informazione

Pressure injury image analysis with machine learning techniques: A systematic review on previous and possible future methods

Pressure injuries represent a tremendous healthcare challenge in many nations. Elderly and disabled people are the most affected by this fast growing disease. Hence, an accurate diagnosis of pressure injuries is paramount for efficient treatment. The characteristics of these wounds are crucial indicators for the progress of the healing. While invasive methods to retrieve information are not only painful to the patients but may also increase the risk of infections, non-invasive techniques by means of imaging systems provide a better monitoring of the wound healing processes without causing any harm to the patients. These systems should include an accurate segmentation of the wound, the classification of its tissue types, the metrics including the diameter, area and volume, as well as the healing evaluation. Therefore, the aim of this survey is to provide the reader with an overview of imaging techniques for the analysis and monitoring of pressure injuries as an aid to their diagnosis, and proof of the efficiency of Deep Learning to overcome this problem and even outperform the previous methods. In this paper, 114 out of 199 papers retrieved from 8 databases have been analyzed, including also contributions on chronic wounds and skin lesions.

Introduction

The average life expectancy of world population has steadily increased during the last decades, and it is expected to continue growing during the next century. At the same time, medical professionals have witnessed an increasing incidence rate of diseases related to sedentary lifestyles and unhealthy eating habits, such as diabetes, especially in the Western hemisphere [12]. The convergence of these two trends has increased the number of long time care patients subject to lengthy periods of immobilization who suffer pressure injuries: a type of chronic wounds resulting from damage caused by pressure over time causing an ischemia of underlying skin structures. These injuries appear most commonly at bony prominences like the sacral area and the heel [10]. Some factors which contribute to pressure injury formation are: skin contact with a bed or a chair without frequent position changes, contact with urine or stool, diseases like diabetes that affect blood flow, injuries which restrict your body positioning and your nutritional status or medications. There are many other factors than may put the skin at risk for pressure injuries, as presented in Fig. 1.

Despite all the prevention measures that are put in place, pressure injuries may still develop. More than 2.5 million people in the United States develop pressure injuries [27]. And the risk of developing them during hospital stays is 3 times greater than the risk of being involved in a car accident. Healthy skin, muscle or fat tissue starts to die when blood flow is slowed or interrupted by pressure. This pressure can be caused by the bones pushing down on a surface like a chair or a bed.

Pressure injury stages are defined based on the deepest parts of the ulcer and the type of tissue affected. Higher stage ulcer represents deeper tissue damage and more serious injury. There are four main stages of pressure injuries [30]:

  • Stage 1: Pressure injury often presents as intact skin with redness, and is usually over a bony area. The skin would stay red after the pressure is removed. This is because the first area affected by pressure is muscle tissue. In fact, due to its aerobic metabolism, muscle tissue requires more oxygen. Hence, when blood circulation is interrupted, lesions begin appearing in deep muscle levels before reaching the skin. The surface area may be painful, warm or softer compared to surrounding tissues.

  • Stage 2: Pressure injury presents as an area where the top layer of the skin is missing with the open area being pink or red.

  • Stage 3: Pressure injury is deeper and more severe than a stage 1 or 2 pressure injury. It goes into fat tissue, and dead tissue may also be present.

  • Stage 4: Pressure injury is the most severe form of this wound. It is the deepest ulcer possible and would reach muscle tissue and possibly bones. Dead tissue may also be present with stage 4 pressure injuries. Fig. 2 depicts the main tissue types that can be present in stage 3 and 4 pressure injuries.

There are also two additional stages of pressure injuries. One is called unstageable because there are such exceedingly dead tissues in the wound that physicians cannot examine its depth. The other type of pressure injuries is called deep tissue pressure injury (DTPI). It usually begins with skin that is not broken, and it is usually purple or dark purple in color. This type of ulcer can open quickly and become a deep pressure injury [29], as it appears in Fig. 3.

Pressure injuries can occur in different areas, including: back or sides of the head, rims of the ears, shoulders, hipbones lower back, backs or sides of the knees, heels, ankles and toes. The age of the patient, coupled with certain medical conditions such as diabetes, slows down the healing of pressure injuries, which often are not healed before the death of the patient [3], and represent a significant economic cost for medical care services [21]. Pressure injuries are both painful and prone to infection, which calls for a continuous monitoring of their evolution by medical staff.

Documenting the pressure injury once the patient is admitted to the hospital is the first essential step in the process of prevention and assessment. Firstly, caregivers need to check if the patient has already an existing pressure injury. In such case, documenting the stage, the size, the color, the drainage and other characteristics is the first step towards a properly done management of the pressure injury. In case the state of the wound is severe, wound care team is to be contacted to prescribe a personalized treatment. This includes reducing pressure and friction by repositioning every one to two hours and using support surfaces to protect sensitive skin and minimize shear, cleaning and dressing the wound depending on its stage, and removing damaged tissue. The latter intervention is crucial to keep the wound free of infected or dead tissue. Depending on the state and stage of the pressure injury, the treatment may differ. When the wound is in its first stages, a gentle cleaning and proper dressing is required to prevent it from getting infected. When the pressure injury is in stage 3 or 4, it may take several months to heal, or may never do, especially if other risk factors are permanently present, such as: poor blood circulation and vascular diseases, diabetes, poor nutrition, immobility. Hence, a user friendly system for a non-intrusive assessment of pressure injuries would represent a valuable tool for frequent assessment and monitoring of these wounds.

To avoid the discomfort of patients in the process, non-invasive wound monitoring techniques are preferred. In this sense, imaging techniques are called to play a key role, as they allow an accurate analysis of its features with no need of contact with the wound itself.

For this reason, this work aims at providing the reader with a comprehensive review of state-of-the-art imaging techniques applied on pressure injury image analysis, and the possibility for deep learning to outperform these techniques.

This study represents a systematic review which gathers the contributions of many studies on the assessment of skin wounds using image processing techniques. Since the number of papers dealing with pressure injuries in particular was limited, we have extended our research on skin wound analysis in general, including: chronic wounds, lower extremity wounds, diabetic ulcers, malignant ulcers, venous ulcers and skin lesions. The studies of skin wounds using image processing have skyrocketed during the beginning of the 21st century. Hence, the majority of the presented contributions were published during the last decade.

The remainder of this paper is organized as follows: Section 2 presents the methodology of paper selection (search terms, inclusion/exclusion criteria, statistical data about the number of papers per year, country and research problem). Then, different categories in pressure injury assessment using image processing are discussed in Section 3, including: wound segmentation, wound measurement, tissue classification and healing prediction. As mentioned previously, other types of skin wounds in the selected papers were addressed to broaden the techniques used for wound imaging analysis. Next, Section 4 outlines Deep Learning techniques used in biomedical image processing. Finally, in Section 5 we discuss the usability of Deep Learning to efficiently assess pressure injuries using image processing. Last but not least, the reader interested in the broader field of non-invasive chronic wound monitoring via imaging techniques is referred to the recent survey by Mukherjee et al. [15].

Section snippets

Searched databases

The literature search was performed using several databases, namely: Web of Science, SpringerLink, IEEE Xplore, PubMed, Science Direct, the Scientific Electronic Library Online (SciELO), SPIE Digital Library and Wiley Online Library.

Search terms

In order to look for papers regarding wound analysis using image processing techniques, different search terms were used, and some of them were combined in the same search. Furthermore, the terminology of pressure ulcer was changed to pressure injury in April 2016

Results and discussion

In the literature, four main problems are addressed in pressure injury image analysis, as depicted in Fig. 4:

  • 1.

    Segmentation

  • 2.

    Measurement

  • 3.

    Tissue classification

  • 4.

    Healing evaluation

Deep Learning techniques as a novel approach for wound image analysis

The previous work in pressure injury image analysis was mostly handled using classical machine learning techniques. However, in the last decade, a subfield of artificial neural networks (ANNs) in machine learning known as deep learning (DL) have gained popularity in many fields of applications such as: object detection and recognition, acoustic modeling, drug discovery and toxicology, speech recognition, language modeling, bioinformatics and many other research topics. Deep neural networks are

Discussion

From the papers analyzed in this survey, it became increasingly apparent that non-invasive solutions for wound monitoring are highly required. Such systems should verify several conditions such as: an efficient detection of the wound whatever the clinical capturing conditions are (lightning, angle and distance), an accurate distinction between the wound and the rest of the image with all possible objects included in the image (bed, bed sheet, a ruler or color panel and a person’s hand or arm),

Conclusion

The increasing prevalence and the associated economic cost of pressure injury care, coupled with the negative impact of this type of wounds on the patient’s quality of life calls for the development of non-invasive tools and techniques that help in the design of more effective healing protocols. In this sense, image processing techniques have proven to provide suitable solutions towards this aim.

Hence, this survey paper provides global knowledge on the previous contributions to wound image…

[Tratto da: www.sciencedirect.com ]

admin

PROGETTO ITALIANO IN WOUND CARE 🇮🇹. SITO UFFICIALE DI LESIONI TOUR ®️. DAL 2017 Premio Eccellenze Italiane Assotutela 2020-2021.

Lascia un commento

Il tuo indirizzo email non sarà pubblicato. I campi obbligatori sono contrassegnati *

Questo sito usa Akismet per ridurre lo spam. Scopri come i tuoi dati vengono elaborati.