Capsule endoscopy: Do we still need it after 24 years of clinical use?

Abstract

In this letter, we comment on a recent article published in the World Journal of Gastroenterology by Xiao et al, where the authors aimed to use a deep learning model to automatically detect gastrointestinal lesions during capsule endoscopy (CE). CE was first presented in 2000 and was approved by the Food and Drug Administration in 2001. The indications of CE overlap with those of regular diagnostic endoscopy. However, in clinical practice, CE is usually used to detect lesions in areas inaccessible to standard endoscopies or in cases of bleeding that might be missed during conventional endoscopy. Since the emergence of CE, many physiological and technical challenges have been faced and addressed. In this letter, we summarize the current challenges and briefly mention the proposed methods to overcome these challenges to answer a central question: Do we still need CE?

Key Words: Capsule endoscopy; Wireless capsule endoscopy; Obscure gastrointestinal bleeding; Artificial intelligence in gastroenterology; Therapeutic capsule endoscopy

Core Tip: In this letter, we comment on a recent article published in the World Journal of Gastroenterology by Xiao et al, wherein the authors explored the use of a deep learning model to automatically detect gastrointestinal lesions during capsule endoscopy. We conclude that while capsule endoscopy remains a valuable tool in clinical practice, it carries many challenges that discourage its routine use. These challenges must be addressed in future studies to enhance its practicality and adoption by gastroenterologists.

TO THE EDITOR

We read with great interest the recent study by Xiao et al[1], published in the World Journal of Gastroenterology, where the author proposed a deep learning model to accurately detect different gastrointestinal lesions using capsule endoscopy (CE). The focus of this letter is to highlight the challenges that motivated the authors to conduct such a study and also to present the other promising solutions proposed in the literature to overcome the challenges and shortcomings of CE.

CE has emerged as a noninvasive modality for investigating and diagnosing small bowel diseases[2,3]. The first CE model in humans was proposed in 2000[2], and the first capsule model was approved by the Food and Drug Administration (FDA) in 2001[4]. This technology brings to mind the concept of “swallowing the surgeon,” proposed by Feynman[5] in his 1959 lecture “There’s Plenty of Room at the Bottom.” The swallowable capsule is propelled via peristalsis, requiring no external pushing force. It records continuously for over 5 hours, transmitting images to a portable recorder, allowing the patient to continue their daily routine[2]. In the era of fiber-optic endoscopy, CE represented a revolutionary advancement by overcoming two significant limitations of fiber-optic endoscopy: (1) The discomfort associated with using flexible, wide fiber-optic cables; and (2) the challenge of visualizing the entire gastrointestinal tract, from the oral cavity to the anal canal[2]. While recent advances in endoscopy devices have addressed many of these issues, the difficulty of visualizing the whole gastrointestinal mucosa remains a persistent challenge in clinical practice.

CE provides a modality for diagnosing esophageal and small bowel diseases due to the difficulty of diagnosis in this specific area via regular endoscopy (Table 1). Jain et al[6] classified the indications for CE into esophagus and small bowel applications. For the esophagus, CE could be used to diagnose Barret’s esophagus, esophagitis, and esophageal varices. However, more research is needed to assess the sensitivity of CE in detecting esophageal lesions, given its lack of therapeutic abilities compared to regular endoscopy. Double-balloon enteroscopy might not be available in low-resource settings endoscopy units. Also, the lack of training programs and the difficult learning curve for this procedure have driven the adoption of CE for detecting small bowel diseases. CE could be used to investigate obscure gastrointestinal bleeding, occult bleeding, iron deficiency anemia, recurrent gastrointestinal bleeding, malabsorptive syndromes (e.g., celiac disease), intermittent colitis, Crohn’s disease, graft-vs-host disease, small bowel tumors, and polyposis syndromes[4,7].

Table 1 Indications for capsule endoscopy.

Esophagus	Small intestine	Colon
Gastroesophageal reflux disease	Obscure gastrointestinal bleeding	Screening for polyps
Barrett’s esophagus	Occult bleeding	Monitoring IBD
Esophageal varices	Iron deficiency anemia	Incomplete colonoscopy
	Suspected Crohn’s disease	Refusing colonoscopy or unfit for anesthesia
	Surveillance of polyposis syndrome
	Evaluation of partial responding celiac

IBD: Inflammatory bowel disease.

Conventional approaches for diagnosing occult gastrointestinal bleeding, such as small bowel series and push enteroscopy, have demonstrated low diagnostic yields[8]. The advent of CE has improved the detection rates in these cases. Meta-analysis data comparing push enteroscopy with CE for the diagnosis of small-bowel diseases revealed that CE was superior, with a 35% to 40% incremental yield and a number needed to treat of 3[9,10]. When CE was compared with double-balloon endoscopy, CE had a higher incremental yield, with rates of 60% vs 57%[11]. CE was also reported to be superior to multislice computed tomography enterography in diagnosing obscure gastrointestinal bleeding, especially in patients with overt bleeding and in individuals younger than 40 years of age[12].

CE has some contraindications that can be classified into absolute and relative. Jain et al[6] reported some absolute contraindications of CE, including any cause of intestinal obstruction and extensive and active Crohn’s disease, which might subsequently lead to fistulas and strictures. These contraindications require a patency test before the procedure, which involves swallowing a dissolvable radiofrequency identification tag that typically dissolves within 30 hours. The tag’s passage can be monitored using imaging modalities to ensure bowel patency. Other debatable contraindications include intestinal pseudo-obstruction and young children. The challenge with young children lies in their ability to swallow the capsule, and the passage of the capsule through their narrow gastrointestinal tract[13]. The use of CE was approved by the FDA for children aged 2 years and older; however, cases where CE was successfully used in infants as young as 8-months-old have been reported[14]. However, the Pillcam manufacturer recommends its use only in children aged 8 years and above[15]. Oikawa-Kawamoto et al[16] reported that CE is safe in infants and young children unable to swallow the capsule if it is delivered directly into the duodenum endoscopically.

Relative contraindications for CE include dysphagia, previous abdominal surgery, pregnancy, diverticulosis, and the presence of cardiac pacemakers or other implanted electro-medical devices[6].

Based on the previously mentioned contraindications, we can understand some of the physiological challenges and pitfalls of CE involving the interactions between the body and the capsule. The most common complications associated with CE are capsule retention[17], difficulty swallowing or aspiration[17,18], incomplete examination, and suboptimal diagnostic results[17,19]. While the risk of these challenges could be minimized using the pre-examination patency test, it cannot be completely eliminated. In addition to physiological challenges, CE also carries some technical limitations involving the capsule itself. These include time-consuming recording and result interpretation, gaps in the recording, short duration of the battery, inability to download images, peristalsis-dependent motion impeding active control of orientation and speed of the capsule, and the inability to interact actively with lesions via obtaining biopsies or delivering therapies[17,20].

These challenges hinder the diagnostic efficacy of CE; thus, efforts have been made to overcome them (Figure 1 and Table 2). Recently, more emphasis has been laid on developing physician-driven capsules, categorized into two types: Those using an internal locomotion mechanism and those using an external locomotion system[21,22]. External locomotion systems using magnetic power are more practical than internal locomotion systems, which rely on mechanisms such as vibratory actuators for self-propulsion, pedundulatory locomotion, legged locomotion, or even swimming capsules with propellers[23]. Internal locomotion mechanisms reduce battery life and increase capsule size beyond the standard, limiting their usability[23]. In contrast, external locomotion systems using magnetic approaches are not limited to batteries and allow the physician to steer the capsule as needed. Many magnetic approaches have been reported. Liao et al[24] reported a comparable diagnostic accuracy between magnetically guided CE and esophagogastroduodenoscopy, with more than 95% of patients preferring the painless, non-invasive magnetically guided capsules.

Figure 1 Graphical abstract for the new technologies of capsule endoscopy. 2D: Two-dimensional; 3D: Three-dimensional; AI: Artificial intelligence; IBC: Intrabody communication; MCCE: Magnetically controlled capsule endoscopy; RF: Radio frequency; UWB: Ultra-wideband; WPT: Wireless power transfer.

Table 2 Pitfalls and technological solutions for capsule endoscopy[45-71].

Pitfall	Solution	Advances	Ref.
Incomplete examination owing to battery runout	Near-field WPT	Enables charging of the capsule during maneuver to enable complex procedures and good visualization with high-quality photos and videos without battery runout	Miah et al[45]; Zhuang et al[46]; Meng et al[47]; Basar et al[48]
	Power management strategies	AI-based technologies to allow rational consumption of the battery and adaptive lighting that further allow good visualization	Hale et al[49]
	UWB/IBC	UWB is a wireless communication technology that operates over a wide range of frequencies to allow lower power consumption while IBC Intrabody communication allows for the delivery of information through the body tissue to not rely solely on wireless communication to allow reduced signal loss, lower power requirement, and improved reliability as it is more stable	Basar et al[50]; Jung et al[51]; Jung et al[52]; Shang and Yu[53]; Li and Guo[54]; Hafezi et al[55]; Lamanna et al[56]
Retention of the capsule	Magnetically controlled capsule MCCE	This revolutionary technology enables external control of the capsule, allowing its maneuver for better visualization. It also supports the development of features that could make CE as effective in diagnosis and therapy as fibro-optic endoscopies. MCCE offers precise, controllable propulsion through the body	Yim et al[27]; Xiao et al[57]; Park et al[58]; Leon-rodriguez et al[59]; Nguyen et al[60]; Guo et al[61]; Hua et al[62]; Hoang et al[63]
	Legged locomotion capsule	These capsules equipped with an internal locomotion system are desirable as they offer enhanced diagnostic capabilities and therapeutic advantages. They do not rely solely on peristaltic movement	Hua et al[62]; Quirini et al[64]
	Paddling-based capsule endoscope	This is another type of internal locomotive capsule that uses the padding technique instead of legging	Kim et al[65]
Limited therapeutic use and inability to take biopsies	Magnetically controlled capsule MCCE	Integrated micro biopsy device, multi-point targeted liquid sampling, Passive and active drug delivery control	Yim et al[27]; Xiao et al[57]; Park et al[58]; Leon-rodriguez et al[59]; Nguyen et al[60]; Guo et al[61]; Hua et al[62]; Hoang et al[63]
Image quality	Multi-element lenses and adaptive illumination	These superior-quality lenses allow a wider angle of view, and the adaptive illumination aids in picture clarity and enhances battery consumption
	CapsoCam SV1	These cams have four-side viewing, allowing for a 360° panoramic view to improve mucosal visualization
	3D imaging reconstruction	to add more comprehensive surface topography using a software-enabled technique to convert 2D images to 3D images
	UWB/IBC	UWB in conjunction with IBC allows for better and more reliable data transmission to permit sending of high-quality photos and videos	Basar et al[50]; Jung et al[51]; Jung et al[52]; Shang and Yu[53]; Li and Guo[54]; Hafezi et al[55]; Lamanna et al[56]
Missed lesions	MCCE, legged locomotion capsules, and paddling-based capsules	The external and internal control by these methods allow for better controllable movement of the capsule without missing lesions by peristalsis	Yim et al[27]; Xiao et al[57]; Park et al[58]; Leon-rodriguez et al[59]; Nguyen et al[60]; Guo et al[61]; Hua et al[62]; Hoang et al[63]; Kim et al[65]
Does not provide real-time feedback	UWB/IBC		Basar et al[50]; Jung et al[51]; Jung et al[52]; Shang and Yu[53]; Li and Guo[54]; Hafezi et al[55]; Lamanna et al[56]
Difficulty and time of interpreting lots of images by the physician	AI-based autonomous lesion detection	Machine learning algorithms to allow for easier analysis of the large number of photos and videos retrieved from the capsule and aid in objective diagnosis	Hale et al[49]; Sharma et al[66]; Hajabdollahi et al[67]; Rustam et al[68]; Alaskar et al[69]
Location problems	Hybrid RF with vision-aware fusion scheme	Multi-sensor approach of the capsule by both RF in addition to vision-based and magnetic type are used simultaneously in the capsule to aid in its locating capabilities instead of relying on only one of them. This problem emerged because of the hard localization of the capsule in the small bowel owing to its length and its compact structure	Vedaei and Wahid[70]; Narmatha et al[71]

AI: Artificial intelligence; 2D: Two-dimensional; 3D: Three-dimensional; IBC: Intrabody communication; MCCE: Magnetically controlled capsule endoscopy; RF: Radio frequency; UWB: Ultra-wideband; WPT: Wireless power transfer.

Regarding the limited ability of CE to take biopsies, several studies have reported advancements in manufacturing procedures that equip CE with tools such as biopsy needles, scrapers, and micro forceps, thus empowering it to take biopsies[25,26]. Yim et al[27] developed micro-jaw forceps and two multiscale magnetic-based robotic devices: a centimeter-scaled untethered magnetically actuated soft capsule endoscope and a submillimeter-scale self-folding-micro-gripper.

Several research groups have developed CE devices capable of performing some therapeutic hemostatic interventions, such as clipping or balloon tamponing[28]. For instance, Leung et al[29] developed a capsule for the treatment of gastrointestinal hemorrhage using a balloon tamponade effect. Another novel CE device is the “Enterion” capsule, which delivers drug formulations to specific regions of the gastrointestinal tract[30].

A single CE examination generates a video containing around 60000 frames, requiring an average of 30-120 minutes for interpretation, depending on the physician’s skill and experience[4,31]. Physicians must meticulously review this large number of frames to identify abnormalities, which are usually seen in only one or two frames and could be easily missed owing to the limited human reading ability. As a result, CE has a significant miss rate: 18.9% for neoplasms, 5.9% for vascular lesions, and 0.5% for ulcers[32].

The incorporation of artificial intelligence (AI) into CE models has reduced the interpretation time, increased the accuracy of abnormality detection, and reduced human error[31,33]. AI has been developed to detect small bowel abnormalities in CE since 2000. Several AI algorithms, such as “express view” and “suspected blood indicator,” are currently incorporated into CE models[34,35]. Studies have shown the good performance of AI in detecting ulcers[36], celiac disease[37,38], polyps/tumors[39,40], hookworms[41], small bowel angiodysplasia[42], and bleeding[43,44]. In the recent study by Xiao et al[1], AI features were employed to assist gastroenterologists in the early detection of 23 different types of gastrointestinal lesions using CE. This approach aims to increase the diagnostic yield and efficacy of CE and also save significant time for gastroenterologists. The multicategory lesion detection model described by Xiao et al[1] is an optimized version of the YOLOv8 model. Thorough ablation experiments were conducted, and the most efficient model was made of P4, a bidirectional feature pyramid network, and the Swin transformer, yielding a mean average precision of 91.5, Giga floating-point operations per second of 203.6, and a frame rate of 129.70 frames per second. Compared to earlier versions such as YOLOv5, YOLOv6, and YOLOv7, Xiao et al’s model[1] significantly improved the mean average precision scores. Moreover, their model demonstrated superior accuracy and frame rates compared to alternative methods such as the single-shot detector, faster recurrent convolutional neural networks, and real-time detection transformers[1].

CONCLUSION

In our opinion, CE remains an important tool for the diagnosis of various gastrointestinal disorders, especially in cases of bleeding in areas difficult to access using regular endoscopy. However, the current challenges faced by both doctors and patients present considerable obstacles. These include the high rates of missed findings, the time-consuming interpretation procedure, the lack of therapeutic abilities, and the other previously mentioned contraindications and complications. As a result, gastroenterologists may be hesitant to order CE in certain cases. Thus, we believe that future research and development are needed to address these limitations and enhance the utility of CE. This could be achieved by using the benefits of modern technologies, such as advanced AI models, to detect more lesions and by developing endoscopist-controlled therapeutic tools. These advancements may help realize the dream of a “capsule endoscopist”, combining diagnostic and therapeutic capabilities in a single, minimally invasive device.