**3. Marshall bundle and arrhythmogenic foci**

**Figure 7.** Endoscopic view of left PVs ostium. LSPV, left superior pulmonary vein; LIPV, left inferior pulmonary vein;

The remodeling process is associated with the structural and electrophysiological abnormality in the PVs and atrium, which could promote local conduction delay and lead to an increased PV/non-PV arrhythmogenicity developing to AF persistency [15, 16, 17, 10, 11]. These eviden‐ ces might imply that the morphological features of the PVs and atrium can contain a crucial role to detect their preexisting arrhythmogenicity, although the evaluation of the morpholog‐ ical features is limited in a quantitative manner because of their variable and unique structure.

The left atrial (LA) roof consisting of the upper wall of the left atrium and upper PVs was demonstrated as the silhouette of LA roof and could simply be visualized by PV angiography or left atrial CT imaging. In addition, the dominancy of morphological PVs/LA and the features of the LA roof silhouette could be easily determined in patients with AF. Thus, the relation between PV/non-PV arrhythmogenicity and LA roof silhouette was examined in this study.

Based on the PVs and LA dominancy, The LA roof shape was classified into a deep V shape (possible PV dominancy), shallow V shape, and flat-coved shape (possible LA dominancy) by cine angiography (Figure 8]. Angiography was conducted by both contrast media from the long sheath locating at the right and left superior PVs. The LA roof shape was assessed by A– P projection and was determined by the angle of the LA roof silhouette between the right and

Table 2 shows the relation between AMF and roof-shape group. In results, 335 AMF were detected. PV/ non-PV foci were observed in 136 of 152 (89 %) AF patients. AF triggers imme‐ diately shifting to AF were found in 114/152 (75 %) AF patients, and AF from PV foci was in 84 of 152 (55 %) AF patients. PV foci containing reproducible atrial premature contractions were observed in 135 of 152 AF patients (89 %) and non-PV foci in 77 of 152 (44 %). The location of non-PV foci was in the superior vena cava (25, 28 %), left atrial posterior wall (19, 21 %), terminal crest (10, 11 %), CS (10, 11 %), left lateral area (9, 10 %), LA roof (7, 8 %), atrial septum

, shallow V shape was 140° to180°

, and flat-

LAA, left atrial appendage.

62 Abnormal Heart Rhythms

**2.5. Left atrial roof shape and PV/non-PV foci**

left LA wall. The deep V shape was defined as <140°

.

coved shape was > 180°

(4, 4 %), and other areas (6, 7 %).

Marshall reported that a "vestigial fold of the pericardium" lies dorsal to the left atrial appendage in 1850. The small oblique vein of Marshall (VOM) is often connected to the vestigial folds going around the ostium of the left PVs. VOM drains into CS and separates the great cardiac vein and CS. The muscle sleeves of the VOM are also connected to CS musculature [26]. The vein or its ligament of Marshall is usually connected to the left PVs [27, 28, 29], and its distal ends are directly connected to the posterior wall of the left atrium [27, 30].

AF can originate from VOM or its ligament because of its catecholamine-sensitive structure [31, 5]. Preserved persistent left superior vena cava as a remnant of VOM can also include the similar electrical and anatomical features [32]. VOM or its ligament has connections to muscle bundles of the left atrium as well as of the surrounding coronary sinus (CS) in histological studies. The distal end often connects to the LA lateral area and to the left PVs [27, 29]. Therefore, recognition of the VOM anatomy in AF patients would help access to non-PV foci around the left PVs, which would lead to favorable clinical procedural result.

#### **3.1. Angiographic vein of Marshall and AMF**

In 100 AF patients, we examined the anatomy of VOM with balloon-occluded venography of coronary sinus using a balloon wedge pressure catheter (Goodtec, Huntington Beach, CA). The landmark of VOM orifice was identified at the junction of the CS and great cardiac vein. The right anterior oblique, left anterior oblique, and anteroposterior views were obtained in enrolled AF patients (Figure 9). To identify the anatomical association for VOM to both the superior and inferior left PVs, we performed selective superior and left inferior left PVs angiography by using injection of 5–10 ml of contrast medium from long sheaths. The grade of VOM development was measured from the AP view and classified into two grades (poor, reaching below superior left PV distributed in LA, and good, above superior left PVs).

**Figure 9.** Representative results of VOM angiography. The location of the VOM is indicated by arrows. VOM runs ob‐ liquely between the left atrial appendage and LSPV.

VOM was visualized by balloon-occluded CS venography in 73 AF patients (73 %). There were no significant differences in clinical characteristics of the two groups. VOM development was poor in 55 patients (75 %) and good in 18 patients. In the anteroposterior image, the VOM running behind the mitral isthmus line was confirmed. VOM going through the mitral isthmus area was observed in 51 patients (51 %). The branches originating from the end of VOM was observed in 49 patients (67 %).

The incidence of PV foci from the left superior PV was significantly higher in patients with VOM than in those without it (66 % vs. 42 %, *P*=0. 05). And also, the incidence of left superior PV foci as direct AF initiator was significantly higher in patients with VOM than in those without it (50 % vs*.* 30 %, *P*<0. 05). The incidences of e left superior PV foci were 41 % in none, 69 % in poor VOM, and 56 % in good VOM. The incidences of the left superior PV foci as direct AF initiator were 30 % in none, 56 % in poor VOM, and 33 % in good VOM.

Twelve patients had non-PV foci in the LA posterior wall, and nine (75 %) of these patients also had PV foci in the left superior PV around VOM structure even after the successful PV isolation procedure at PV ostium level. The correlation between angiographic VOM anatomy and surrounding non-PV foci is shown in Figure 4. After ablating the site at the branch of VOM connecting to the LA wall, we can often successfully terminate AF. Twenty-eight patients had 30 non-PV foci surrounding left superior PVs including LA posterior free wall, LA roof, and LA lateral wall, and 12 of 30 non-PV foci were directly shifted to AF.

AF can originate from VOM or its ligament because of its catecholamine-sensitive structure [31, 5]. Preserved persistent left superior vena cava as a remnant of VOM can also include the similar electrical and anatomical features [32]. VOM or its ligament has connections to muscle bundles of the left atrium as well as of the surrounding coronary sinus (CS) in histological studies. The distal end often connects to the LA lateral area and to the left PVs [27, 29]. Therefore, recognition of the VOM anatomy in AF patients would help access to non-PV foci

In 100 AF patients, we examined the anatomy of VOM with balloon-occluded venography of coronary sinus using a balloon wedge pressure catheter (Goodtec, Huntington Beach, CA). The landmark of VOM orifice was identified at the junction of the CS and great cardiac vein. The right anterior oblique, left anterior oblique, and anteroposterior views were obtained in enrolled AF patients (Figure 9). To identify the anatomical association for VOM to both the superior and inferior left PVs, we performed selective superior and left inferior left PVs angiography by using injection of 5–10 ml of contrast medium from long sheaths. The grade of VOM development was measured from the AP view and classified into two grades (poor, reaching below superior left PV distributed in LA, and good, above superior left PVs).

**Figure 9.** Representative results of VOM angiography. The location of the VOM is indicated by arrows. VOM runs ob‐

VOM was visualized by balloon-occluded CS venography in 73 AF patients (73 %). There were no significant differences in clinical characteristics of the two groups. VOM development was poor in 55 patients (75 %) and good in 18 patients. In the anteroposterior image, the VOM running behind the mitral isthmus line was confirmed. VOM going through the mitral isthmus area was observed in 51 patients (51 %). The branches originating from the end of VOM was

The incidence of PV foci from the left superior PV was significantly higher in patients with VOM than in those without it (66 % vs. 42 %, *P*=0. 05). And also, the incidence of left superior PV foci as direct AF initiator was significantly higher in patients with VOM than in those without it (50 % vs*.* 30 %, *P*<0. 05). The incidences of e left superior PV foci were 41 % in none, 69 % in poor VOM, and 56 % in good VOM. The incidences of the left superior PV foci as direct

AF initiator were 30 % in none, 56 % in poor VOM, and 33 % in good VOM.

around the left PVs, which would lead to favorable clinical procedural result.

**3.1. Angiographic vein of Marshall and AMF**

64 Abnormal Heart Rhythms

liquely between the left atrial appendage and LSPV.

observed in 49 patients (67 %).

The branches of the VOM were a good landmark to identify the location of non-PV foci around left PVs (Figure 10). We could successfully ablate the residual non-PV foci at the distal end of VOM in 11 patients (39 %) after PV ostial isolation (AF termination after RF delivery, 3; disappearance of reproductive atrial premature contractions, 8). Successful terminations of non-PV foci were observed in 5 in left LA posterior wall, 4 in LA lateral wall, and 2 in LA roof. Among 28 patients with non-PV foci surrounding left PVs, non-PV foci were successfully deleted in 17 patients, whereas 11 patients of them had AF recurrence. Seven of 11 (67 %) with successful non-PV foci termination were free from AF recurrence.

**Figure 10.** Spontaneous PV and coronary sinus angiography (3a). The VOM reached between LSPV and LIPV and ran laterally to the LIPV ostium. Super-selective VOM angiography revealed that the end of VOM had a branch that spreads in the area of the lateral and posterior wall below the LSPV ostium (3b). Arrows indicate the end of the VOM.

The presence of VOM is associated with a higher incidence of AF triggers of the left superior PVs. The incidence of left superior PV foci was significantly higher in patients with visible VOM than in those without visible VOM. In angiographic findings, the distal braches of VOM are commonly distributed around both the left superior and inferior PVs, especially in patients with good visual VOM. The VOM and its ligament richly innervated by sympathetic nerves could be served as a cause of isoproterenol-sensitive automatic activity [5, 33]. These evidences support arrhythmogenic foci from the VOM, and its ligament can be inducible by using highdose isoproterenol administration.

Left PVs foci and non-PV foci adjacent to the left PVs can have an influence on each other. Approximately 40 % of non-PV foci around the left superior PV were successfully ablated by targeting the distal end of VOM or its branches. These evidences demonstrate the angiographic data of VOM, and its branches may indicate the site of catheter ablation of non-PV foci. Radiofrequency energy applied to the areas of VOM distal ends occasionally delineated non-PV foci originating from the surrounding area of left PVs. Thus, we believe that understanding of the VOM anatomy can improve the clinical outcome of ablation in cases with catecholaminesensitive AF.

#### **3.2. Conduction along the left lateral ridge and the arrhythmogenicity of the left pulmonary veins**

The ligament and VOM containing the Marshall bundle (MB) with richly innervates the sympathetic and parasympathetic nerves is within the left lateral ridge (LLR) which is longitudinally running between the left atrial appendage and left pulmonary vein (LPVs), and they can serve as a source of triggers and the substrate of reentry of atrial fibrillation (AF) [1,2]. If the distinctive dominant conduction along the LLR is present, possibly due to the continuous and/or partial MB conduction, its conduction may be associated with the increased arrhyth‐ mogenicity of the LPVs. In this study, we examined the relationship between the preferential conduction properties of the MB and the arrhythmogenicity of the LPVs in 40 AF patients.

A 20-pole diagnostic catheter was positioned in the CS for pacing and recording. The upper and lower LPVs were simultaneously mapped with two adjustable 20-pole catheters (Optima, Irvine, USA) (Figure 11a). At first, RF energy during CS pacing was delivered along the LLR as a part of the LPV ablation (Figure 11b), and each ablation site and the conduction pattern during the RF delivery were monitored and recorded by fluoroscopy and a 3D electroana‐ tomical system.

The earliest activated site of the upper LPV during CS pacing was observed at the carina lesion in 32 of 40 patients (80 %), anterior wall in four of 40 (10 %), and posterior wall in four of 40 (10 %). The earliest activated site was at the upper LPV in 34 of 40 (85 %), bottom of the lower LPV in four of 40 (10 %), and posterior site in two of 40 (5 %).

After the RF delivery along the LLR, the PV potentials of the upper LPV completely disap‐ peared in one patient and that of the lower LPV in two patients. The conduction time between the LPVs and CS stimulus site was significantly prolonged during the RF delivery (before vs. after; upper, 91±26 ms vs. 127±38 ms, p<0.001; lower, 86±21ms vs. 103±22ms, p<0.001). A remarkable prolongation of more than 30 ms was observed in 19 of 40 patients (48 %) (both LPVs, 6; only the upper LPVs, 12; and only the lower LPV, 1). The sites of the remarkable prolongation during the RF delivery were observed at the carina between the LPVs [4], anterior site of the upper LPV carina [10], anterior wall of the lower LPV [3], and bottom of the lower LPVs [2].

Thirty-three AMF from LPVs (upper, 22; lower, 11) were observed in 23/40 patients (56 %). Fifteen of the detected foci directly shifted to AF, and 16 of them exhibited premature atrial contractions and/or transient frequent repetitive firings. The earliest activated site of the AMF

Left PVs foci and non-PV foci adjacent to the left PVs can have an influence on each other. Approximately 40 % of non-PV foci around the left superior PV were successfully ablated by targeting the distal end of VOM or its branches. These evidences demonstrate the angiographic data of VOM, and its branches may indicate the site of catheter ablation of non-PV foci. Radiofrequency energy applied to the areas of VOM distal ends occasionally delineated non-PV foci originating from the surrounding area of left PVs. Thus, we believe that understanding of the VOM anatomy can improve the clinical outcome of ablation in cases with catecholamine-

**3.2. Conduction along the left lateral ridge and the arrhythmogenicity of the left pulmonary**

The ligament and VOM containing the Marshall bundle (MB) with richly innervates the sympathetic and parasympathetic nerves is within the left lateral ridge (LLR) which is longitudinally running between the left atrial appendage and left pulmonary vein (LPVs), and they can serve as a source of triggers and the substrate of reentry of atrial fibrillation (AF) [1,2]. If the distinctive dominant conduction along the LLR is present, possibly due to the continuous and/or partial MB conduction, its conduction may be associated with the increased arrhyth‐ mogenicity of the LPVs. In this study, we examined the relationship between the preferential conduction properties of the MB and the arrhythmogenicity of the LPVs in 40 AF patients.

A 20-pole diagnostic catheter was positioned in the CS for pacing and recording. The upper and lower LPVs were simultaneously mapped with two adjustable 20-pole catheters (Optima, Irvine, USA) (Figure 11a). At first, RF energy during CS pacing was delivered along the LLR as a part of the LPV ablation (Figure 11b), and each ablation site and the conduction pattern during the RF delivery were monitored and recorded by fluoroscopy and a 3D electroana‐

The earliest activated site of the upper LPV during CS pacing was observed at the carina lesion in 32 of 40 patients (80 %), anterior wall in four of 40 (10 %), and posterior wall in four of 40 (10 %). The earliest activated site was at the upper LPV in 34 of 40 (85 %), bottom of the lower

After the RF delivery along the LLR, the PV potentials of the upper LPV completely disap‐ peared in one patient and that of the lower LPV in two patients. The conduction time between the LPVs and CS stimulus site was significantly prolonged during the RF delivery (before vs. after; upper, 91±26 ms vs. 127±38 ms, p<0.001; lower, 86±21ms vs. 103±22ms, p<0.001). A remarkable prolongation of more than 30 ms was observed in 19 of 40 patients (48 %) (both LPVs, 6; only the upper LPVs, 12; and only the lower LPV, 1). The sites of the remarkable prolongation during the RF delivery were observed at the carina between the LPVs [4], anterior site of the upper LPV carina [10], anterior wall of the lower LPV [3], and bottom of the lower

Thirty-three AMF from LPVs (upper, 22; lower, 11) were observed in 23/40 patients (56 %). Fifteen of the detected foci directly shifted to AF, and 16 of them exhibited premature atrial contractions and/or transient frequent repetitive firings. The earliest activated site of the AMF

LPV in four of 40 (10 %), and posterior site in two of 40 (5 %).

sensitive AF.

66 Abnormal Heart Rhythms

tomical system.

LPVs [2].

**veins**

**Figure 11.** The location of the PV/non-PV foci showing the earliest atrial focus was determined by a reference for the local electrogram and the earliest activation site of the foci. The earliest activation sites from arrhythmogenic foci and during CS pacing were determined by the direction of the spontaneous activation recorded by double spiral catheters located in both LPVs. The pacing site during the RF application was delivered from the posterolateral CS, possibly from the takeoff site of the MB. The RF application along the LLR was sequentially delivered in a lower to upper man‐ ner (from the bottom of the inferior LPV, anterior wall of the inferior LPV, and LPV carina to the anterior wall of the superior LPV) during CS pacing.

from the upper LPV was found at the carina region in 12 of 22 (55 %), anterior wall in three of 22 (14 %), roof site in three of 22 (14 %), and posterior wall in four of 22 (18 %). The earliest activated site of the AMF from the lower LPV was found at the carina region in six of 11 (55 %), anterior wall in two of 11 (18 %), bottom in one of 11 (9 %), and posterior wall in two of 11 (18 %).

The conduction time from the CS to the earliest activated upper PV after the RF delivery was significantly longer in patients with AMF from the upper LPV than in those patients without (107±36 ms vs. 146±40 ms, p<0.01), and the conduction time from the CS pacing site to the earliest activation site of the upper LPV was significantly prolonged in the patients with AMF than in those without during the RF delivery (44±22ms vs. 17±11ms, p<0.01).

In this study, the earliest site of AMF from the LPVs was often determined to be around the carina region. These observations are likely to be consistent with the previous report [9]. In addition, the complex crossing of the muscular connections, bridges, neural inputs, and the adjoining muscle sleeves, possibly related to the MB conduction in the inter-PV carina, might promote electrical arrhythmogenicity including spontaneous ectopies of AF [10]. And also, the earliest activated site of the upper LPVs during CS pacing was highly observed around the carina region, and also a remarkable prolongation jump during the RF delivery was highly observed around the carina and/or adjacent anterior area. A previous report suggested that the distal exit of the MB into the upper LPV is commonly located around the inter-PV junction, possibly bypassing the LPV junction to the left atrium [34]. These specific muscle orientations and the dominant MB conduction toward the carina region could promote the preferential conduction properties.

In addition, the prolongation of the conduction time between the CS and LPVs during the RF delivery was significantly more commonly observed in patients with upper LPV AMF than in those without. The preferential properties of the MB connecting to the LPVs might involve cross talk that promotes an increased LPV arrhythmogenicity [3, 4, 11]. A larger amount of preserved MB muscle as a remnant of the LSVC, which is related to the conduction properties of the LPVs, may be crucial for determining the increased arrhythmogenicity of the LPVs.
