procedures associated with the lesion can be mitigated. Furthermore, subsequent procedures (e.g., stent deployment) can be more successful due to the intervention of the recommended procedure associated with the lesion. For example, modifying coronary calcified lesions before stent placement can ensure better stent performance and prevent stent failure. If the calcified lesion is not modified first (e.g., by balloon dilation or other modification therapy), stent under-dilation modification may occur. Therefore, optimizing stent dilation performance and outcomes is crucial, and calcium modification prior to percutaneous coronary intervention (PCI) is an important step in ensuring adequate stent dilation during this procedure.

[0103] The model can be trained on a database, which is generated based on previously captured vascular data, patient data, surgeries, and / or the results of surgeries subsequently performed on the vessels in the captured vascular data. Data can be collected from cadaveric and / or living subjects. The model can be updated based on the surgeries performed on the vessels using the currently captured vascular data. In this regard, the model can include feedback loops that allow it to be continuously updated based on the latest surgeries performed on the vessels in the most recently captured vascular data, and user input indicating whether the surgeries were correct or whether further surgeries are needed. As described herein, in addition to image data, patient data can also be input into the model. Patient data can be stored in a database record along with intravascular image data corresponding to the respective patient. Feedback in the form of patient data can include measurements (e.g., heart rate, blood pressure, and / or blood flow velocity), as well as details provided by the physician performing the vascular preparation surgery. In some examples, blood flow velocity can be determined based on intravascular image data. Furthermore, patient data can include descriptions of whether and / or how the recommended surgery was deviated from when the surgery was performed. This data and / or other data can be used as training examples for further training or fine-tuning of the AI ​​model.

[0104] Figure 1The illustration shows a data collection system 100 for collecting intravascular data, calculating further data from the collected data, and providing a graphical user interface (GUI) that allows a user to view and interact with the collected and calculated intravascular data, as further described below. The system may include a data collection probe 104 for imaging a blood vessel 102. In some examples, probe 104 may be an intravascular device (e.g., an OCT probe, IVUS catheter, miniature OCT probe, near-infrared spectroscopy (NIRS) sensor, optical frequency domain imaging (OFDI), and / or any other device that can be used to image the blood vessel 102). In some examples, probe 104 may be a pressure line or flow meter, etc. Probe 104 may include a device tip, one or more radiopaque markers, optical fibers, torque lines, and / or the like. Furthermore, the device tip may include one or more data collection subsystems, such as a beamguide, acoustic beamguide, pressure detector sensor, other transducers or detectors, and combinations of the above components.

[0105] A lead (not shown) can be used to introduce probe 104 into blood vessel 102. Probe 104 can be introduced and withdrawn along the length of the blood vessel while data is being collected. As probe 104 is withdrawn or retracted, multiple scan datasets or OCT datasets and / or IVUS datasets can be collected. Frames of these datasets or image data can be used to identify features (e.g., vessel size and / or pressure and / or flow characteristics).

[0106] The probe 104 can be connected to the subsystem 108 via optical fiber 106. The subsystem 108 may include a light source (such as a laser), an interferometer with a sample arm and a reference arm, various optical paths, a clock generator, a photodiode, and / or other OCT components and / or IVUS components.

[0107] The probe 104 can be connected to the optical receiver 110. According to some examples, the optical receiver 110 can be a system based on a balanced photodiode. The optical receiver 110 can be configured to receive light collected by the probe 104.

[0108] The data collection system 100 may further include or be configured to receive data from the non-invasive imaging system 120. The non-invasive imaging system 120 may be, for example, an imaging system based on angiography, fluoroscopy, X-ray, magnetic resonance imaging (MRI), and / or computer-aided computed tomography (CT). The non-invasive imaging system 120 may be configured to perform non-invasive imaging of the blood vessel 102. According to some examples, the non-invasive imaging system 120 may acquire one or more images before, during, and / or after the withdrawal of the data collection probe 104. The non-invasive imaging system 120 may be used to image a patient, enabling decision-making and allowing the execution of various possible treatment options (e.g., stent placement). These and other imaging systems may be used to perform external or internal imaging of a patient to obtain raw data, which may include various types of image data.

[0109] The non-invasive imaging system 120 can communicate with the subsystem 108. According to some examples, the non-invasive imaging system 120 can be wirelessly connected to the subsystem 108 via a network. For example, the non-invasive imaging system 120 can be wirelessly connected to the subsystem 108 via a communication interface (e.g., Wi-Fi and / or Bluetooth). In some examples, the non-invasive imaging system 120 can communicate with the subsystem 108 via a cable (e.g., fiber optic cable). In yet another example, the external imaging device 120 can be indirectly communicatively connected to the subsystem 108 or the computing device 112. For example, the non-invasive imaging system 120 can be connected to a separate computing device (not shown) that communicates with the computing device 112. As another example, data from the imaging system 120 can be transferred from a storage device to the computing device 112 via a network and / or the like using a computer-readable storage medium.

[0110] The subsystem 108 includes a computing device 112. The computing device may include one or more processors 113, memory 114, instructions 115, data 116, and one or more modules 117.

[0111] The one or more processors 113 can be any conventional processor, such as a commercially available microprocessor. Alternatively, the one or more processors can be special-purpose devices (e.g., application-specific integrated circuits (ASICs), graphics processing units (GPUs), field-programmable gate arrays (FPGAs), and / or other hardware-based processors). Although Figure 1The illustration shows the processor, memory, and other components of device 112 within the same block; however, those skilled in the art will understand that a processor, computing device, or memory may actually include multiple processors, computing devices, or memories stored or not stored within the same physical housing. Similarly, memory may be a hard disk or other storage medium located in a housing different from that of device 112. Therefore, references to processors or computing devices will be understood to include a collection of processors, or computing devices, or memories that may or may not operate in parallel.

[0112] Memory 114 may store information accessible to a processor (including instructions 115 executable by processor 113) and data 116. Memory 114 may be a type of memory operable for storing information accessible to processor 113, including non-transitory computer-readable media, or other media storing data readable by an electronic device, such as hard disks, memory cards, read-only memory (“ROM”), random access memory (“RAM”), and / or optical discs, as well as other writable and read-only memories. The subject matter disclosed herein may include different combinations of the above, whereby different portions of instructions 115 and data 116 are stored on different types of media.

[0113] The processor 113 may retrieve, store, and / or modify the memory 114 according to instructions 115. For example, although this disclosure is not limited to a particular data structure, the data 116 may be stored in a computer register, stored in a relational database as a table with multiple different fields and records, stored in an XML document, and / or a flat file. The data 116 may also be formatted in a computer-readable format, such as, but not limited to, binary values, ASCII codes, or Unicode. By further example only, the data 116 may be stored as a bitmap composed of pixels (which are stored in a compressed or uncompressed manner), or various image formats (e.g., JPEG), vector-based formats (e.g., SVG), or computer instructions for drawing graphics. Furthermore, the data 116 may include information sufficient to identify the relevant information, such as numbers, descriptive text, proprietary code, pointers, references to data stored in other memory (including other network locations), and / or information used by functions to calculate the relevant data.

[0114] Instruction 115 can be any set of instructions (e.g., machine code) that is directly executed by processor 113, or any set of instructions (e.g., a script) that is indirectly executed by processor 113. In this regard, the terms “instruction,” “application,” “step,” and “program” may be used interchangeably herein. Instructions may be stored in object code format for direct processing by the processor, or in any other computing device language, including scripts, or collections of independent source code modules that are interpreted on demand or compiled in advance. The function, methods, and routines of the instructions will be explained in detail below.

[0115] According to some examples, computing device 112 may receive data from probe 104 via a wired connection or via a wireless connection. The data may include, for example, intravascular data, including intravascular imaging data, pressure data, temperature data, flow data, and / or the like. In some examples where probe 104 is an intravascular data collection device, the data received from the device may be used to determine characterizations associated with: lesions, plaque burden, fractional flow reserve (“FFR”) measurements at one or more locations along the vessel, calcium angle, external elastic layer (“EEL”) detection, calcium detection, proximal frames, distal frames, EEL-based metrics, stent / stent-free decisions, scoring, recommendations for debulking and other procedures, evidence-based recommendations notified by automatic detection of regions / features of interest, and / or stent planning, etc.

[0116] Module 117 may include various modules configured to perform a variety of functions, which are described in more detail later in this document. Such modules may include suggested surgical modules and display modules. In some examples, other types of modules may be included (e.g., modules for calculating other vascular characteristics, and / or stent detection modules, etc.). According to some examples, these modules may include image data processing pipelines or component modules thereof. The image processing pipeline can be used to transform collected vascular data (e.g., image data) into two-dimensional (“2D”) and / or three-dimensional (“3D”) views and / or representations of vessels, stents, and / or detection areas.

[0117] Based on some examples, the module may additionally or alternatively include other types of image processing modules (e.g., video processing software modules, preprocessing software modules, image file size reduction software modules, catheter removal software modules, shadow removal software modules, vessel enhancement software modules, speckle enhancement software modules, Gaussian Laplacian filter or transform software modules, guide detection software modules, anatomical feature detection software modules, fixed marker detection software modules, background subtraction modules, Frangi vessel software modules, image intensity sampling modules, moving marker detection software modules, iterative centerline testing software modules, morphological closure operation software modules, feature tracking software modules, catheter detection software modules, bottom cap filter software modules, path detection software modules, Dijkstra). The software modules include Dijkstra, Viterbi, fast-travel-based software modules, vessel centerline generation software modules, vessel centerline tracking software modules, Hessian software modules, intensity sampling software modules, image intensity overlay software modules, and other suitable software modules as described herein. For example, modules may include the Frangi vessel software module, which can be highly effective for enhancing tubular structures (e.g., vessels) in intravascular imaging. This module can improve the contrast and depiction of vessel boundaries, which is crucial for accurate lumen segmentation and lesion characterization. This module can be particularly useful in preprocessing OCT and IVUS data to support downstream tasks such as calcium detection and stent planning.

[0118] According to some examples, modules may include software (e.g., preprocessing software), transformations, matrices, and / or other software-based components for processing image data or responding to patient triggers to facilitate co-registration of different types of image data by other software-based components and to facilitate functionality suitable for implementing various features of this disclosure. Modules may include lumen detection using scan-line-based or image-based methods, stent detection using scan-line-based or image-based methods, indicator generation, generation of juxtaposition bars for stent planning, lead shadow indicators to prevent confusion with dissections, lateral branches, and missing data, and / or others. For example, a module may include lumen detection using scan-line-based methods, which can provide robust and computationally efficient detection of vascular lumen boundaries. This module may be well-suited for real-time applications and can be seamlessly integrated with OCT pullback data. Accurate lumen detection is fundamental to calculating metrics (e.g., minimum luminal area or percentage of stenosis) and identifying lesion locations relative to anatomical landmarks.

[0119] To facilitate co-registration of different types of image data, the module can co-register two or more representations from two or more retractions of image data. Therefore, two or more representations of retractions along the overlapping portions of a blood vessel can be integrated together. The retractions can utilize one or more types of intravascular imaging devices (e.g., optical coherence tomography (OCT) probes, intravascular ultrasound (IVUS) catheters, miniature OCT probes, near-infrared spectroscopy (NIRS) sensors, optical frequency domain imaging (OFDI) devices, or any other device that can be used to image blood vessels).

[0120] Registration of representations may include, for example, identifying a location in one representation and identifying comparable or equivalent locations in other representations. As an example, if a location is identified in a first representation, a comparable location will be identified in a second representation. Continuing this example, if a location in the first representation is identified as location X along a blood vessel, then a comparable location in the second representation will also be location X along a blood vessel or substantially location X. Locations can be identified based on input or user instructions received by the system. For example, the system is configured to receive input or user instructions via a GUI or other component corresponding to the selection of a location along a first or second representation of a blood vessel. This location can be, for example, an easily identifiable location along a blood vessel, such as a lateral branch, bifurcation, bend, proximal or distal end of a previously implanted device, etc. The system is further configured to receive another input or user instruction corresponding to the selection of a comparable location along another representation of a blood vessel. Based on the received input, the system performs co-registration of the representations based on the understanding that "the selected location is in the same or substantially the same location along the blood vessel in each withdrawal."

[0121] In some examples, location can be automatically identified. For example, the system can automatically identify anatomical features or previously implanted medical devices (e.g., stents). The system can compare anatomical features in representations to determine which anatomical features are identical or substantially identical. In some examples, the system can determine that anatomical features are identical or substantially identical based on patterns or sequences of anatomical features. For example, if the system identifies a lateral branch, then a tortuosity, and then a vascular region with increased calcium load in a first representation, the system can identify a corresponding region in a second representation that includes the same sequence of lateral branches, tortuosity, and then increased calcium load. Based on the comparison, the system can register a given location in the first representation with a comparable location in the second representation. The system can similarly register the remainder of these representations.

[0122] In some examples, the module can be configured to use AI models, ML algorithms, and / or the like to process vascular data obtained by probe 104 and / or imaging system 120.

[0123] Subsystem 108 may include a display 118 for outputting content to a user. As shown, display 118 is separate from computing device 112. However, according to some examples, display 118 may be part of computing device 112. Display 118 may output image data relating to one or more features detected in a blood vessel. For example, output may include, but is not limited to, cross-sectional scan data, longitudinal scans, diameter maps, and / or image masks. Output may further include visualization indicators of lesion and vascular characteristics or lesion features, such as calculated pressure values, vessel size and shape, and / or similar indicators. Output may further include visual markers for candidate stent placement (e.g., overlays that highlight selected vascular areas for possible stent placement). In some examples, output includes instructions for suggested procedures. Display 118 may be identifiable by features including: text, arrows, color coding, highlighting, outlines, and / or other suitable human or machine-readable symbols.

[0124] According to some examples, display 118 can be used to present a graphical user interface (“GUI”) to a user, allowing the user to interact with computing device 112, thereby typically outputting specific content on display 118 using input forms such as a mouse, keyboard, touchpad, microphone, gesture sensor, and / or any other type of user input device. One or more steps can be performed automatically or without user input to navigate an image, input information, select input, and / or interact with input, etc. Display 118 and the input devices, along with the computing device 112 instrument, can allow transitions between different stages of a workflow, different viewing modes, etc. For example, a user can select a vascular segment for analysis, input data or commands in response to prompts while transitioning through different stages of a workflow.

[0125] In some examples, information input by the user using a GUI may include annotations. For example, the system may be configured to use a GUI to receive annotations for one or more representations of a displayed blood vessel. In some examples, annotations may be indications of plaque burden, fractional flow reserve (“FFR”) measurements at one or more locations along the vessel, calcium angles, EEL detection, calcium detection, proximal frames, distal frames, EEL-based metrics, stent / stent-free decisions, scoring, recommendations for debulking and other procedures, evidence-based recommendations notified by automatic detection of regions / features of interest, and / or stent planning, etc. In some examples, annotations may be related to and / or associated with suggested treatments. For example, annotations may be treatment device landing areas, balloon device areas, vessel preparation device areas, and / or lesion-related areas. For example, the system may receive user input corresponding to proximal and distal locations along the vessel, which correspond to the proximal and distal locations of the treatment device landing area, balloon device area, vessel preparation device area, and / or lesion-related area. Based on some examples, the system can receive input corresponding to proximal and distal locations along the blood vessel, selecting frames for the detection module. The detection module can then correct the luminal region of the selected frame by applying linear interpolation.

[0126] According to some examples, annotations may also be determined automatically by the data collection system 100. For example, the system may determine one or more of the following based on vascular data: plaque burden, FFR measurements at one or more locations along the vessel, calcium angle, EEL detection, calcium detection, proximal frame, distal frame, EEL-based indicators, stent / stentless decision, scoring, recommendations for debulking and other procedures, evidence-based recommendations notified by automatic detection of regions / features of interest, stent planning, treatment device landing area, balloon device landing area, vascular preparation device area, and / or lesion-related area, etc. The system can automatically provide plaque burden, FFR measurements at one or more locations along the vessel, calcium angle, EEL detection, calcium detection, proximal frame, distal frame, EEL-based indicators, stent / stentless decision, scoring, recommendations for debulking and other procedures, evidence-based recommendations notified by automatic detection of regions / features of interest, stent planning, treatment device landing area, balloon device landing area, vessel preparation device area, and / or lesion-related area, etc., as one or more annotations on at least one of the vessel representations.

[0127] A display 118, either alone or in conjunction with computing device 112, may allow switching between one or more viewing modes in response to user input. For example, a user may be able to switch between different intravascular data, images, etc., recorded during each of multiple retractions of probe 104 within blood vessel 102. In some examples, a user may be able to switch between different representations (e.g., longitudinal representation, cross-sectional representation, three-dimensional representation, intravascular image, color image, black and white image, live image, and / or the like).

[0128] In some examples, a display 118, either alone or in conjunction with computing device 112, may present one or more menus to a user (e.g., a physician), who can provide input by selecting an item from one or more menus in response. For example, the menu may allow the user to show or hide various features. As another example, a menu may exist for selecting vascular features to be displayed. Additionally, as yet another example, a menu may exist for selecting treatments available within a treatment facility (e.g., a hospital). For example, the menu may allow the user to select the specifications of available treatments (e.g., the diameter and / or length of available balloons) and / or the brand of available treatments. According to some examples, the display may output a menu that includes one or more inputs for analyzing and / or processing information associated with vascular data. For example, inputs may include adding and / or removing lateral branches, recalculating virtual flow reserve (VFR), measuring a selected region of interest, correcting for false positives (e.g., lumen graft modules), and / or the like.

[0129] The content output on display 118 may include one or more representations of blood vessel 102. For example, the representation may include image data, a longitudinal representation, a three-dimensional representation, a live representation, and / or the like. Image data may include, for example, image frames captured during the withdrawal of an intravascular imaging probe (e.g., OCT, NIRS, IVUS, and mini-OCT). Image data may include extraluminal image frames (e.g., angiography, computed tomography (CT), magnetic resonance imaging (MRI), and / or fluoroscopy, etc.). Image data may include pixel data representing each pixel of the image. Image data may also include metadata associated with the image (e.g., capture time, equipment used to capture the image, whether any filters or settings were used to capture the image, and more). A longitudinal representation may include, for example, a blood vessel representation based on luminal diameter that is symmetrical about the longest axis of the representation. In some examples, the representation may include a graphical representation (e.g., a graphical representation of VFR, pressure values, flow rates, and / or the like).

[0130] In some examples, one or more visual representations of an image may include indications of lesion location, lesion severity, lesion length, and / or the like. Additionally or alternatively, lesion indications may be color-coded, with each color representing severity, length, and / or other measurements associated with the lesion.

[0131] In some examples, the content output on display 118 may include candidate treatment areas. Such candidate treatment areas may be, for example, candidate stent landing areas. For example, the output may include indications corresponding to candidate proximal landing areas and candidate distal landing areas for stents. Candidate treatment areas may be determined based on determined plaque burden, lesion location, lesion length, and / or the like. Indications may be provided on any of the vascular representations (e.g., three-dimensional representation, longitudinal representation, graphical representation, and / or image data (e.g., external images)).

[0132] In some examples, the candidate treatment area may be associated with a suggested procedure. For instance, the candidate treatment area may include an indication corresponding to the location used for lesion modification treatment. This indication can be any graphic element (e.g., arrows, boxes, highlighting, colors, text labels, and / or magnification of the location, and more).

[0133] According to some examples, display 118 and / or computing device 112 may be configured to receive one or more inputs from a user corresponding to a selection of one or more representations (e.g., a representation of a blood vessel). For example, input may be received from the user corresponding to a selection of an image frame on a longitudinal representation of a blood vessel. In response, other representations being output may be updated to display corresponding indications or image frames. For example, the displayed extraluminal image may be updated to have an indication along the blood vessel corresponding to the position of the image frame selected in the longitudinal representation, a circumferential indication may be provided on a three-dimensional representation corresponding to the position of the image frame selected in the longitudinal representation, and / or a cross-sectional image frame may be updated to correspond to the image frame selected in the longitudinal representation, and so on. In some examples, blood vessel data associated with the selected position may be updated and provided for display.

[0134] Figure 1The data collection system 100 can be used to collect vascular data (e.g., image data) that can be used to identify recommended procedures associated with lesions within the vessels as part of the treatment of calcified coronary artery disease (CAD) or peripheral artery disease (PAD). CAD affects the coronary arteries that supply blood to the heart. PAD affects the peripheral arteries that supply blood from the heart to the limbs (including arms, hands, legs, and feet). Intravascular images can be IVUS images and / or OCT images. IVUS and / or OCT images provide images for evaluating coronary artery and peripheral lesions. The system can analyze the images to determine information associated with the vessels and / or lesions (e.g., detailed morphological assessment of the lesions and evaluation of the results of lesion modification techniques). In some examples, the system can analyze and / or use the images when determining recommended planning and / or guidance for PCI by selecting proximal and distal landing zones, stent diameter, and / or length. According to some examples, Figure 1 The data collection system 100 can use vascular data (e.g., image data), vascular information determined from the vascular data, and / or lesion characteristics determined from the vascular data to identify recommended surgical procedures for modifying lesions.

[0135] In fact, based on findings from intracoronary imaging, lesions (such as coronary calcifications) can be subdivided into morphological subtypes. Eccentric calcifications extend across two or fewer quadrants, thus having an arc of <180°; concentric calcifications have an arc of >180°; and nodular calcifications appear as eruptive calcium protrusions entering the lumen. For example, the depth and length of calcium are also important predictors of PCI outcomes. As determined by OCT, calcium arcs >180°, depths >0.5 mm, and lengths >5 mm increase the risk of stent under-expansion. Although both OCT and IVUS can assess the length of calcium, OCT provides a better assessment of calcium depth due to its ability to penetrate calcium. Intravascular ultrasound cannot penetrate calcium and creates acoustic shadows, thus hindering depth assessment. However, alternative markers can be used to determine calcium thickness by IVUS, as the presence of after-reverberation is associated with thinner calcium sheets (<0.5 mm), while significant shadowing indicates thicker calcifications (>1 mm). Recently, it was found that an IVUS-specific scoring system can help predict stent underexpansion using the following four criteria: (1) a calcium arc of 5 mm length > 270°, (2) presence of calcium at 360°, (3) presence of calcified nodules, and (4) adjacent vessel diameter < 3.5 mm. A score of 2 indicates that calcium modification should be performed.

[0136] Lesion modification techniques (including balloon-based techniques, plaque resection, ablation techniques, and more recently, lithotripsy-based techniques) are available to interventional cardiologists. Given the lesion-related characteristics, Figure 2The data collection system 100 can be used to identify recommended surgeries (e.g., modification techniques).

[0137] By training an AI model as described in this paper, the imaging system avoids hard-coded rules or determinism. For example, some deterministic algorithms propose different forms of cutting or notching balloons to handle eccentric calcium deposits with different degrees of curvature. Based on labeled data used to train the model, the AI ​​model can suggest procedures consistent with standard medical procedures at output, enforced by a feedback loop that allows the labels to be updated by the technician or physician performing the vascular preparation procedure. Furthermore, the AI ​​model can be updated to suggest procedures more recent or more recent than when it was initially created, and reinforced by operator feedback. For example, when new medical procedures are developed or modified, these new procedures can be fed into the AI ​​model to serve as suggested surgical outputs.

[0138] Lesion modification can be determined based on the characteristics of the lesion. Lesion modification techniques include, for example, balloon-based therapies for eccentric calcification, and ablation or lithotripsy-based therapies for concentric and nodular calcification. Laser-assisted coronary angioplasty (ELCA) has yielded a variety of different results for lesion modification. Increasingly, lesion modification techniques are considered complementary, and combinations of techniques are often advocated in the treatment of coronary calcification. Imaging after calcium modification can be used to evaluate outcomes and determine whether further modification is needed before stent implantation.

[0139] refer to Figure 1 The IVL system 10 may include a console or power supply 12 (in the form of a generator, but alternatively in the form of a laser system), a handle 14 with a therapy delivery controller 15, and a catheter 20 with two lithotripsy emitters 22 (shown as a pair of arcuate electrodes, but alternatively, they may include optical emitters or laser emitters), and a fluid-filled balloon 24. In some examples, the IVL system 10 may be combined with... Figure 2The data collection system 100 communicates with the patient. In some examples, the data collection system 100 may be configured to track the position of the balloon 24 in real time during the procedure. As described above, the AI ​​model can be used at different stages of vascular preparation to determine whether to stop or modify the procedure, for example, for preparing various types of orbital plaque resection for vessels with various conditions such as: concentric plaque, eccentric plaque, nodular plaque, severe stenosis (e.g., cross greater than 99% and pre-dilation less than 2%), long and diffuse lesions (e.g., length up to 60 mm), multi-vessel disease, and / or small-diameter vessels (e.g., 2.5 mm). For example, the AI ​​model may propose one or more modalities of vascular preparation and, upon completion and reassessment, propose one or more additional modalities of vascular preparation that are the same as or different from the previously proposed one or more modalities of vascular preparation. Once vascular preparation is complete, the AI ​​model may propose one or more additional therapies.

[0140] In some examples, the system can dynamically adapt surgical recommendations based on real-time or sequential imaging data, thereby improving safety and surgical success rates. One example of intraoperative modification can occur as an adjustment within the procedure. For example, the AI ​​model can propose an initial OA for a concentric calcified lesion with an arc >180° and moderate thickness. After OA passage, additional OCT withdrawal can indicate one or more indications that the initial OA was insufficient for complete lesion modification. In some examples, such indications may include partial calcium fracture, residual nodular calcification, and persistent luminal stenosis >70%. Once the system reprocesses the updated vascular data, the AI ​​model can suggest passage through the additional OA with a larger coronary size and / or IVL to address deeper calcium that was not fractured by the OA. This updated recommendation from the AI ​​model can be a result of identifying that the initial OA was insufficient for complete lesion modification. Therefore, the AI ​​model can recommend a complementary modality to optimize vessel compliance before stent implantation. Another example of intraoperative modification can occur as a postoperative adjustment. For example, after IVL completion and stent deployment, postoperative OCT withdrawal can indicate one or more indications of insufficient dilation. For example, these indications may include stent under-expansion (<80% of the expansion index) and residual calcium thickness >0.5 mm beneath the stent strut. Once the system reprocesses the updated vascular data, the AI ​​model can suggest mechanical optimizations (e.g., using a high-pressure non-compliant balloon after expansion and / or using IVL after stent implantation). The AI ​​model can be trained on imaging and outcomes after stent implantation, so such updated recommendations from the AI ​​model could identify outcomes where "insufficient expansion predicts restenosis." Therefore, the AI ​​model can recommend further mechanical optimizations. As another example, if dissection is detected during the procedure, the AI ​​model can change the recommendation from a drug-coated balloon (DCB) to a drug-eluting stent (DES) or a drug-eluting reabsorbable stent (DRS). As yet another example, if thrombosis is detected during imaging, the AI ​​model can recommend embolization protection or postpone lesion preparation until the thrombus burden is reduced.

[0141] An optional marker band B may be provided. The catheter 20 preferably includes a central tube defining a guidewire lumen through which the guidewire G passes to deliver the balloon 24 at a desired location along the guidewire G. A sheath may surround the central tube to define a delivery lumen through which saline solution can be controllably delivered to inflate the balloon 24. Thus, the lumen can provide a concentric space around the central tube through which electrode wires (not shown) can extend from the controller 15 to the transmitter 22 in other components according to the invention. The sheath may be connected proximally to a hub 17, which may include any number of ports allowing the electrode wires, along with saline solution for inflation, the guidewire G, and any number of other components, to be inserted into the lumen as desired.

[0142] The balloon 24 can be positioned in a constricted position to more easily traverse the patient's vascular system to reach the calcified site. In use, the balloon 24 will expand to a pressure commonly used for angioplasty procedures (e.g., 4 atm) and treatment will be initiated via the delivery controller 15.

[0143] Figure 2 A balloon 24 enlarged to a therapeutic delivery state is shown, in which a lithotripter 22 can be "launched" to destroy vascular calcifications. An optional indicator band B may be provided to provide visualization and proper positioning using known imaging techniques. The balloon 24 is enlarged to a typical angioplasty pressure (e.g., 4 atm) and treatment is performed. The balloon 24 may expand spontaneously during or after treatment to clear the vessel and allow blood to flow. Although Figure 3 Two pairs of electrode emitters 22 are shown, but in alternative embodiments, the number of electrode emitters may include only 1 pair or 2, 3, 4, 5, 6, or even more emitter pairs to address longer lesions (e.g., lesions encountered in the peripheral vascular system).

[0144] The controller 15 is used to generate one or a series of voltage pulses according to the vessel wall or lesion treatment plan. According to the illustrated embodiment, a high-voltage pulse is provided to one of the transmitters 22 (which includes a pair of spaced-apart electrodes), and then it is connected in series to a second transmitter 22 (which also includes a pair of spaced-apart electrodes). The high-voltage pulse sequentially induces a spark across the first electrode pair within the balloon 24, and then, or also induces a spark across the second electrode pair. The slightly conductive saline solution within the balloon 24 allows the high-voltage spark to cross each electrode pair, thereby creating an energy wave propagating towards the vascular calcification within the balloon.

[0145] Based on some examples, the identified procedure could be RA surgery. For instance, RA may be more useful for lesions that cannot be crossed or expanded, or for concentric calcifications, and can be combined with other modification techniques.

[0146] In another example, the identified recommended procedure could be an OA (osteovascular surgery). Vascular data can be used to identify the recommended procedure and can also provide indicators to guide OA surgical techniques. In some examples, vascular data may include: information associated with the vessel determined based on image data and / or patient history data, characterization of lesions within the vessel of interest determined based on image data and / or patient history data, and / or similar features. For example, the size and placement of the coronary relative to the landmark band or other characteristics of the OA catheter can be determined prior to surgery. In such an example, the model can identify the size and placement of the coronary relative to the OA catheter as part of the recommended procedure.

[0147] Figure 1 The diagram illustrates the use of Figure 3 An exemplary method for using an imaging system to determine the following: information associated with the imaged blood vessel, characteristics associated with lesions within the imaged blood vessel, and identification of recommended surgical procedures for modifying the lesions. The following operations need not be performed in the precise order described below. Instead, various operations can be processed in different orders or simultaneously, and operations can be added or omitted.

[0148] The data collection system 100 captures vascular data via probe 104. Vascular data can be captured prior to treatment (e.g., lesion modification surgery, vascular preparation, and / or similar treatments). The vascular data can be, for example, intravascular imaging data. Figure 3 In this example, the vascular data is OCT image data. The data collection system 100 uses the vascular data to determine information associated with the vessel and / or characteristics of the lesion. Information associated with the vessel may include, for example, plaque burden, fractional flow reserve (“FFR”) measurements at one or more locations along the vessel, calcium angle, external elastic layer (“EEL”) detection, calcium detection, proximal frame, distal frame, EEL-based indices, stent / stentless decision, lumen diameter, mean diameter, percentage of stenosis, lesion type, and / or the like. Characteristics of the lesion may include, for example, the presence of the lesion, the location of the lesion, the lesion type, and / or the lesion arc / coverage, etc.

[0149] In some examples, vascular data can be provided as input to an AI model trained to provide predictions of information associated with vascular characteristics and / or lesion characteristics. Information associated with the characteristics of the vascular vessels and / or lesions can be stored in association with the vascular data. Vascular data may include, for example, patient data corresponding to patients whose vessels are imaged, image data, and / or the like. For example, information associated with the characteristics of the vessels, lesions, and / or images can be stored in the memory of: the data collection system 100, a remote storage device, a repository, and / or the like.

[0150] The appropriate corrective surgery can be determined, at least in part, based on the characteristics of the lesion. For example... Figure 5BAs shown, when the lesion is calcified, the modification procedure can be determined based on the degree of the calcified arc, the thickness of the calcified lesion, and the length of the calcified lesion within the region of interest (ROI). In some examples, the choice of vascular modification procedure can be highly dependent on specific lesion characteristics (e.g., the degree of the calcified arc, the thickness of the calcified lesion, and / or the length of the calcified lesion within the RPI). These parameters, as predictors of surgical success and long-term outcomes, can directly influence the mechanical resistance of the lesion to expansion and the likelihood of achieving optimal stent expansion. For example, the presence of a calcified arc >270°, a thickness >0.5 mm, and a length >5 mm is associated with a significantly increased risk of stent underexpansion. Therefore, tailoring modification strategies based on these parameters can optimize stent deployment and reduce the risk of restenosis. For example, orbital plaque resection is effective for severe calcification, lesion accessibility, and / or bidirectional modification. Specifically, orbital plaque resection (OA) can modify severely calcified lesions that other devices may find difficult to traverse. Furthermore, OA is useful when the lesion is accessible but requires precise modification to facilitate stent placement. Additionally, OA can modify calcified lesions in either direction along the vessel, which is particularly advantageous for complex lesions. Therefore, track-guided plaque resection can be identified as a modifying procedure when there are higher-degree calcium arcs, greater calcium thickness, and / or longer calcium lengths in the region of interest of the vessel. As another example, IVL is effective for balloon crossing and less severe calcification. Specifically, IVL is useful when balloon crossing of the lesion is easy but adequate calcium management is uncertain; and IVL is also effective when balloon crossing is feasible and the calcified lesion is less severe. Furthermore, IVL is simple to use and does not require lead swapping, making it a convenient option in some cases. Therefore, IVL can be identified as a modifying procedure when lower-degree calcium arcs, smaller calcium thickness, and / or shorter calcium lengths are detected in the region of interest of the vessel.

[0151] As an example of surgical modification of calcium arcs, lesions with calcium arcs less than 180° respond well to puncturing balloons and / or cutting balloons. These devices can create controlled ruptures in the calcium, improving vascular compliance without high-pressure dilation. Therefore, for calcium arcs less than 180°, puncturing balloons and / or cutting balloons can be identified as modification procedures. On the other hand, lesions with calcium arcs greater than 180° may be more rigid and circumferentially constrain the vessel. In this case, IVL and / or OA may be preferred to modify the lesion and facilitate stent dilation. Therefore, for calcium arcs greater than 180°, OA and / or IVL can be identified as modification procedures.

[0152] As an example of a modification procedure for calcium thickness, IVL or high-pressure balloon angioplasty can be used to fracture calcium thicknesses less than 0.5 mm. Post-reverberation artifacts on IVUS typically indicate thin calcium, which is more suitable for balloon-based therapies. Therefore, for calcium thicknesses less than 0.5 mm, IVL and / or high-pressure balloon angioplasty can be identified as modification procedures. On the other hand, calcium thicknesses greater than 0.5 mm can be more resistant to fracture. In this case, OA is generally more effective due to its ability to ablate and abrade dense calcium (especially when IVL cannot produce sufficient fractures). Therefore, for calcium thicknesses greater than 0.5 mm, OA can be identified as a modification procedure.

[0153] As an example of surgical modification for calcium length, focal therapies (e.g., scarified balloon or short OA run) can be used to treat calcium lengths less than 5 mm. Therefore, for calcium lengths less than 5 mm, scarified balloon or short OA run can be identified as modification procedures. As another example, calcium lengths greater than 5 mm may require more aggressive therapies and / or combination therapies. For example, OA can be used to shrink the lesion, followed by IVL to achieve deeper calcium modification. Long lesions also increase the risk of stent underdistension, requiring thorough lesion preparation. Therefore, for calcium lengths greater than 5 mm, OA and / or IVL can be identified as modification procedures.

[0154] A corrective surgery can be performed, and postoperative vascular data can be captured. Postoperative vascular data may include, for example, intravascular imaging. The data collection system 100 can use the postoperative vascular data to determine the effectiveness of the corrective surgery. In some examples, the data collection system 100 may use vascular data acquired during surgery (e.g., when surgery is paused and new image data is acquired, for example, by withdrawing an intravascular imaging probe). The intravascular imaging probe may be, for example, an OCT probe, an IVUS probe, and / or the like. The system can identify the percentage improvement of the target vessel between instances of imaging taken at different points before, during, and / or after the proposed surgery. For example, the percentage improvement may be determined using the same or different AI models. Exemplary indicators for improvement may be based on changes in calcium breakage and / or changes in the level of stenosis in the vessel.

[0155] Vascular and / or patient data that may be included as features in the training data examples may include one or more of the following: age, presence of hypertension, presence of dyslipidemia or diabetes, whether the patient is currently or has been a smoker, BMI, whether the patient has a history of myocardial infarction, renal insufficiency, and an angina score from 0 to IV. Additional examples include characteristics of the target vessel (including the left anterior descending artery, circumflex artery, right coronary artery, or left main coronary artery). Other examples may include: the amount of severe calcification identified in the target vessel, and the location of the lesion, including whether the lesion is proximal, middle, distal, or at the ostium. Other lesion characteristics include lesion length and calcification length. Other examples include patient history data, such as other conditions the patient may have, patient allergy data, patient medication data, or patient-specific adverse event data (including but not limited to hypersensitivity reactions, excessive bleeding, and / or product performance failure). Patient history data is stored securely and in compliance with any applicable privacy regulations. Patient history data can be useful if the patient is undergoing subsequent surgery at the same facility, or if the patient is undergoing surgery at a different facility than the one where the patient's history data was collected. Training data can be from cadaveric subjects and / or living subjects. Examples of training data may include reference vessel diameter, minimum lumen diameter, and / or diameter stenosis. Training data can be readily available and may include, for example, preoperative image data, patient history, and / or postoperative success or failure indicators. In some examples, training data may be publicly available (e.g., information on anonymized individual patients with privacy protected).

[0156] If a percutaneous coronary intervention (PCI) has been performed previously, other exemplary features may include reference vessel diameter, lumen diameter, diameter stenosis, stent length, and / or sharp gain. Post-PCI complications may also be included as features, including severe dissection, slow flow or no return, sudden closure, and / or vessel perforation.

[0157] Other examples of features that can be provided as part of the training data include, for example, calcified coronary lesions identified using OCT, lesion length, minimum luminal area, average luminal area, regional stenosis, calcium length, maximum continuous calcium arc, average calcium arc, minimum calcium thickness, and / or calcium volume index. Other examples include characteristics associated with the minimum luminal area (MLA) site prior to IVL, percentage of area stenosis, luminal area, calcium angle, and / or maximum calcium thickness. Other examples include characteristics associated with post-stent implantation at the MLA site (e.g., area stenosis, stent area, stent expansion, sharp area gain, and calcium fracture). Other examples include characteristics associated with the maximum calcium site prior to IVL (e.g., area stenosis, luminal area, calcium angle, and / or maximum calcium thickness). Other examples include characteristics associated with the final minimum stent area (MSA) site prior to IVL (e.g., area stenosis, stent area, and / or stent expansion, sharp area gain, and / or calcium fracture). Other examples include characteristics associated with the final MSA site after stent implantation (including area stenosis, stent area, stent expansion, sharp area gain, and / or calcium fracture).

[0158] Other examples of features that can be provided as part of the training data include: OCT characteristics of calcium fractures, such as the presence of any fracture, one fracture, two fractures, or more than three fractures, fracture length, fracture depth, maximum calcium arc at the calcium fracture, minimum calcium angle at the calcium fracture, calcium fractures per lesion, calcium fractures per mm, lumen gain at the fracture site, lumen area at the fracture site, and / or stent expansion percentage. Other examples of features that can be provided as part of the training data include features for OCT symmetry, eccentricity, and poor location characteristics.

[0159] Other examples of features that can be provided include whether the patient has stable ischemic heart disease, acute coronary syndrome (non-ST-segment elevation myocardial infarction (NSTEMI) or unstable angina), or stable recent ST-segment elevation myocardial infarction (STEMI). Table 1 below shows exemplary features that can be provided as part of the training examples in the training data.

[0160]

[0161] Table 1 The selected subset reflects a practical and evidence-based approach to feature selection, balancing clinical significance, technical feasibility, and data availability.

[0162] Because predictors of surgical success have been established clinically, this subset holds promise as a strong foundation for training AI models. Lesion length, calcium arc, calcium thickness, and minimum luminal area have been established as predictors of stent under-dilation and surgical complexity. For example, a calcium arc >180°, calcium thickness >0.5 mm, and calcium length >5 mm are characteristics associated with poor stent dilation. Additionally, minimum luminal area and diameter stenosis are commonly used to assess lesion severity and guide PCI decisions.

[0163] Due to its alignment with imaging modalities, this subset also holds promise as a robust foundation for training AI models. The features in Table 1 can be measured using OCT and / or IVUS, which are the primary imaging modalities integrated into the system. Therefore, AI models can be trained on data that is both accessible and reliable within the real-world workflow of clinical practice.

[0164] Because of the balanced representation of pre- and post-operative data, this subset also holds promise as a strong foundation for training AI models. The subset in Table 1 includes pre-interventional (e.g., calcium angle and / or lesion length) and post-interventional (e.g., stent expansion and / or lumen gain) metrics. Therefore, AI models can be trained within the complete surgical context, and thus can be better predicted for outcomes or suggested modifications.

[0165] Due to its practicality and interpretability, this subset also holds promise as a robust foundation for training AI models. The features in Table 1 are interpretable by clinicians and aligned with existing scoring systems, such as OCT-based calcium scores and / or IVUS-based predictive factors. This facilitates the clinical acceptance and validation of AI-generated recommendations.

[0166] Due to data availability and consistency, this subset also promises to serve as a robust foundation for training AI models. The features in Table 1 are typically documented in clinical practice and registries, making them more likely to be used in sufficient quantity and quality for AI model training. Their inclusion supports the generalizability and robustness of AI models.

[0167] In these examples, an AI model can be trained based on a supervised learning problem to classify changes in blood vessels based on input image data and predict percentage changes or improvements. For instance, the AI ​​model can be trained using training examples labeled according to percentage changes, where backpropagation with gradient descent and weight updates is used to update model parameters. The objective function can be a distance-based loss type (e.g., L1 or L2 loss). The system can compare the predicted changes from the AI ​​model to a predetermined threshold for improvement, which can be set, for example, by the technician or physician performing the procedure.

[0168] Based on some examples, modifying procedures performed on vascular and / or postoperative vascular data can be associated with initial information, characteristics, and images. Commonly, the information, characteristics, images, performed procedures, and postoperative vascular data can be used to form a database of training data for training a model to identify recommended procedures. For example, the database can provide indications that a given modifying procedure is effective for lesions of "x" degree, "y" thickness, and "z" length, while different modifying procedures are ineffective for similar lesions.

[0169] During execution, the model trained to provide suggested procedures (e.g., suggested lesion modification procedures) may receive current vascular data as input. This current vascular data may be vascular data collected by the data collection system 100 and not yet part of the training data used for the model. In some examples, in addition to these or alternatively, the model may receive information associated with the blood vessel and / or characteristics of the lesion within the vessel as input.

[0170] Machine learning (ML) can be used to provide preparation guidance. Specifically, training an ML algorithm that incorporates labeled OCT data and various vascular preparation techniques that can be used in conjunction with predicted surgical outcomes can be useful. The labeled OCT data can include calcium information from an artificial intelligence (AI) / ML algorithm, which can be used to detect calcium and the external elastic layer (EEL) of the blood vessel. U.S. Patent Publication No. 2020 / 0294659 discloses an imaging and evaluation system for detecting EEL-based indicators.

[0171] AI models can include models or modules for image processing. For example, an AI model can include a convolutional neural network, or one or more convolutional layers, one or more pooling layers, and one or more fully connected layers or other neural network layers. Other exemplary architectures that can be implemented at least partially include other types of deep neural networks, recurrent neural networks (RNNs) (e.g., Long Short-Term Memory (LSTM), Generative Adversarial Networks (GANs), and / or transformer architectures, including transformers with self-attention and / or cross-attention mechanisms.

[0172] A variety of activation functions, model layer sizes, types, and / or other model hyperparameter values ​​can be used. For example, the model can implement Rectified Linear Unit (ReLU), Swish, or tanh as exemplary activation functions.

[0173] AI models can be trained, for example, by an OCT system or another system, to perform one or more machine learning tasks. AI models can be trained to perform individual tasks in a pipeline or to perform a task end-to-end. For example, an AI model can be trained to receive image data (e.g., intravascular and / or extravascular image data) and provide characteristics associated with the vessel (e.g., morphology or tissue properties, classification of the lesion type depicted in the image data, location of intravascular lesions, calcium or other abnormalities, and / or size or severity of the lesion, and more) as output. AI models can be trained to perform the task using supervised learning methods. For example, an AI model can receive training data including instances of image data and annotate that training data with characteristics predicted by the trained AI model. AI models can be initialized with random or semi-random model parameter values, and the AI ​​model receives and processes the training data to generate a set of predictions. The loss between model predictions and annotations can be calculated using one of several possible loss functions (e.g., L1 or L2 loss, Jaccard loss, cross-entropy loss, or any of a variety of other loss functions used to measure the difference between the predicted output and the true label).

[0174] The total error of each loss calculated by processing the training data can be used to perform backpropagation with model parameter updates to update the model parameter values ​​of the AI ​​model. The AI ​​model can be trained through one or more iterations, each iteration including this processing / forward pass, loss calculation, and backpropagation with model parameter updates. The AI ​​model can be trained until one or more stopping criteria are met. Stopping criteria can include exceeding a maximum wall-clock time, minimum loss improvement between unmet iterations, minimum accuracy in the model output when met, and more. Any of a variety of techniques can be used to train the model or improve the performance of the model training and / or the trained model, such as stochastic gradient descent with model parameter updates, batch gradient descent, momentum gradient descent, regularization, exit, and / or scheduling for changing model hyperparameters (e.g., learning rate, batch size of training inputs, etc.).

[0175] AI models can be trained or fine-tuned through offline training processes, online training processes, or a combination of both. For example, an AI model can be periodically retrained based on the contents of an updated database, or fine-tuned using updated examples. In some examples, such as when the system receives feedback from an operator regarding a suggested surgery, the AI ​​model can be fine-tuned using single or batches of vascular data annotated with that feedback. This may occur, for example, when the model has already been deployed on the system or another device communicating with the system, or when the model has been updated offline and then updated again at a future point in time.

[0176] AI models can further include components or models for generating recommendations based on received or generated OCT image data. For example, an AI model could perform an end-to-end process in which OCT image data is generated by processing received OCT images, and then blood vessel preparation recommendations are generated based on the learned OCT image data. (See references in this document.) Figure 4 The AI ​​model can be trained based on a multi-class classification task, where categories correspond to different vascular preparation options. The model can output instructions associated with a suggested vascular preparation procedure. The system implementing the model can further process the model output to provide, for example, data corresponding to the suggested vascular preparation procedure (including instructions, images, and / or other data that the user performing the vascular preparation may need or benefit from).

[0177] The system can also prompt the user for input to confirm or reject the recommended vascular preparation. The system can receive user input in the form of a confirmation / rejection instruction, along with possible additional information explaining why the recommended vascular preparation was rejected or accepted. The suggested recommendations, along with the OCT image data and the user's confirmation / rejection, can be stored in a database 199, which can be accessed later to fine-tune the AI ​​model using updated training examples.

[0178] The system can also prompt the user for input to limit the suggested procedure to treatments available within a treatment facility (e.g., a hospital). The system can receive user input regarding available treatments (including specifications such as the diameter and / or length of available balloons) and / or the brand of available treatments. The system can then output a suggested procedure that corresponds to the input available treatments and their compatibility with the specified lesion.

[0179] Figure 4 Figure 5 illustrates an example that uses OCT as an imaging-based modality to train an AI model (e.g., a convolutional neural network (CNN)). The suggested procedures output by this AI model can be used to build a database, as described below, and for suggesting vascular preparation options and potential outcomes. As mentioned above, the AI ​​model may include one or more modules that implement CNNs or neural network layers.

[0180] exist Figure 5B During this process, OCT surgery can be performed, including one or more retractions (as described above). Following this, vascular preparation can be carried out, and the aspects and parameters of the preparation, as well as the conclusions or results of the vascular preparation, are input into the AI ​​model. These steps are primarily used in the AI ​​model building phase.

[0181] Figure 2This is a flowchart of an exemplary process 500B according to various aspects of this disclosure, used to train an AI model to generate suggested vascular preparation options. A system with one or more processors (e.g., system 100) can execute process 500B. Some operations in the process described herein may be omitted or performed multiple times (e.g., iteratively or in parallel). In some examples, other operations are added to the process and / or performed in a different order.

[0182] According to box 510B, the system receives training data, which includes examples of OCT image data annotated with labels corresponding to the types of vascular preparations performed on the vessels characterized in the OCT image data. The training data may include OCT image data as described herein, which may be stored in a database (e.g., database 199) along with labels annotating the types of vascular preparations performed, such as the types of vascular preparations performed on patients from whom the OCT image data was obtained. See the flowchart referenced herein. Figure 4 To process Figure 5A The database can be managed and updated by the system, which in turn can provide additional training examples for further training or fine-tuning of the AI ​​model.

[0183] Training data may further include other data (e.g., patient history data corresponding to the OCT image data). Historical data may include additional medical information about the patient that differs from the information found in the OCT image data (e.g., other conditions the patient may have, patient allergy data, patient medication data, or patient product-specific adverse event data, including but not limited to hypersensitivity reactions, excessive bleeding, and / or product performance failure). Patient history data is stored securely and in compliance with any applicable privacy regulations. Patient history data can be useful if the patient undergoes subsequent surgery at the same facility, or if the patient undergoes surgery at a different facility than the one where the patient's history data was collected. Training data may also include the results of performing vascular preparation techniques on vessels indicated by the OCT image data.

[0184] According to box 520B, the system processes training data through an AI model to determine one or more losses. For example, the AI ​​model can generate indications of suggested vascular preparation procedures from a list of multiple categories. The system can generate a loss between the generated indications and ground truth labels, which indicate vascular preparations associated with input training examples. The system can compute any of a variety of different loss functions used for multi-class classification problems (e.g., classification cross-entropy loss). To generate indications corresponding to the suggested recommendations, the model can implement a softmax layer to generate output probabilities corresponding to various different vascular preparation procedures represented by multiple categories. The system can receive the output probabilities and determine the recommended recommendation, for example, based on the recommendation with the highest probability in the list of output probabilities.

[0185] According to box 530B, the system updates one or more model parameter values ​​based on one or more determined losses. The system can perform, for example, stochastic or batch gradient descent with backpropagation to update the model parameter values ​​of the AI ​​model. The model can be trained using any of a variety of different techniques for machine learning training, including supervised learning methods. The system can compute the loss based on the cross-entropy loss between the predicted class and the ground truth label in the training data.

[0186] In some examples, the model can be further trained using one or more unsupervised learning techniques. For instance, training in an unsupervised context can be based on clustering groups of unlabeled training data to discover similarities or differences between examples. In other examples, dimensionality reduction techniques can be applied to train the model to reduce the total amount of data required for the model to derive recommended vascular preparation procedures.

[0187] According to box 540B, the system determines whether the stopping criteria have been met. As described herein, stopping criteria may include exceeding the maximum wall-clock time, failing to meet the minimum loss improvement between iterations, and / or meeting the minimum accuracy in the model output, and more. According to box 550B, if the stopping criteria have been met (“Yes”), the system outputs the trained model. Otherwise (“No”), the system may perform another training iteration to process the training data and update the model parameter values ​​according to boxes 520B and 530B, respectively. Figures 6 to 8 This refers to an AI model that has undergone sufficient OCT procedures (with retraction) and been fed into it along with any other relevant patient data (such as the patient's biometrics). The trained AI model is then ready to recommend one or more preparation methods or treatments with optimal outcomes, or options associated with possible outcomes, which may require consideration of risks or other options. For example, the AI ​​model can be trained on 250 or more examples.

[0188] For example, a trained AI model can be deployed in the system so that when a new OCT retraction is performed, the system can process the new data using the AI ​​model to provide recommendations on the most appropriate preparation method or technique, or to offer multiple options to choose from. In some examples, the OCT system may include a display device that can show information on any lesion type for the operator to view and / or any vascular modifications or treatments recommended or offered as options.

[0189] Figure 6 An exemplary method is illustrated in which the above-described techniques (e.g., using an OCT system to image lesions and to recommend and / or predict the outcomes of specific treatments or techniques) are built, deployed, and used to create a final usable database for analyzing lesions.

[0190] Figure 1 This is an exemplary logical flow and data provided by OCT to create a database of vascular preparation and results. According to box 810, the system can generate image data corresponding to the region of interest from the pre-treatment OCT. The system can generate vascular data as described herein or receive OCT image data from another source. Vascular data may include, for example, data characterizing lesion arc, morphology, lesion length, and / or vessel diameter. OCT image data can be used as data stored in a database (e.g., a reference database). Figure 7 A portion of the records in the database 199 shown and described. Additionally, as described herein, the system can co-register two or more representations of image data. For example, the system can co-register two or more overlapping retractions of one or more types of intravascular imaging devices (e.g., optical coherence tomography (OCT) probes, intravascular ultrasound (IVUS) catheters, miniature OCT probes, near-infrared spectroscopy (NIRS) sensors, optical frequency domain imaging (“OFDI”), any other device that can be used to image blood vessels, or combinations thereof). The system can also receive patient history data.

[0191] According to box 820, the system receives input suggestions for a vascular preparation procedure corresponding to the vascular data. According to box 830, vascular preparation may be performed, for example, by an automated device or a user (e.g., a physician). According to box 840, vascular data for the region of interest is generated or received after surgery. According to box 850, the vascular preparation results may be recorded as part of a subject database record. According to box 860, the system may receive instructions regarding whether additional or different treatments are needed. Vascular data with instructions for the corresponding vascular preparation techniques to be performed may be recorded as an item in database 199 and used to train an AI model to generate recommendations based on new data. Updates to the database (including updated vascular data, updated patient data, results of performed surgeries, and updates to surgical capabilities) may be provided as additional training examples for training or fine-tuning the AI ​​model.

[0192] Figure 8 This is an exemplary logical flow when creating a database of OCT information and vascular preparation results, and when the model recommends which vascular preparation to use. The user can agree or disagree with the vascular preparation recommendations to continue enhancing the database. According to box 910, the system generates or receives vascular data for the region of interest. According to box 920, the system outputs the recommended vascular preparation. According to box 930, the system can prompt and receive user input indicating whether the recommended vascular preparation is correct. The system can update the database 199 based on the user input provided according to box 930, for example, to add additional database records, including vascular data and recommended vascular preparation (if the model is correct) or different vascular preparation techniques (if the recommended vascular preparation is incorrect). According to box 940, vascular preparation can be performed, and according to box 950, updated vascular data can be received or generated on the region of interest after surgery. According to box 970, the vascular preparation results can be input into the AI ​​model to determine if additional or different vascular preparations are needed. If additional or different vascular preparations are needed, the system can receive the vascular data generated as in box 960 and provide new recommendations according to box 920.

[0193] Figures 9 to 15The illustration depicts a suggested procedure (e.g., a vascular preparation modality) based on the type of lesion. Multiple categories used by the AI ​​model for possible classification can include these modalities, as well as other possible modalities. For example, the model can be trained to output an indication suggesting a percutaneous transluminal coronary angioplasty (PTCA) balloon preparation method. Exemplary categories, as shown in flowchart 4, include PTCA balloon, notched balloon (for mixed plaques), notched balloon (for fibrous plaques), IVL, OAS (for calcified lesions with an arc greater than 180 degrees), or IVL (for calcified lesions with an arc less than 180 degrees). As described herein, the model can be trained to generate a list of output probabilities, where each probability represents the probability of a prediction made by the model, i.e., how likely it is that the recommended procedure corresponds to the correct vascular preparation to be performed, based on the received OCT image data.

[0194] The recommended procedure may be limited to available treatments within a treatment facility (e.g., a hospital). The system may receive user input regarding available treatments, including the specifications of the available treatments (e.g., the diameter and / or length of the available balloon) and / or the brand of the available treatments. The system can then output a recommended procedure that corresponds to the input available treatments and their compatibility with treating the specified lesion.

[0195] The recommended procedures may include instructions on the methods used to treat the patient. For example, instructions may include limitations on specific procedures, such as whether there are limitations on how long IVL can be used to treat the lesion (e.g., IVL applicability for lesions ≤20 mm). In addition to instructions regarding lesion length limitations, the system may provide a variety of other context-specific instructions to guide the user. These instructions may be dynamically generated based on lesion characteristics, patient-specific data, and institutional constraints, designed to enhance surgical safety, efficacy, and adherence to best practices. These instructions may include device compatibility and size, surgical sequences, lesion-specific warnings, imaging and validation, and / or institutional or operational constraints. Device compatibility and size instructions may include guidance on the balloon-to-artery ratio (e.g., “Use a balloon with a diameter 1.1 times the reference vessel diameter”), stent size recommendations (which may be based on proximal and distal reference diameters), and / or device length selection (which may be based on lesion length plus margins (e.g., “Select a stent at least 5 mm longer than the lesion”)). Surgical sequence instructions may include pre-dilation requirements (e.g., “Pre-dilation with a semi-compliant balloon before IVL”), adjunctive therapy (e.g., “Consider balloon scoring after RA to optimize lesion compliance”), and / or post-dilation pressure range (e.g., “Dilate to 16-18 atm after using a non-compliant balloon”). Lesion-specific warning instructions may include calcium morphology alerts (e.g., “Nodular calcium detected – consider OA instead of IVL”), thrombosis alerts (e.g., “Thrombosis detected – consider using an embolization-protective filter”), and / or dissection risk alerts (e.g., “Avoid high-pressure dilation due to eccentric calcium and thin vessel wall”). Imaging and validation instructions may include imaging follow-up (e.g., “Repeat OCT withdrawal after lesion modification to assess calcium fracture”), and / or landing zone confirmation (e.g., “Ensure the stent landing zone does not contain calcium with an arc >180°”). Institutional or operator constraint instructions may include equipment availability filters (e.g., “Recommended equipment is unavailable – select an alternative equipment from the approved list”), and / or use-based instructions (IFU) constraints (e.g., “This equipment is not suitable for vessels with a diameter <2.5 mm”).

[0196] The recommended surgeries can include updates to the capabilities for each surgery. As each surgery is further developed, these additional capabilities can be added to the database (e.g., database 199) to output updated recommended surgeries.

[0197] The system can transform AI model outputs into executable surgical and / or therapeutic options through a structured and interpretable pipeline. This process may include internal model logic and optional post-processing steps. For example, the system may include structured post-processing and context filtering to ensure the final output is clinically executable, interpretable, and / or aligned with institution- and patient-specific constraints. In some examples, the AI ​​model can be trained to output a ranked list of surgical options with a single class label (e.g., “IVL”, “OA”, “RA”, “crippled balloon”) and / or associated confidence scores (e.g., IVL: 0.82; OA: 0.65; RA: 0.41). These outputs can be derived from a multi-class classification model, which can be trained on imaging data, annotations, and / or surgical outcomes. To transform the output into recommended surgeries, a lookup table and / or rule engine can be used to map the raw model output into surgical recommendations. This lookup table and / or rule engine associates each category with: surgery name, equipment type and size range, indications and contraindications, and / or IFU-based constraints (e.g., lesion length, vessel diameter). This mapping ensures that the output is a clinically meaningful recommendation. Optionally, additional post-processing can be applied to refine or filter the output. For example, post-processing may include: equipment availability filtering (e.g., removing options not in stock at the facility), IFU compliance checks (e.g., ensuring recommendations align with regulatory indications), conflict resolution logic (if multiple surgeries are equally ranked, the system can prioritize recommendations based on lesion morphology, operator preference, and / or historical success rate), and / or patient-specific contraindications (e.g., filtering options based on allergies, comorbidities, and / or previous adverse events). Recommended procedures can be presented in a GUI, which may include the primary recommendation (optionally including logical justification), alternative options (optionally utilizing confidence scores), visual overlays on imaging data (e.g., highlighting the calcium arc or stent landing area), and / or interactive elements for user feedback or overriding. Feedback loops may include recording user choices and outcomes to update the model over time, improve future recommendations, and enable personalization.

[0198] Figure 1 This is an example GUI display, which can be based on using Figure 9The data collection system 100 collects and / or processes vascular data as output. The output includes at least one representation of the vascular data. As shown, the representation includes a cross-sectional intravascular image, an angiographic image including the vessel of interest, and a two-dimensional representation of the vessel. The two-dimensional representation is symmetrical about the longest axis of the axis. Each representation may include one or more indicators corresponding to information associated with characteristics of the vessel or lesion. Indicators may include, for example, numerical values ​​(e.g., total angle of the lesion, maximum thickness of the lesion, EEL value, lumen diameter value, and / or the like). Indicators may be color-coded, where each color represents a different type of tissue or lesion. In some examples, indicators may be color-coded to indicate the severity of the lesion and / or the like.

[0199] Based on some examples, a display may include one or more inputs. The inputs can be configured to receive user input corresponding to a selection (e.g., a selection from a drop-down menu or typed input). Figure 10 As shown, the input may include instructions for providing the performed vascular preparation and / or any post-vascular preparation therapy. Vascular preparation and / or post-vascular preparation may correspond to lesion modification surgery. In some examples, a user (e.g., a physician) may provide this information, causing the data collection system 100 to store it in a database for training patterns to provide recommended surgeries.

[0200] Following vascular preparation surgery, postoperative vascular data can be captured by data collection system 100. This postoperative vascular data can be processed by data collection system 100 or another remote system to determine postoperative information about the blood vessels and / or postoperative characteristics of the lesions. Based on the postoperative information and / or postoperative characteristics, the system can automatically determine the outcome of the vascular preparation (e.g., success, need for additional modifications, failure, and / or the like). Figure 9As shown, in some examples, the display may include input configured to receive selections (e.g., selections from a drop-down menu, or typed input associated with the vascular preparation result). Postoperative vascular data may indicate that no lesions remain in the vessel. In some examples, when there are no lesions to be treated in the vessel, the result may be indicated as "good" or "successful." In some examples, after a retouching procedure, when the lesions within the vessel are below a first threshold, the result of the retouching procedure may be determined to be good. The first threshold may correspond to a first threshold such as the length, thickness, and / or angle of the lesion. In examples where, after a retouching procedure, the lesions within the vessel are above the first threshold but below a second threshold, the retouching procedure may be determined to be "good" or "requires additional retouching." The second threshold may correspond to a second threshold such as the length, thickness, and / or angle of the lesion. In yet another example, after a retouching procedure, when the lesions within the vessel are above a third threshold, the output of the retouching procedure may be determined to be a failure. The third threshold may correspond to a third threshold such as the length, thickness, and / or angle of the lesion.

[0201] via Figure 10 and Figure 11 The information received from user input can be stored in database 199 and used as training data for the model.

[0202] Figure 12 and Figure 1 An exemplary display is illustrated, which is used based on via Figure 9 The data collection system 100 receives vascular data to provide surgical recommendations. When the database 199, for example, is accessed via... Figure 10 and Figure 11 When storing more information about vascular data and associated modifying procedures based on user input, the data collection system 100 can execute one or more models to provide suggested procedures and / or follow-up procedures after the initial modifying procedure.

[0203] For example, in Figure 12In this system, the data collection system 100 can receive initial vascular data associated with at least one region of the vessel of interest. The data collection system 100 executing the model can determine information associated with the vessel of interest and / or the characteristics of the lesion within the vessel of interest. Information associated with the characteristics of the vessel of interest and / or the lesion can be provided as input to the model, providing a suggested procedure (e.g., a “suggested vascular preparation procedure”) during model execution. In some examples, the model or another model can provide a subsequent suggested procedure (e.g., a “suggested post-preparation” procedure). The same AI model as described herein can be used to suggest a post-preparation procedure (e.g., by providing the AI ​​model with updated vascular data to generate a suggested procedure to be performed as part of post-preparation). The subsequent suggested procedure can be a PCI procedure performed after lesion modification surgery.

[0204] Figure 1 The diagram illustrates the process of... Figure 12 An exemplary display of the output of the data collection system 100. The output may be based on vascular data received after the performance and / or completion of the recommended procedure. For example, after performing the recommended procedure, another withdrawal of probe 104 may be performed, allowing the imaging subsystem 108 to receive postoperative vascular data. The data collection system 100 can determine postoperative information associated with blood vessels and / or lesions within blood vessels by executing a model. Figure 13 As shown, postoperative vascular data indicates that a small amount of localized lesions (e.g., calcium) remain. This model can identify additional procedures (e.g., "recommended further preparation" procedures) based on the amount of lesion remaining before post-vascular preparation surgery.

[0205] Figure 14 and Figure 11 Similar to Figure 12 and Figure 13 However, it shows the recommended vascular preparation for "OAS-1.25 mm classic". Figure 13 In this process, the imaging system can provide suggested surgical procedures, including suggested equipment for performing the procedures. Based on postoperative vascular data, the data collection system 100 can provide suggested follow-up procedures (e.g., subsequent vascular preparation procedures), such as... Figure 15 As shown.

[0206] Figure 12 Similar to Figure 14 and Figure 15 ,However, Figure 16The diagram illustrates a suggested treatment for a "stent" and a total angle of zero degrees. The data collection system 100 can determine and provide suggested treatments for output, rather than providing suggested follow-up procedures (e.g., suggested further preparatory procedures) for output. Exemplary suggested treatments include the application of a drug-coated balloon; stent placement (e.g., a bare-metal stent or a drug-eluting stent); and balloon angioplasty. For example, the data collection system 100 can also generate these suggested treatments in addition to suggested vascular preparatory procedures.

[0207] In another aspect, the present invention may include imaging to detect the need for embolization filters. AI models can be trained to detect thrombi using intravascular imaging modalities. For example, in OCT images, thrombi may appear as low-signal, irregular intraluminal masses accompanied by signal-free shadows. In IVUS images, thrombi may appear as hypoechoic or hyperechoic masses protruding into the lumen. In NIRS images, lipid-rich plaques, which are prone to rupture and thrombus formation, can be detected. AI models can be trained to recognize these characteristics in intravascular imaging modalities and classify them according to: thrombus presence, type (e.g., red vs. white thrombi), and / or burden (e.g., length, volume, and / or degree of occlusion). The presence of thrombi is a known risk factor for distal embolism during interventional procedures (e.g., plaque resection, balloon angioplasty, and / or stent deployment). Manipulating thrombus-filled lesions can remove material, which may lead to microvascular obstruction, no-reflow phenomenon, or distal embolism and infarction. To mitigate these risks, AI models can recommend embolization protection devices (e.g., filters or occlusion balloons). These devices can be recommended if the thrombus burden exceeds a defined threshold (e.g., >5 mm in length or >50% of the luminal area) and / or if the procedure involves high-risk procedures (e.g., track-based or rotational plaque resection). Upon detection of a thrombus, the system can flag the lesion as high-risk, recommend embolization protection as part of the surgical plan, and / or suggest alternative strategies (e.g., delayed intervention and / or pharmacological thrombolysis). For example, thrombus information from an NIRS or OCT scan can trigger a recommendation to use an embolization filter. For instance, an AI model can receive vascular data (including image data from IVUS, NIRS, or OCT retractions) and generate a recommendation to use an embolization filter as a suggested vascular preparation procedure. During intervention on a vessel containing a thrombus, an embolization filter can be used to prevent embolic debris from traveling downstream and causing blockages in smaller vessels. During some procedures (e.g., angioplasty or plaque resection), manipulating the thrombus may dislodge debris. Embolization filters can capture detached debris, preventing it from causing blockages in smaller downstream blood vessels.

[0208] It should also be noted that users can still add feedback to the database about the results in all cases, and machine learning can be used to continue to improve the algorithm, which in turn keeps more data available to support recommendations for vascular preparation and additional treatments.

[0209] Figure 1 Depicting aspects of this disclosure (e.g., using) Figure 4 The data collection system provides a flowchart of an exemplary procedure 1600 for the proposed surgery. The following operations do not necessarily have to be performed in the exact order described below. Instead, various operations can be processed in different orders or simultaneously, and operations can be added or omitted.

[0210] According to box 1610, the system receives vascular data associated with at least a portion of a blood vessel from the imaging modality. For example, and as referenced... Figure 8 As described in Figure 5, the imaging modality can be OCT, but other modalities (e.g., micro-OCT) are also possible. Vascular data may include, for example, data characterizing the lesion arc, morphology, lesion length, and / or vessel diameter. Vascular data may also include patient data (e.g., patient data corresponding to the patient whose vessels are being imaged). In some examples, the system may further determine whether the lesion is a medial or medial lesion, and based on this determination, output an indication regarding whether the lesion comprises a medial or medial lesion.

[0211] According to box 1620, the system determines initial characteristics of a lesion within at least a portion of a blood vessel based on vascular data. As described above, the initial characteristics may include, for example, numerical values ​​(e.g., total angle of the lesion, maximum thickness of the lesion, EEL value, lumen diameter value, and / or the like). Indications may be color-coded, where each color represents a different type of tissue or lesion. In some examples, indicators may be color-coded to indicate the severity of the lesion, and / or the like. The lesion may be, for example, a calcified lesion.

[0212] According to box 1630, the system provides the initial characteristics of the identified lesion as input to the AI ​​model. The AI ​​model can be a single model or a combination of multiple models. For example, the system can use a first AI model to determine the initial characteristics and provide the output of the first AI model to a second AI model. The AI ​​model can be a CNN, a transformer, an RNN (e.g., LSTM), and / or the like.

[0213] According to box 1640, the system identifies a suggested procedure by executing an AI model to modify a lesion within at least a portion of a blood vessel. For example, see the reference above. Figure 17The AI ​​model described can output suggested procedures (e.g., PTCA balloon, puncture balloon (for mixed plaques), puncture balloon (for fibrous plaques), IVL, OAS (for calcified lesions with an arc greater than 180 degrees), or IVL (for calcified lesions with an arc less than 180 degrees).

[0214] According to box 1650, the system provides suggested procedures as output. For example, the output can be provided to a GUI display and / or saved as part of a record in a database (e.g., database 199) for updating the AI ​​model.

[0215] In some examples, process 1600 may further include: after the recommended hand surgery is completed, the system receives postoperative vascular data from the imaging modality. The system can determine updated characteristics of the lesion based on the postoperative vascular data.

[0216] Figure 1 Depicting aspects of this disclosure (e.g., using) Figure 1 The data collection system provides a flowchart of an exemplary procedure 1700 for the proposed surgery. The following operations do not necessarily need to be performed in the exact order described below. Instead, various operations can be processed in different orders or simultaneously, and operations can be added or omitted.

[0217] According to box 1710, the system receives intravascular imaging data of blood vessels from the imaging modality. For example, and as referenced... Figure 1 As described, intravascular imaging data can be received from an image probe (e.g., for OCT, NIRS, IVUS, and / or miniature OCT).

[0218] According to box 1720, the system determines information associated with blood vessels based on intravascular imaging data. For example, and as referenced... Figure 1 The system described can determine the detailed morphological assessment of lesions and the evaluation of the results of lesion modification techniques.

[0219] According to box 1730, the system determines the presence of intravascular lesions based on intravascular imaging data. As described herein, the system can provide an indication of presence, for example, as a type of visual indicator.

[0220] According to box 1740, the system determines the characteristics of the lesion based on intravascular imaging data. The characteristics of the lesion may include, for example, the presence of the lesion, its location, its type, and / or its arc / coverage.

[0221] According to box 1750, the system provides lesion characteristics as input to the AI ​​model. The AI ​​model can be a single model or a combination of multiple models. For example, the system can use a first AI model to determine the presence and / or characteristics of a lesion and provide the output of the first AI model to a second AI model. The AI ​​model can be a CNN, a transformer, an RNN (e.g., LSTM), and / or the like. The AI ​​model can receive other vascular data as input (e.g., data characterizing the imaged vessel and / or patient data corresponding to the imaged vessel).

[0222] According to box 1760, the system identifies a suggested procedure by executing an AI model, the suggested procedure being used to modify a lesion at least within at least a portion of a blood vessel. According to box 1770, the system provides the suggested procedure as output.

[0223] According to box 1780, the system stores information related to blood vessels, characteristics of identified lesions, and suggested surgical procedures from the database in memory. For example, the database could be... ​ Database 199.

[0224] According to box 1790, the system updates the AI ​​model based on the database. For example, the system can perform one or more training iterations based on updated data in the database to update the model parameter values ​​of the AI ​​model.

[0225] Unless otherwise stated, the foregoing alternative examples are not mutually exclusive, but can be implemented in various combinations to achieve unique advantages. These and other variations and combinations of the above features can be utilized without departing from the subject matter defined by the claims; therefore, the foregoing description of the embodiments should be understood in an illustrative rather than restrictive manner. Furthermore, the provision of examples described herein and the use of terms such as “e.g.,” “comprising,” and the like should not be construed as limiting the subject matter of the claims to the specific examples; rather, these examples are intended to illustrate only one of many possible embodiments. Additionally, the same reference numerals in different figures can identify the same or similar elements.

Claims

1. A system comprising: One or more processors, said one or more processors being configured to: Receive vascular data associated with at least a portion of the blood vessel from the imaging modality; The initial characteristics of the lesion within at least a portion of the blood vessel are determined based on the vascular data. The initial characteristics of the determined lesion are provided as input to the model; The model is executed to identify a proposed surgical procedure for modifying lesions within at least a portion of the blood vessel; as well as Provide the suggested surgical procedure for output.

2. The system of claim 1, wherein the one or more processors are further configured to: After the recommended surgery is completed, postoperative vascular data is received from the imaging modality; and The updated characteristics of the lesion are determined based on the postoperative vascular data.

3. The system according to claim 1 or 2, wherein the one or more processors are further configured to: The initial features, the proposed surgery, and the updated features are stored in memory as training data for the model; and The model is updated based on the initial characteristics, the suggested surgery, and the updated characteristics.

4. The system according to any one of the preceding claims, wherein the imaging modality is optical coherence tomography or intravascular ultrasound.

5. The system according to any one of the preceding claims, wherein the characteristics associated with the lesion include at least one of the total angle, thickness, or length of the lesion.

6. The system according to any one of the preceding claims, wherein the proposed procedure is orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

7. The system according to any one of the preceding claims, wherein the lesion is a calcium lesion.

8. The system according to any one of the preceding claims, wherein the one or more processors are further configured to: Based on the vascular data, it is determined whether the lesion includes a medial lesion or an inner lesion; and Based on the determination, an indication is provided regarding whether the lesion includes a middle layer lesion or an inner layer lesion for output.

9. A method comprising: One or more processors receive vascular data associated with at least a portion of a blood vessel from an imaging modality; The one or more processors determine the initial characteristics of the lesion within at least a portion of the blood vessel based on the vascular data; The one or more processors provide the initial characteristics of the determined lesion as input to the model; The one or more processors identify a proposed procedure by executing the model, the proposed procedure being used to modify a lesion within at least a portion of the blood vessel; as well as The proposed surgery is provided by the one or more processors for output.

10. The method of claim 9, further comprising: After the recommended surgery is completed, the one or more processors receive postoperative vascular data from the imaging modality; as well as The one or more processors determine the updated characteristics of the lesion based on the postoperative vascular data.

11. The method according to claim 9 or 10, further comprising: The one or more processors store the initial features, the proposed surgery, and the updated features as training data for the model in memory; as well as The model is updated by the one or more processors based on the initial characteristics, the proposed surgery, and the updated characteristics.

12. The method according to any one of claims 9 to 11, wherein the imaging modality is optical coherence tomography or intravascular ultrasound.

13. The method according to any one of claims 9 to 12, wherein the characteristics associated with the lesion include at least one of the total angle, thickness, or length of the lesion.

14. The method according to any one of claims 9 to 13, wherein the proposed procedure is orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

15. The method according to any one of claims 9 to 14, wherein the lesion is a calcified lesion.

16. The method according to any one of claims 9 to 15, further comprising: The one or more processors determine, based on the vascular data, whether the lesion includes a medial lesion or an inner lesion; as well as The one or more processors provide an indication, based on determination, that the lesion includes a middle layer lesion or an inner layer lesion for output.

17. One or more non-transitory computer-readable storage media that encodes instructions, which, when executed by one or more processors, cause the one or more processors to perform operations, the operations including: Receive vascular data associated with at least a portion of the blood vessel from the imaging modality; The initial characteristics of the lesion within at least a portion of the blood vessel are determined based on the vascular data. The initial characteristics of the determined lesion are provided as input to the model; The model is executed to identify a proposed surgical procedure for modifying a lesion within at least a portion of the blood vessel; and Provide the suggested surgical procedure for output.

18. The one or more non-transitory computer-readable storage media of claim 17, further comprising: After the recommended surgery is completed, the one or more processors receive postoperative vascular data from the imaging modality; as well as The one or more processors determine the updated characteristics of the lesion based on the postoperative vascular data.

19. The one or more non-transitory computer-readable storage media according to claim 17 or 18, further comprising: The one or more processors store the initial features, the proposed surgery, and the updated features as training data for the model in memory; as well as The model is updated by the one or more processors based on the initial characteristics, the proposed surgery, and the updated characteristics.

20. One or more non-transitory computer-readable storage media according to any one of claims 17 to 19, wherein the imaging modality is optical coherence tomography or intravascular ultrasound.

21. One or more non-transitory computer-readable storage media according to any one of claims 17 to 20, wherein the characteristics associated with the lesion include at least one of the total angle, thickness, or length of the lesion.

22. One or more non-transitory computer-readable storage media according to any one of claims 17 to 21, wherein the proposed procedure is orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

23. One or more non-transitory computer-readable storage media according to any one of claims 17 to 22, wherein the lesion is a calcium lesion.

24. One or more non-transitory computer-readable storage media according to any one of claims 17 to 23, further comprising: The one or more processors determine, based on the vascular data, whether the lesion includes a medial lesion or an inner lesion; as well as The one or more processors provide an indication, based on determination, that the lesion includes a middle layer lesion or an inner layer lesion for output.

25. A system comprising: One or more processors, said one or more processors being configured to: Receive intravascular imaging data of blood vessels from the imaging modality; Information associated with the blood vessel is determined based on the intravascular imaging data; The presence of lesions within the blood vessel is determined based on the intravascular imaging data. The characteristics of the lesion are determined based on the intravascular imaging data. The characteristics of the lesion are provided as input to the model, wherein the model is trained to identify the recommended surgery based at least in part on information associated with the blood vessel and the determined characteristics of the lesion; The model is executed to identify a proposed surgical procedure for modifying a lesion within at least a portion of the blood vessel. Provide the suggested surgical procedure for output; The information associated with the blood vessel in the database, the characteristics of the identified lesion, and the recommended surgery are stored in the memory; as well as The model is updated based on the database.

26. The system of claim 25, wherein the one or more processors are further configured to: After the recommended surgery is completed, postoperative vascular data is received from the imaging modality; and The updated characteristics of the lesion are determined based on the postoperative vascular data.

27. The system of claim 25 or 26, wherein the imaging modality is optical coherence tomography or intravascular ultrasound.

28. The system according to any one of claims 25 to 27, wherein the characteristics associated with the lesion include at least one of the total angle, thickness, or length of the lesion.

29. The system according to any one of claims 25 to 28, wherein the proposed procedure is orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

30. The system according to any one of claims 25 to 29, wherein the lesion is a calcium lesion.

31. The system according to any one of claims 25 to 30, wherein the one or more processors are further configured to: Based on the vascular data, the lesion is determined to include either a medial lesion or an inner lesion; and Based on the determination, an indication is provided regarding whether the lesion includes a middle layer lesion or an inner layer lesion for output.

32. A method comprising: Intravascular imaging data of blood vessels are received from the imaging modality by one or more processors; The one or more processors determine information associated with the blood vessel based on the intravascular imaging data; The presence of lesions within the blood vessel is determined by the one or more processors based on the intravascular imaging data; The characteristics of the lesion are determined by the one or more processors based on the intravascular imaging data. The characteristics of the lesion are provided as input to the model by the one or more processors, wherein the model is trained to identify the recommended surgery based at least in part on information associated with the blood vessel and the determined characteristics of the lesion; The one or more processors identify a proposed procedure by executing the model, the proposed procedure being used to modify a lesion within at least a portion of the blood vessel; The proposed surgery is provided by the one or more processors for output; The one or more processors store information associated with the blood vessel from the database, the characteristics of the identified lesion, and the recommended surgery into memory; as well as The model is updated by the one or more processors based on the database.

33. The method of claim 32, further comprising: After the recommended surgery is completed, the one or more processors receive postoperative vascular data from the imaging modality; as well as The one or more processors determine the updated characteristics of the lesion based on the postoperative vascular data.

34. The method according to claim 32 or 33, wherein the imaging modality is optical coherence tomography or intravascular ultrasound.

35. The method according to any one of claims 32 to 34, wherein the characteristics associated with the lesion include at least one of the total angle, thickness, or length of the lesion.

36. The method according to any one of claims 32 to 35, wherein the proposed procedure is orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

37. The method according to any one of claims 32 to 36, wherein the lesion is a calcified lesion.

38. The method according to any one of claims 32 to 37, further comprising: The one or more processors determine, based on the vascular data, whether the lesion includes a medial lesion or an inner lesion; as well as The one or more processors provide an indication, based on determination, that the lesion includes a middle layer lesion or an inner layer lesion for output.

39. One or more non-transitory computer-readable storage media that encodes instructions, which, when executed by one or more processors, cause the one or more processors to perform operations, the operations including: Receive intravascular imaging data of blood vessels from the imaging modality; Information associated with the blood vessel is determined based on the intravascular imaging data; The presence of lesions within the blood vessel is determined based on the intravascular imaging data. The characteristics of the lesion are determined based on the intravascular imaging data. The characteristics of the lesion are provided as input to the model, wherein the model is trained to identify the recommended surgery based at least in part on information associated with the blood vessel and the determined characteristics of the lesion; The model is executed to identify a proposed surgical procedure for modifying a lesion within at least a portion of the blood vessel. Provide the suggested surgical procedure for output; The information associated with the blood vessel in the database, the characteristics of the identified lesion, and the recommended surgery are stored in the memory; as well as The model is updated based on the database.

40. The one or more non-transitory computer-readable storage media of claim 39, further comprising: After the recommended surgery is completed, the one or more processors receive postoperative vascular data from the imaging modality; as well as The one or more processors determine the updated characteristics of the lesion based on the postoperative vascular data.

41. One or more non-transitory computer-readable storage media according to claim 39 or 40, wherein the imaging modality is optical coherence tomography or intravascular ultrasound.

42. One or more non-transitory computer-readable storage media according to any one of claims 39 to 41, wherein the characteristics associated with the lesion include at least one of the total angle, thickness, or length of the lesion.

43. One or more non-transitory computer-readable storage media according to any one of claims 39 to 42, wherein the proposed procedure is orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

44. One or more non-transitory computer-readable storage media according to any one of claims 39 to 43, wherein the lesion is a calcium lesion.

45. One or more non-transitory computer-readable storage media according to any one of claims 39 to 44, further comprising: The one or more processors determine, based on the vascular data, whether the lesion includes a medial lesion or an inner lesion; as well as The one or more processors provide an indication, based on determination, that the lesion includes a middle layer lesion or an inner layer lesion for output.

46. ​​A method comprising: Vascular data associated with at least a portion of a blood vessel is received by one or more processors; The one or more processors determine the initial characteristics of the lesion within at least a portion of the blood vessel based on the vascular data; The determined initial characteristics of the lesion are provided as input to the model by the one or more processors; The one or more processors identify a proposed procedure by executing the model, the proposed procedure being used to modify a lesion within at least a portion of the blood vessel; as well as The proposed surgery is provided by the one or more processors for output.

47. The method of claim 46, further comprising: After the recommended surgery is completed, the one or more processors receive the postoperative vascular data. as well as The one or more processors determine the updated characteristics of the lesion based on the postoperative vascular data.

48. The method according to claim 46 or 47, further comprising: The one or more processors store the initial features, the proposed surgery, and the updated features as training data for the model in memory; as well as The model is updated by the one or more processors based on the initial characteristics, the proposed surgery, and the updated characteristics.

49. The method according to any one of claims 46 to 48, wherein the characteristics associated with the lesion include at least one of the total angle, thickness, or length of the lesion.

50. The method according to any one of claims 46 to 49, wherein the proposed procedure is orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

51. The method according to any one of claims 46 to 50, wherein the lesion is a calcified lesion.

52. The method according to any one of claims 46 to 51, further comprising: The one or more processors determine, based on the vascular data, whether the lesion includes a medial lesion or an inner lesion; as well as The one or more processors provide an indication, based on determination, that the lesion includes a middle layer lesion or an inner layer lesion for output.

53. A method comprising: One or more processors update a model based on postoperative intravascular imaging data of blood vessels and a database, wherein the model is trained to identify recommended surgery based at least in part on determined vascular characteristics and determined characteristics of the lesion. Updating the model includes: Postoperative intravascular imaging data of the blood vessel is received by the one or more processors; The one or more processors determine vascular characteristics based on the postoperative intravascular imaging data. The presence of lesions within the blood vessel is determined by the one or more processors based on postoperative intravascular imaging data. The characteristics of the lesion are determined by the one or more processors based on the intravascular imaging data after the surgery; The characteristics of the lesion are provided as input to the model by the one or more processors; The one or more processors identify a proposed procedure by executing the model, the proposed procedure being used to modify a lesion within at least a portion of the blood vessel; The proposed surgery is provided by the one or more processors for output; and The one or more processors store the determined vascular characteristics, the determined lesion characteristics, and the recommended surgery from the database into a memory.

54. The method of claim 53, wherein the characteristics associated with the lesion include at least one of the total angle, thickness, or length of the lesion.

55. The method according to any one of claims 53 to 54, wherein the proposed procedure is orbital plaque resection, rotational plaque resection, intravascular lithotripsy, scarification balloon, laser ablation, or angioplasty balloon surgery.

56. The method according to any one of claims 53 to 55, wherein the lesion is a calcified lesion.

57. The method according to any one of claims 53 to 56, further comprising: The one or more processors determine, based on the vascular data, whether the lesion includes a medial lesion or an inner lesion; as well as The one or more processors provide an indication, based on determination, that the lesion includes a middle layer lesion or an inner layer lesion for output.

Images