Oral Tongue Lesions Research Articles

Hardware Deployable Edge-AI Solution for Prescreening of Oral Tongue Lesions Using TinyML on Embedded Devices

Diagnosing oral cavity cancer (OCC) in its initial stages is an effective way to reduce patient mortality. However, current prescreening solutions are manual and the resultant clinical treatment is not cost effective for the average individual, primarily in developing nations. In this letter, we present an automated and inexpensive prescreening solution utilizing artificial intelligence (AI) deployed on embedded edge devices to detect benign and premalignant superficial oral tongue lesions. The proposed machine vision solution utilizes a clinically annotated photographic dataset of nine types of superficial oral tongue lesions to retrain a MobileNetV2 neural network using transfer learning. In this approach, we also utilized TensorFlow Lite for Microcontrollers to quantize a 32-bit floating point (float32) precision model into an 8-bit integer (int8) model for deployment on power and resource-constrained OpenMV Cam H7 Plus embedded edge device. The quantized int8 model was able to detect the nine tongue lesions with an accuracy of 98.69% on the test set. More than 60% reduction in on-device RAM and flash memory usage were measured for the int8 model when compared to the equivalently performing float32 model for relatively the same inference speed (~1.1 ms) on the target edge device.

Automated Detection of Oral Pre-Cancerous Tongue Lesions Using Deep Learning for Early Diagnosis of Oral Cavity Cancer

Abstract Discovering oral cavity cancer (OCC) at an early stage is an effective way to increase patient survival rate. However, current initial screening process is done manually and is expensive for the average individual, especially in developing countries worldwide. This problem is further compounded due to the lack of specialists in such areas. Automating the initial screening process using artificial intelligence (AI) to detect pre-cancerous lesions can prove to be an effective and inexpensive technique that would allow patients to be triaged accordingly to receive appropriate clinical management. In this study, we have applied and evaluated the efficacy of six deep convolutional neural network (DCNN) models using transfer learning, for identifying pre-cancerous tongue lesions directly using a small dataset of clinically annotated photographic images to diagnose early signs of OCC. DCNN models were able to differentiate between benign and pre-cancerous tongue lesions and were also able to distinguish between five types of tongue lesions, i.e. hairy tongue, fissured tongue, geographic tongue, strawberry tongue and oral hairy leukoplakia with high classification performances. Preliminary results using an (AI + Physician) ensemble model demonstrate that an automated pre-screening process of oral tongue lesions using DCNNs can achieve ‘near-human’ level classification performance for diagnosing early signs of OCC in patients.

Magnetic resonance imaging: A predictor of pathological tumor dimensions in carcinoma of anterior two-thirds of tongue - A prospective evaluation.

Preoperative imaging is mandatory for deciding the extent of surgery in tumors of oral tongue. Previous studies have shown the significance of depth of tumor invasion in predicting nodal involvement. This prospective study aimed to assess the correlation between tumor dimensions in all three planes obtained through preoperative imaging and histopathological findings, as well as the correlation between these and pathological node positivity. Fifty-nine consecutive patients with nonmetastatic, operable, squamous cell carcinoma of anterior two-thirds of the tongue were included in the study. Preoperative imaging findings were compared with pathological findings and analyzed. Histopathological dimensions were concordant with imaging findings. Anteroposterior, transverse, and craniocaudal (CC) dimensions obtained through imaging showed a significant correlation with corresponding pathological findings (0.730, 0.621, 0.810, respectively; P < 0.001). Among all three, only CC dimension showed a significant correlation with pathological nodal involvement (odds ratio [OR] = 7.875, P = 0.03, relative risk = 0.236). Pathological tumor thickness of >3 mm had a positive predictive value of 54.9% for nodal involvement (OR = 7.875, P = 0.03). With widespread availability of state-of-the-art magnetic resonance (MR) scanners, CC dimension needs to be emphasized as the most significant prognostic tumor parameter. Recent evidence, including our study, suggests that MR imaging is concordant with pathological findings, justifying its use in the pretreatment evaluation of oral tongue lesions.

Endoscopic Sentinel Lymph Node Biopsy in a Porcine Model

The objective of this study was to investigate the feasibility of endoscopic sentinel lymph node biopsy in a porcine model. One hundred microcuries of technetium-labeled sulfa colloid (Tc-SC) was injected into the right and left ventrolateral surfaces of the oral tongue of six adult Yorkshire pigs. A handheld gamma probe was used to locate the region of focal radioactivity on the neck that corresponded to the sentinel lymph node (SLN). Next, 0.25 mL of isosulfan blue dye was injected into the Tc-SC injection sites on the tongue. Endoscopic SLN dissection was then performed using a combination of balloon dissection and CO2 insufflation. The operative time, blood loss, and radioactivity of the SLN were measured for each animal. The SLN was detected transcutaneously with the gamma probe, and endoscopic SLN excision was successful. Endoscopic visualization and an endoscopic gamma probe confirmed the presence of both isosulfan blue dye and radiopositivity in the SLN in each pig. The procedure lasted 22 to 61 minutes (median duration, 35 minutes). There was no measurable blood loss in any of the animals. Mean radioactivity measured 14,466 counts/second per lymph node. Endoscopic SLN biopsy for oral tongue lesions is feasible and warrants further study.

Oral cancer at a Tasmanian tertiary referral clinic, The Royal Hobart Hospital 1996–2002

The Holman Clinic at the Royal Hobart Hospital includes a multi-disciplinary head and neck clinic which functions as a tertiary referral centre for Southern Tasmania and involves Ear Nose and Throat surgeons, Oral and Maxillofacial Surgeons, Plastic and Reconstructive Surgeons, Radiation Oncologists and Medical Oncologists. The aim of this study was to examine retrospectively the number, gender distribution, age, site of lesion, histology, mortality and treatment modalities of the oral cancers referred to the Holman clinic at the Royal Hobart Hospital. The medical histories and a database of the Holman clinic were used as the sources of data for this study. A total of 101 patients were treated for oral cancer in the Holman clinic at the Royal Hobart Hospital from 1996 to 2002. There were 64 males and 37 females. The distribution of anatomical sites of the oral cancers in this study was as follows: 36 oral tongue lesions, 17 floor of mouth, 13 lip, five retromolar trigone, five mandibular alveolus, six buccal mucosa, nine palatal and 10 minor and major salivary gland cancers. The most common site of oral cancer was the tongue (35.6 per cent), followed by the floor of mouth (16.8 per cent) and lip (12.9 per cent). The majority of oral cancers were squamous cell carcinoma, except for the salivary gland cancers. The incidence of squamous cell carcinoma was between 67 and 100 per cent, depending upon the site involved. The trends found in this study are similar to those previously documented over the past 20 years.

Dental factors in the genesis of squamous cell carcinoma of the oral cavity

This study investigated the role of dental factors in the genesis of squamous cell carcinomas of the oral cavity. 99 newly diagnosed patients with advanced head and neck cancer were evaluated for potential demographic, dental and prosthetic risk factors. Patients with extra-oral malignancies were used as controls in the analysis of dental factors in patients with intra-oral cancers. We performed multiple comparisons of these factors against the primary tumour sites. An analysis of dental and prosthetic factors and patients with intra-oral versus extra-oral tumours revealed no significant differences. However, all 28 intra-oral malignancies arose in areas in contact with teeth and/or appliances. For example, all 10 oral tongue lesions arose on the lateral borders and all six floor of (drop) mouth primary tumours developed at the flange extension of a lower denture. Patients with partial or full dentures had significantly lower node (N) scores than patients without a prosthesis. Although this paper suggests a relationship between dental factors and the genesis of oral squamous cell malignancies, a larger prospective study would be required to achieve definitive data. Patients at risk should be followed carefully for the development of chronic irritation from teeth and appliances.

Postoperative radiation therapy for squamous cell carcinomas of the oral cavity and oropharynx: Impact of therapy on patients with positive surgical margins

The presence of a positive or close margin after resection of a squamous cancer of the head and neck is associated with a significant risk of local recurrence. To determine the efficacy of postoperative radiation therapy for patients with advanced oral cavity and oropharyngeal cancers with inadequate margins of resection, the present retrospective analysis was undertaken. One hundred and two patients were treated with surgery and postoperative radiation therapy for advanced squamous cell carcinomas of the oral cavity and oropharynx. The anatomic subsites treated include oral tongue (n = 29), floor of mouth (n = 22), base of tongue (n = 31) and tonsillar fossa (n = 20). Twenty-five patients (25%) had positive margins, 41 patients (40%) had close margins (< or = 0.5 cm from the surgical margin) and 36 (35%) had negative margins. The median radiation dose was 6000 cGy. With a median follow-up of 7 years, the actuarial control rate for patients with positive, close and negative margins was 79%, 71%, and 79%, respectively. When postoperative doses of > or = 60 Gy were delivered to patients with positive/close margins (excluding patients with oral tongue lesions), the 7-year actuarial control was 92%. In similar patients receiving < 60 Gy, the actuarial control was 44% (p = 0.0007). Compared to other anatomic subsites, inferior control rates were obtained with oral tongue lesions. For this subsite, the control rates for positive, close, and negative margins were 50%, 62% and 69% respectively. We conclude that excellent local control can be achieved with postoperative radiation therapy, despite the presence of inadequate margins of resection, when doses of > or = 60 Gy are used. Future strategies must be directed at further improving these results in patients with oral tongue lesions.

Analysis of time-dose factors in squamous cell carcinoma of the oral tongue and floor of mouth treated with radiation therapy alone

This is a retrospective analysis of 147 patients with squamous cell carcinoma of the oral tongue and floor of mouth who were treated with radical radiation therapy at the University of Florida between October 1964 and December 1977. Control of the primary lesion was analyzed in respect to total dose, time, treatment modality, and split vs. continuous course therapy. There is an increase in local control of T2 lesions of the oral tongue with treatment by radium alone or radium and less than 3000 rad external beam irradiation as compared to radium and 3000 rad or more external beam (p = 0.02) and in T3 lesions of the floor of mouth with treatment by radium and 3000 rad or more external beam as compared to external beam alone (p = 0.11). Analysis of time-dose factors reveals an increase in local control with increasing dose over shorter periods of time in T3 lesions of the floor of mouth and T2–3 lesions of the oral tongue. Whereas dose appears to be important in all of these groups, overall treatment time appears to influence local control more in oral tongue lesions than in floor of mouth lesions.

Elective neck irradiation in squamous-cell carcinoma of the head and neck.

One hundred fifty-nine patients with squamous-cell carcinoma of the oral cavity, oropharynx, nasopharynx, hypopharynx, and supraglottic larynx, and with clinically negative neck (Stage NO) were reviewed to determine the value of elective neck irradiation (ENI). The primary cancer was controlled in 125 patients; in these patients the neck failure rate was 1.9% with ENI and 18% without ENI. Recurrence in the neck in cases where the primary cancer was controlled was associated with an increased risk of death due to cancer. The primary cancer was not controlled in 34 patients; in these patients the neck failure rate was 25% with ENI and 67% without ENI. Ten of 11 patients in whom treatment failed both at the primary site and in the neck died of cancer despite attempts at surgical salvage. ENI may be associated with an increased risk of local failure in oral tongue lesions as a result of a decreased proportion of treatment given with interstitial therapy.

Carcinoma of the oral tongue and floor of mouth: fifteen years' experience with linear acceleration therapy.

One hundred fourteen patients with carcinoma of the oral tongue and floor of mouth were treated with high-dose megavoltage radiation at Stanford University from 1956 to1970. Actuarial 5-year survival for 56 patients with oral tongue lesions was: T', 73; T', 37%; and T', 19%. Similarly, for 58 patients with lesions of the floor mouth, 5-year survival was: T', 73%; T', 37%; and T', 25%. Local control of the primary was obtained in oral tongue T' lesions 10 of 11 times; T', 5 of 8; and T', 13 of 36 times. For floor of mouth, local control was: T', 22 of 26;T', 7 of 14; andT', 3 of 15. There was an indication that better control was obtained if interstitialtherapy was a planned part of the treatment. Dose for local control when external radiation alone was utilized was usually over 1900 rets. Patients with initially clinicallynegative nodes (TXNO) who had a low radiation dose to primary echelon lymph nodes developed later cervical lymph node metastases 38% of the time. In no case did late metastatic disease appear in patients whose necks were treated prophylactically.